11.19.2007

The QuadFX Disaster Part Two

AMD today released its much awaited quad core desktop, the Phenom. After reading the reviews I just felt a sense of deja vu. Almost a year ago when AMD released the QuadFX to compete against Intel's newly released quad-cores, here is what some had to say:

Ars Technica (2006):
"Today, AMD officially launched their much-anticipated 4x4 enthusiast platform under the name QuadFX. The results of the rash of QuadFX reviews that just came out are uniformly disappointing; Intel's quad-core offering, the QX6800, outperforms AMD's much hotter, more power-hungry offering in almost every benchmark".

Anandtech (2006):
"When only running one or two CPU intensive threads, Quad FX ends up being slower than an identically clocked dual core system, and when running more threads it's no faster than Intel's Core 2 Extreme QX6700. But it's more expensive than the alternatives and consumes as much power as both, combined... Until then, there's always Quad FX but you're better off with Kentsfield".

Then again a year later while trying to take another stab at the "enthusiast" market:

Ars Technica (2007)
"Current benchmark results from Anandtech, HardOCP, and Hexus indicate that the Phenom, while notably more efficient than Athlon 64 X2 in certain scenarios, still lags the Q6600 clock-for-clock. Phenom may have finally given AMD the ability to offer a quad-core processor to compete with Q6600, but Intel's chip still holds a better price/performance ratio".

Anandtech (2007)
Phenom is, clock for clock, slower than Core 2 and the chips aren't yet yielding well enough to boost clock speeds above what Intel is capable of... today's launch confirms that Intel is still the king of the quad-core market.

Hexus (2007)
We can debate all day whether the majority of consumer software is threaded enough to take advantage of four execution cores, but the immutable fact remains that AMD's fastest quad-core offering is slower than Intel's slowest. Compounding this depressing statement for AMD is the January 2008 launch of Penryn-based Core 2 Quads, furthering Intel's performance dominance.

In order to grasp the magnitude of how much AMD messed up on execution, it's important to realise that even after a year of humiliating itself with the QuadFX disaster it has yet to beat it's old nemesis the QX6700. Intel was kind enough to lower it's quad-core offering to QX6600 but it seems AMD's Phenom can't beat that either. Never mind taking back the performance crown or beating the opposition on some important benchmark, but failing to beat a year old processor is simply inexcusable. This is indeed a wake up call and should desensitise anyone who's heard far too many excuses from AMD.

113 comments:

Anonymous said...

oh damn
the fabs 'bout to close in a minute

Tonus said...

As I understand it, early next year Intel is planning to add a 45nm 2.66 quad-core processor in the $166 range, and eventually drop it to $133.

That could be a case of Intel smelling blood in the water and moving to strike. If they can produce a quad-core that performs as well as most of AMD's line and costs ~$100 less it will force AMD to drop prices both quickly and drastically.

I doubt they will BK, but I think Intel intends for them to continue to hurt financially as long as possible.

Andy said...

And then there is Nehalem. Unless AMD is skipping a uarch along with a process node they are going to have to pull some major performance from K10 in the future. I say skipping, they claim 45nm in a few months less than 18 months after 65nm so why not dump k10 early for a new uarch?

GutterRat said...

Chernobyl
The Exxon Valdez
The Hindenburg
Bhopal
The Titanic
"rev 10h"

Great Disasters

Anonymous said...

Where is that retard abinstein now? Is he hiding under Scientia?

abinstein, how do you like getting your ass kicked?

Blabbermouth.

Anonymous said...

Seems like AMD won't be regaining any lost ground in the Top 500 supercomputer list anytime soon.
Intel EM64T triples share

Anonymous said...

From page 4 of Anandtech's review:

Currently, the L3 cache/NB on these chips runs at a fixed frequency that's actually lower than the rest of the CPU frequency: 2.0GHz. We tested Phenoms running from 2.2GHz all the way up to 2.6GHz, and in all cases the L3 cache and North Bridge ran at 2.0GHz. We're not sure if this will ever get fixed, but it's somewhat disappointing as it was supposed to be a major reason for upgrading to Socket-AM2+ (but it's good news for current AM2 owners).

one point of argument for AMD supporters at the launch of Barcelona was that the K10 will scale much better than Core2 (some even went so far as proclaiming superlinear performance gains) as frequency increases due to the L3/IMC. looks like that is not going to be the case either.

JumpingJack said...

Anand is not clear on how the architecture is put together. The L3 clocks independently of all the cores, it must because AMD is independently clock each core for power reasons, but sharing L3. As such, the L3 must remain fixed and the logical place to fix it is with respect to the IMC.

Kanter's K10 article best summs it up...

http://www.realworldtech.com/page.cfm?ArticleID=RWT051607033728

"The frequency for each core is independent of both the other cores, and the various non-core regions; the voltage for all four cores is shared, but separate from the non-core. As a result, power can be aggressively managed by lowering frequency and voltage whenever possible. To support independent clocking and modular design, asynchronous dynamic FIFO buffers are used to communicate between different cores and the northbridge/L3 cache. These FIFOs absorb any global skew or clock rate variation, but the latency for passing through depends on the skew and frequency variance – which is why the L3 cache latency is variable. "

This is part of the complexity of this design, and the addage that the more complicated the pipes the easier it is to clog up the works comes to mind.

Anonymous said...

that is why you don't make unrealistic claims and claims without backup. In the end, if you're wrong, you're just going to look like an idiot.
Scientia and Abinstein are walking examples of that.

Abinstein thought Phenom/K10 was the next coming of Jesus....

Now they are both in the hiding, neither have posted anything, anywhere.

SPARKS said...

Yeah, Yeah, Yeah, it's frig'in dog, big surprise there.

Ok, let’s get serious, real serious.

Hey, GIANT, are you paying attention?

Look how nicely INTC has pee’d on Pheromones parade. INTC slipped this one quietly to Anand on the same day AMD lifted the embargo! LOL

http://www.anandtech.com
/cpuchipsets/intel/showdoc.aspx?i=3154

This is the bad boy I’ve been waiting for, and it’s a monster!

Honey, fire up the credit cards, I’m in.

SPARKS

Giant said...

Wow! What a monster of a CPU there sparks!

PC Perspective tests the awesome new QX9770 as well!

http://www.pcper.com/article.php?aid=484

What was their impression of this beast?!

The Intel QX9770 does what Intel wanted it do though: it embarrasses AMD by showing how much untapped potential the Core 2 design really has in it and that it will take a series of miracles for AMD to catch up with Intel anytime soon.


What a MONSTER! 3.2GHz clockspeed, 1600mhz FSB with dual channel DDR3 running at 1600mhz!

But then just think, in late 2008 -> early 2009 Intel will release quad core and octa core Nehalem! QuickPath with triple channel DDR3 1600mhz!

I thought my Q6600 was a Houdini rendering monster. Yorkfield cranks it up a huge notch and Nehalem takes it to another whole new level! In January it's Yorkfield for me. Then in late '08 quite probably a Nehalem machine!

Meanwhile, AMD is still trying to release a CPU as fast as Intel's more than a year old QX6700!

Giant said...

Abinstein thought Phenom/K10 was the next coming of Jesus....

I do recall that. He kept quoted his 'source' to support his wild claims that Pheromones has a higher IPC that Kentsfield or Yorkfield.

Anonymous said...

"I do recall that. He kept quoted his 'source' to support his wild claims that Pheromones has a higher IPC that Kentsfield or Yorkfield."

:) Well, wait until he becomes active again... it will be another excuse... "The TLB Errata bug is holding back performance" or "Intel did not disclose the SSE4 before it was too late and AMD could not get it into Agena" or "The cow was tired and could not jump over the moon" ... ;)

Anonymous said...

giant wrote:
"I do recall that. He kept quoted his 'source' to support his wild claims that Pheromones has a higher IPC that Kentsfield or Yorkfield."

What was his source, Rahul Sood?

http://www.rahulsood.com/2007/08/benchmarks-are-wiggedy-wiggedy-whack.html

"And for the record, if you were to benchmark Phenom at 3GHz you would see that it kicks the living crap out of any current AMD or Intel processor—it is a stone cold killer"

Hahahaha! Oh dear.

Anonymous said...

I think folks should re-read JJ's last post (and Lex's post on Scientia's before it gets deleted).

Simplicity is what you need to strive for in design. There is no doubt IMC, HT, independent clocking are all great design features but the problem is that's all they are...design features. If you can't implement these well into a manufacturable solution, then really what is the point? The world is littered with great designs and innovations that have failed miserably due to the fact they could not be manufactured.

I have no doubt K10 down the line will prove to be a good design - with good power capability and good scalability. The clear problem is it is not quite manufacturable on AMD's 65nm yet (don't think for a second the issues are solely due to this TLB errata crap being floated around the net). So now the problem is AMD has gone too far down the road (in their mind) and will likely try to gut it out until the process capability/manufacturing hopefully catches up to the design.


I keep hear these wonderful stories about when Barcelona came out, it reminded them of K8 (or was it K7?) in that it started off poorly but then proved to be a good design over time, and that people need to just give it time.

My though to this was EXACTLY - why the @%^#! did AMD not learn from that and try to marry the design with the manufacturing capability. If it will be better 6 months or a year down the line then they should have waited! Or targeted a simpler design that matches their current capabilities.

AMD needs to take a long hard look at their process capability and manufacturing expertise and ask themselves some hard questions"

1) Can the process/manufacturing keep up with the design team? (Intel had the same issue until they forced the designers to work more closely with the process development folk) Someone in the company needs to tell the designers that they can't design in a box - it doesn't matter how good the design is if you can't mass-produce it.

2) Is trying to catch up to Intel on technology nodes helping or hurting AMD? Are they implementing costly and sophisticated solutions (like SOI and immersion) just to stay on the same pace as Intel and is this costing them in terms of manufacturability? This is a very difficult tradeoff - is it better to slow down the tech node scaling (which hurts a bit from a cost perspective in terms of die size), but will it lead to better process performance (which may mean better ASP's) or better yield and/or binsplits. This would definitely require some of the technologists in the AMD organization swallowing their pride and realizing that even with IBM, they may not be able to keep up with Intel in this area.

I think the first test case will be 45nm - does it really make sense to move to this node without high K? In my view 65nm was not ready and it was rushed out the door - this was apparent when you looked at how K8 ported from 90nm to 65nm.

Time will tell but the benefits of the smaller die may be offset by cost increases with immersion litho and difficulty with binsplits due to lack of gate oxide scaling.

JumpingJack said...

"(don't think for a second the issues are solely due to this TLB errata crap being floated around the net). "

I tend to agree... I think this TLB crap is just a smoke screen (my opinion, no proof, just a theory).... actually, it could be the TLB affected, but it is not a logic bug, it is a physical bug much like the cold bug is not a logic bug. Let me explain my rationalization.

Remember AMD recalled about 3000 opeterons because of a 'test escape' (this was real and this is real jargon in the industry, it was a true escape). The escape was an overheating section (hot spot) on the chip when under load running one particular instruction consecutively over and over.

Well, my guess (and again, it is just a theory guess) is that AMD has a hot spot that is shutting down or affecting the circuit around the TLB (perhaps). This would explain why only 2.4 GHz and higher are affected, and why some CPUs are good and some are bad (die to die, wafer to wafer variation in the processing). It also explains why a few will OC well and other crap out before 2.5 GHz (as we see in some reviews).

Anyway, this is my guess it could be wrong.

Jack

Tonus said...

In fairness to Rahul Sood, his article was from mid-August, and I am not sure what the fastest Intel CPU was at the time. 2.66GHz Core2Quad? 2.4GHz perhaps? A 3GHz Phenom would beat a 2.4/2.6Ghz C2Q, wouldn't it?

The problem is that they took so long to get the Phenom to market and when they finally did, they are only running at 2.2/2.3GHz. And Intel has already introduced one new architecture (Penryn) and is readying another one (Yorkfield) for early 2008.

It is just another reminder that what matters is how well your shipping products perform. A 3GHz Phenom may well have been the fastest desktop CPU in August of 2007. That doesn't help me if I cannot buy a 3GHz Phenom until August 2008, or whenever they finally begin to offer them for sale.

Anonymous said...

When was the last time the Intel-AMD performance delta was this big?

Ho Ho said...

tonus
"2.66GHz Core2Quad?"

Actually QX6800 at 2.83GHz was launched at 9'th April so there is no excuse to his ramblings.

My guess is he simply talked to someone from AMD and too their every word as gospel.

Ho Ho said...

Whoops, that was actually 2.93GHz

Tom said...

Great blog and comments..

I only have one thing to add - does ANYONE, ANYWHERE REALLY believe what AMD's says anymore?

My wife works at Intel and she even says a healthy, niche, AMD is a driving force for Intel to innovate.

However, when AMD talks about 2.4 errata and 2.6, 2.8, 3.0 GHZ promised for this year/next year, how can even the Scientia's/Abinstein continue to be AMD fans and look themselves in the mirror everyday?

This is beyond ridiculous.

Anonymous said...

"how can even the Scientia's/Abinstein continue to be AMD fans and look themselves in the mirror everyday?"

fans ---> short for fanatical --> fanatical is not generally associated with logic and reason.

Fans appear rationale when things are going well (the same can be said about Intel fans too), however it is when things are not going well the tiger finally shows its stripes.

It's at that point where things become desparate and you have blogs like "K10: A Good start" (Oh this was meant to be serious!?!?) or a blog on process technology when the author has nary a clue of what Si is.

Anonymous said...

I am still waiting to hear how Scientia will spin this Phenom launch.

Let's see if Scientia sticks to his timeline from his "2008 and Beyond" post from August where he thinks AMD will close the manufacturing gap with intel in the 32nm node:

Q2 06 - Woodcrest
Q3 07 – Barcelona Trailing by 5 quarters.

Q4 07 - Penryn
Q3 08 – Shanghai Trailing by 3 quarters.

Q4 08 - Nehalem
Q2 09 – Bulldozer Trailing by 2 quarters.

Q4 09 - Westmere
Q1 10 - 32nm Bulldozer Trailing by 1 quarter.

SPARKS said...

Um, just as a point of interest, I believe Pheromones belchmarks have reached the financial street, and it ain’t pretty.

If you’re keeping score, AMD closed at 11.28, shedding 79 cents. This is basically a 4 year low. Although not as bad Feb 2003, a little over 5 dollars a share, back then, they didn’t have the massive debt and a failing product line up.

I’ll probably get beat up for saying this (I enjoy the abuse from the financial people), but I’ll say it again. The time is coming when the big institutional brokers are going to unload AMD, cut their losses, and write it off as a 2007 LOSS. This will not be pretty. Today’s 7 percent drop is telling me someone’s getting out, and getting out BIG.


http://moneycentral.msn.com/
stock_quote?Symbol=amd


SPARKS

Anonymous said...

abinstein is a coward. Now making excuses on AMDZOne

http://www.amdzone.com/index.php/Forums?func=view&id=154&catid=6

Anonymous said...

Whaaaa-at?? I just read on the "Joke of Perspiring 2-bit Computing" where Phenoms frag all glued-up P3's by 40% across a wide range of tasks. Surely Sharikou doesn't lie??

Anonymous said...

"I believe Pheromones belchmarks have reached the financial street, and it ain’t pretty."

Well I think people are digesting the 700Mil cash infusion as relatively meaningless (this theoretically could have given the stock a bit of a poop - though it does dilute shares).

What's scary is the over 13% in short positions on this company (source Yahoo finance, ket stats link). This is rather high for a company this size and one that is not a financial institution these days. The Price/book is also moving steadily toward one (not a good sign).

So much for the people who were happy the stock was going down prior to Barcy launch, so they could pick it up on the cheap and watch ir rake0 off after K10 launched! I guess they may have enough money to sell the stock and buy some of the heavily discounted AMD chips.

Giant said...

abinstein is a coward. Now making excuses on AMDZOne

Indeed. AMDZone's site is FUBAR though. I tried to register an account but was unable to do so.

On Scientia's blog I reminded Abinstein of his earlier words:-


Remember what you said here, because I'm quite sure you'll find yourself dead wrong when Phenom and Penryn are released.

According to motherboard makers who have tested both chips, Phenom does have somewhat better IPC than Penryn. The latter's advantage is SSE4 (which Phenom lacks) and higher clock frequency. If however AMD can speed up the release Phenom at 3.0GHz, Penryn will lose the lead, which may not be taken back until Nehalem.


This was in response to my estimates that Yorkfield would enjoy a 15% IPC advantage over Phenom. How close was I? According to Toms Hardware, a 2.4Ghz Q6600 is 13.5% faster than a 2.3Ghz Phenom. There's a 4% clock speed difference, so the Intel system enjoys a 9.5% IPC advantage. Yorkfield adds 5% onto that (excluding SSE4), giving Yorkfield an IPC advantage of 14.5%.

Anonymous said...

Abinstein is still holding onto his claim that K10 is clock for clock faster than Core2. Just look at spec_rate!

surely since K10 won there, it means that all the other "toy benchmarks" are just not indicative of K10's potential.

now if you excuse me, I have to go back to measuring the size of the dies now to determine who will have better IPC.

;)

GutterRat said...

abinstein the tool?

I thought he shriveled up and vanished, along with the rest of the harlequins, flaneurs, and faqirs claiming "rev 10h" was going to save Arabian Micro Devices from the jaws of evil Intel.

One of the kind folks in Austin alerted me of
this brand logo. I am told the reason for the delay of some high-end products was because they needed to redesign the packaging to accomodate the new logo.

LOL

Anonymous said...

Serious question folks (not looking for flames):

AMD had to know about this TLB errata earlier then it was announce no? If this really is the cause of the performance hit, why launch Phenom at all? Why not use the Phenom capacity to make more Barcelonas which will get as good if not better margins? They could still have launched the rest of the Spider platform and if people wanted to buy early they could plug in 2 cheap K8's with the intent of upgrading to K10's in the future.

Or why not launch dual core K10 versions? Or use their capacity to make more K8's? If they are hoping to sell the K10's in the 250 range - aren't they better off selling 2 K8's in the ~$125 range? The K8 must get better yield and binsplits so this would seem to be better from a biz/cash flow perspective, no?

I don't think it is to appease the enthusiasts - the real enthusiasts are already pissed off and will want to get the next (B3) stepping not this dog.

Unless AMD was truly surprised by the errata issue at the last minute (which I find hard to believe), it seems like they did not look at this thing from a business or logic sense and just said screw it let's launch anyway and we'll fix it later.

Thoughts? The strategic thinking at AMD seems, well, curious...

Ho Ho said...

"Or why not launch dual core K10 versions?"

Has anyone actually seen or heard anything about K10 dualcores? For some reason I think they are hoping to skip those entirely as K8 seems to do OK and is cheaper to manufacture.

Tonus said...

why launch Phenom at all?

Because you've got to launch it at some point, and with the stock price dipping the way it is, shareholders may be bringing pitchforks and torches to your next shareholder's meeting. I don't know that there is any choice they could make that would be a good choice, but this is probably better than not launching at all.

I do think it's worse than launching the way they have, slowly whittling the line down. First it was 2.2, 2.4, and 2.6. Then it was 2.2, 2.3, and 2.4. Now it's 2.2 and 2.3, with 2.4 delayed due to a bug and 2.6 being pushed back and showing a TDP of 140W. Considering their track record so far, you have to wonder if they're telling the truth about the 2.6, or if there will be another delay when Q2 rolls around.

After all that has happened, I think that they HAD to launch something. Had AMD been more up-front and honest, the launch would still be disappointing but at least we'd know what the deal is. As it stands, the launch seems really ugly and there is no reason to trust AMD's announcements going forward. And that is a really bad thing, IMO.

Anonymous said...

It seems AMD's Barcelona = Intel's Prescott. Sounds like a lot of enhancements, but ultimately disappointing in performance and heat output.

AMD fanboys shouldn't hold their breath waiting for that magical stepping that will double performance and enable clockspeeds in excess of 3ghz.

The 9900 model is the 2.6Ghz right? What happens to the 2.7GHz model? Phenom 10000? 2.8GHz, Phenom 10100?

Anonymous said...

"After all that has happened, I think that they HAD to launch something..."

I guess this is my point/question. Why did they HAVE to? From a business and financial aspect it makes no sense. If you see someone who plays poker and have a crappy hand that you know is beat and say well I already have put $30 in the pot so I might as well put in another $30 - I'll show you a losing poker player.

This is now a low price/low margin chip for AMD and they run the risk of diluting any potential brand image of "phenom" with such a poor performing initial chip. Now even if they ratchet up the clock and somehow get better IPC through a bug fix (not sure how this would happend), the Phenom is going to be associated with cheaper and will have a difficult time commanding a premium.

Sure they would take some PR lumps if they delayed the launch but if I were a stockholder, I would rather them sell better margin K8 dual core chips (or even low clocked Barcelonas which have a decent ASP) until they can sort out the problems on the quads. Business is about making money not trying to save face and compete in a niche market (quads on desktop) for what I think are largely ego concerns.

Anonymous said...

"Has anyone actually seen or heard anything about K10 dualcores?"

I think originally they were planned for Q1'08 (maybe even Q4'07 originally), but were pushed to Q2 - not sure if that will hold up given the issues and seeming focus on quads.

"The 9900 model is the 2.6Ghz right? What happens to the 2.7GHz model? Phenom 10000? 2.8GHz, Phenom 10100?"

You are assuming there will be 2.7 and 2.8GHz chips? :) I would guess they start skipping the 100MHz increments above 2.6GHz. As for the number they may start some enthusiast type name crap like FX or some other designation.

Maybe instead of Phenom it will be Phenomenal!

GutterRat said...

AMD shares slump to four-year low on funding concern

AMD shares slump to four-year low on funding concern

AMD is the worst performer on the Philadelphia Semiconductor Index

By Ian King

BLOOMBERG NEWS

Thursday, November 22, 2007

Shares of Advanced Micro Devices Inc., the second-largest maker of computer processors, fell to their lowest price in more than four years Wednesday on concerns that the chipmaker will have to raise more money.

AMD, which employs 2,600 people in Austin, dropped 45 cents, or 4 percent, to close at $10.83 at the New York Stock Exchange after falling as much as 6.7 percent. The shares had declined 45 percent this year before Wednesday.

Investors are worried that AMD's sale of an 8.1 percent stake to an Abu Dhabi investment company for $622 million last week won't be enough, analysts including Jefferies & Co.'s John Lau said. Another stock sale would dilute investors' holdings and lower earnings per share.

"Investors fear another round of financing may be forthcoming," Lau said. "The recent secondary offering still isn't enough for them to fund their future plans."

AMD, based in Sunnyvale, Calif., has reported losses that total more than $2 billion in the past four quarters after spending more to compete with larger rival Intel Corp.

AMD won't return to profitability until 2009, according estimates in a survey by Bloomberg.


What do you think? Shall we put a fork in this and call it a done?

SPARKS said...

“Well I think people are digesting the 700Mil cash infusion as relatively meaningless”

Man, I wish I said that before you did. I really thought I’d get beaten up for taking that view; then again, ‘Jim, I’m an electrician, not a broker’.

That said, why would Middle East Investors throw money at (away?) a failing company? Do they know AMD’s current position with both finances and INTC? If so, why go with them anyway? What’s the motive? Are these guys taken in by the AMD spin machine like everyone else, or is their something else far more financially tactical which I’m not savvy enough to understand???

Hell, I just don’t get it.


SPARKS

Giant said...

$10.83 and falling!

Anonymous said...

"That said, why would Middle East Investors throw money at (away?) a failing company? Do they know AMD’s current position with both finances and INTC? If so, why go with them anyway? What’s the motive?"

My best guess is that this is an M&A play and they are hoping that someone buys AMD - the price/book is now so low, it is a real possibility. Even putting say a 30% premium on the current share price, this would be ~8Bil.

What I couldn't understand was why the stock did not drop right after news of the investment got out - the purchase immediately diluted the stock ~8%, yet somehow the market didn't correct for that right away. I guess folks assumed the cash infusion somehow would increase earnings? (sarcasm)

I think at this point with the stock so cheap, it may actually not fall that much more as the rumors of a buyout will increase as the stock goes down (which would prop up/stabilize the stock price). Unless they start selling off pieces of the company, I'm not sure how much lower the stock price could go (maybe $9?) - this of course assumes no additional stock offerings - obviously any new stock offering should dilute the share price by the amount of the offering.

SPARKS said...

"shareholders may be bringing pitchforks and torches to your next shareholder's meeting."

No they won’t, they are NUMB!

Case in point, I had a friend, (past tense) who back in January told me how AMD Barcelona was going to “DESTROY” C2D! I told him don’t buy into it and get the hell out as soon as possible.

Months later, after Wrecktor De La Ruinz took his 16 million, after AMD lost big in two consecutive quarters, I mentioned how CEO’s can lose billions, fire thousands and still collect their pay. I cited Wrector as an example. This guy yelled at me and chastised me as if it were not permitted to speak the unspeakable! If he weren’t a friend, I would have racked him up against the nearest wall.

We no longer speak. Further, he has lost (forgive me) “Tens of thousands” of dollars. To this day, he still will not admit how wrong he was and he can’t look me directly in the eye.

I truly believe there are many more like him. Despite all to the contrary, they still believe! They are in denial. They were lied to and deceived. They believed all the spin we, on this site, saw right through.

I’m sure Henri Richard saw this coming. However, he left with some integrity and character still intact. He wanted nothing to do with it. That much is clearly obvious, and in retrospect, his timing was perfect.

Someone once said there’s a sucker born every minute. If it weren’t so sad it would be pathetic, if not criminal. At this juncture, reality hasn’t set in,--- yet. By then it WILL be too late.

I’m still waiting for my old buddy to come to me and apologize. I will then know AMD is truly finished. Till then, there’s still hope, and P.T. still Barnum lives on at Advanced Micro Devices.

SPARKS

SPARKS said...

“My best guess is that this is an M&A play”

Sorry, forgive me, I’m not familiar with the term, M&A.


“stock did not drop right after news of the investment got out - the purchase immediately diluted the stock”

This left me scratching my head, too. I guessing, but perhaps they were waiting for Pheromones release as the stock drop steadily immediately after the benchmark numbers were in/out.

You’re absolutely correct about the nine bucks. 4Q earnings are a little over a month away, and they still have nothing to sell.

Share dilution? Wrecktor will sell is mothers house at this juncture. In fact, 5 dollars a share is no stranger to AMD, January 2003.

Again, what is the M&A play?

SPARKS

Orthogonal said...

M&A = Merger and Acuisition.

I also believe the 8% stake is pure speculation on a possible take-over. I'm sure someone out there is crazy enough to take it one when they are trading at a relatively cheap price to entry in the market. (Forbidding possible x86 license issues)

I haven't commented on Phenom yet, but I guess I don't need to. The reviews speak for themselves.

Anyway, I haven't been around much lately since last week was one of those weeks from hell, and nothing gets done this week due to holidays plus I'm taking some time off.

Happy Thanksgiving and have a great weekend.

InTheKnow said...

This from another blog...

Yields are broken at AMD

I call Yield a good die at the right speed. In this case Yield is broken by all measures


This is both right and wrong.

It is right in the sense that yields that really matter are representative of both bin splits (i.e. speed) and functional die. A die that falls below a certain speed level can't be sold.

On the other hand it is wrong in the sense that the only data I have ever seen Intel or AMD release is a plot of defect density described as yield. This would imply that what both Intel and AMD are referring to when they speak publicly about yields is strictly in reference to functional die.

So in this sense AMD probably does have good yields. The problem is, they are only telling half of the story. A perfect wafer isn't worth the silicon it is made from if all of the die are too slow to sell. And I think that is the real story with Barcelona. It isn't that AMD can't make a functional die. It is that their process window is so tight that they can't produce a fast enough part to sell.

But I'm sure that APM 3.0 will fix all that in short order. With the near magically ability to repair individual damaged die that has been ascribed to the system we should be up to our armpits in 3GHz Barcelona's in no time. (sarcasm intended)

SPARKS said...

"Anyway, I haven't been around much lately since last week was one of those weeks from hell"

Don't worry about a thing, buddy. You just keep cookin' up those fat X48's. Save the fastest ones and send 'em this way. Tell the boss to hurry up with QX9770. I'm dyin', and my Raptor's are look'n for a new home.

QX9770,1600+ FSB
HOO YA!

SPARKS

Roborat, Ph.D said...

from a technical analysis point of view, the $622Mil investment made sense, purchasing the stock at around ~$12.5. AMD's stock has been hovering above $13 and from a TA point of view the stock has reached bottom. But as we can see now that that investment has already lost considerably. Sometimes investing based purely on TA isn't good.

I think the reason why AMD's stock is being beaten is because of the nervousness in the market about companies with huge debts. Much worse are companies with huge debts AND negative cash flow.

See Micron as an example. It's only a matter of time before AMD gets treated the same.

JumpingJack said...

"I think the reason why AMD's stock is being beaten is because of the nervousness in the market about companies with huge debts. Much worse are companies with huge debts AND negative cash flow."

I cannot help but agree. Poor AMD, the actions they take are not necessarily bad, just badly timed.

* AM2 needed to come as the memory technology was needed, but they force people to upgrade via new MB, opening up a C2D/new MB combo as an attractive alternative... likely losing some install base.

* ATI purchase, not a bad idea but at a time when they face the toughest market in years and a jittery credit market.

Now they are caught in a neg. cash flow with mounting debt and the possibility of a down swing in the market, followed by the launch of underperforming products.... eeegads.

Poor execution is only compounded by the poorest of timing.... and there is no doubt, Conroe and the NGMA (aka core uarch) caught them way way off guard.

Jack

Anonymous said...

"AMD's stock has been hovering above $13 and from a TA point of view the stock has reached bottom"

No offense but I think your TA is based on the previous # of outstanding shares. Don't forget this was ~8% new issues - from that alone, the stock price should have adjusted down ~8% in price for a net neutral - this would mean ~$1 off the $12.50 value at the time or put it at $11.50 "real" value (and that would reset your TA bottom to $12).

It's also not clear to me what you are basing your "bottom" on to begin with. If I do some amateurish chart reading and you look at the low in beg Apr and the low in mid-Aug, it looks like the resistance level would be somewhere in the high 10's (~10.7-10.8) - which is where the stock is right now. So if it goes lower than it is, the stock could be in for a bigger drop. Of course with the change is share #, I'm not sure if you can chart the bottoms like that.

So take the 8% dilution, sprinkle in the overall tech decline recently, take a dash of some difficulty raising new capital if needed and I think this turkey is ready for basting! All that is needed is a 4x4 board, fill it up with some non-existent Phenom 9900's and this thing should cook in no time!

I don't think the launch really matters that much - it's not like K10 was going to have any material impact on earnings for the next quarter or two. It may impact those looking to buy/hold long term - but I think traders didn't care one way or the other.

Anonymous said...

".... and there is no doubt, Conroe and the NGMA (aka core uarch) caught them way way off guard."

I tend to agree, but given the mobile data as people were starting to plug these into modified desktop boards, it amazes me that they could have been that surprised.

Regardless I don't think there is much they could have done (other than maybe not buying ATI) - it's not like they could have sped 65nm up - remember they are on IBM's development schedule, not their own! I also doubt they could have sped up K10, the only thing the probably could have done would be to start it earlier, but that was years ago and well before core u-arch.

I keep hearing perfect storm, but the problem is pretty much a matter of debt they took on from ATI - technologically (in terms of CPU's) they'd be in a similar position with or without the acquisition and the only difference is they would be bleeding cash more slowly.

pointer said...

InTheKnow said...
...
It is right in the sense that yields that really matter are representative of both bin splits (i.e. speed) and functional die. A die that falls below a certain speed level can't be sold.

On the other hand it is wrong in the sense that the only data I have ever seen Intel or AMD release is a plot of defect density described as yield. This would imply that what both Intel and AMD are referring to when they speak publicly about yields is strictly in reference to functional die.

So in this sense AMD probably does have good yields. The problem is, ...


actually not true ... a chip will be tested within its frequency range (with guardband), for instance, if the lowest product frequecy is 1.8GHz, then it won't be freq-tesed at anywhere lower. Thus, even a particular chip would have passed at 1.6GHz, AMD and Intel won't know that and thus not within the statistic collection.

Test time is money. The shorter it is, the better. Thus I don't think the CPU will be tested anywhere lower than its lowest freq bin.

Roborat, Ph.D said...

Actually clock frequencies are assigned by fusing based on the frequency capability of the device under test. What the Test process does is it runs the device at normal operating conditions while using a frequency detect circuit. And based on its frequency, the CPU is assigned a pre-determined clock speed by fusing its bus clock multiplier (normally 4-5 speed grades). To prevent over clocking, the closest lower speed grade is always assigned. If a CPU doesn’t reach the lowest speed grade, it’s rejected (as an electrical yield loss or in some cases shipped to some console maker – *couch*xbox).

Marketing CPUs in different speed grade is quite complex therefore the number of different speed bins/grades is generally financially influenced (i.e., target=maximum profits). Technically CPUs failing to pass the lowest speed grade are functional; the major problem is selling those affects your profit as a whole as the market tends to shift to anything cheaper. Except of course if your process is broken, and all you have are rejects and you’re well past your promised launch dates. In extreme cases like these you have no choice but to sell.

On a side note, there is some degree of attractiveness in buying the highest bin because it doesn’t have an upper limit and nobody really knows how high its true clock really is. It is unfortunate that it’s limited but there’s nothing stopping you from imagining that you possess the fastest and the most perfect CPU ever built.

pointer said...


Roborat, Ph.D said...
Actually clock frequencies are assigned by fusing based on the frequency capability of the device under test. What the Test process does is it runs the device at normal operating conditions while using a frequency detect circuit. And based on its frequency, the CPU is assigned a pre-determined clock speed by fusing its bus clock multiplier (normally 4-5 speed grades).


well, I was wrong then, since your explaination is indeed very convincing. I don't such design for test feature exist ... must be because the tester can't handle the frequecy as high s what the currect CPU has for testing then (guess), beside this could really shorten the test time.

SPARKS said...

“doesn’t have an upper limit and nobody really knows how high its true clock really is.”

Hold on right there, DOC. Explain something to me like I’m a two year old. Are XE chips going out the door where NO ONE has tested their absolute limits!?!?!?! I’m shocked and appalled! There must be something besides ‘ok, it reached its target frequency, keep the multiplier unlocked, and sell it’!

Lets say there one choice area of one wafer out of, say, lot # i8U2 where everything went perfect. Every layer, every step, every etch, every photo exposure, on each layer looked like two years of ‘Playmate of the Month’ centerfolds stacked one on top the other. Basically, every frigg’n atom was where it was supposed to be.

Do you mean to tell me that this one VERY SPECIAL bad boy is slipping out the door AND NO ONE KNOWS WHERE IT’S GOING OR WHOS GETTING IT??????

DOC, YA KILLIN ME! I’M PACING THE FLOOR, AND I WON’T SLEEP TONIGHT!

IF SO, THEY GOTTA FIND A BETTER WAY, MAN!!

I’ve got it “Extreme Edition PRIME CUTS”, tested and priced for lunatics, HOO YA!!!

SPARKS

InTheKnow said...

Roborat said...

If a CPU doesn’t reach the lowest speed grade, it’s rejected (as an electrical yield loss or in some cases shipped to some console maker – *couch*xbox).

So if I am reading this right, you concur that it would be possible to separate defect based failures from parametric failures, at least to some degree. One could debate whether or not it would be ethical to do so, but it is possible.

The conversation is, of course, purely hypothetical since AMD would always take the higher moral ground and never stoop to such a questionable practice. ;)

Roborat, Ph.D said...

intheknow said: "So if I am reading this right, you concur that it would be possible to separate defect based failures from parametric failures, at least to some degree."

There is a burn-in test that screens silicon type defects either instantly (catastrophic) or after being stressed (infant mortality). It's a little generalised but the answer is YES.

Roborat, Ph.D said...

Sparks said: Are XE chips going out the door where NO ONE has tested their absolute limits!?!?!?! I’m shocked and appalled!...

all i'm saying is that a QX9650 at 3GHz for example could be at any frequency above 3GHz before it was fused down to 3GHz. Of course there are theoretical limits based on circuit and thermal designs which would cap the frequency at some unknown level. I personally don't think it's outrageously far from 3GHz but ofcourse the results overclockers get suggest that Core2 has a healthy headroom.

enumae said...

"6.00% Convertible Senior Notes due 2015

On April 27, 2007, we issued $2.2 billion aggregate principal amount of 6.00% Convertible Senior Notes due 2015 (the 6.00% Notes). The 6.00% Notes bear interest at 6.00% per annum. Interest is payable on May 1 and November 1 of each year beginning November 1, 2007 until the maturity date of May 1, 2015.
"

How much does this November 1st payment end up costing AMD and what effect could/will it have on AMD's Q4 numbers?

Thanks

SPARKS said...

“personally don't think it's outrageously far from 3GHz but of course the results overclockers get suggest that Core2 has a healthy headroom”

Healthy headroom, is like saying Angelina Jolie has a cute figure, hello!

The QX9650 is a PRIME example. They’re selling this MONSTER at a base line 3 Gig,----- cake! From what I’m reading, EVERY reviewer is clockin’ this thing at 4 Gig, on air, no sweat.

We discussed this previously, and we’ve all have come to the conclusion INTC is sandbagging the chip. They could have released this bad boy at 3.33 Gig, no sweat.

But hey, more power to ‘em, no complaints here. Fast cars, fast boats, fast women, a fast life style, and fast chips are gonna cost ya, here’s my money. But, Maximum PC, even in its review USED the word "sandbagging".

Orthogonal said testing chip is time consuming, hence expensive, no doubt. They MUST have a way, which I would love to know, by the way, just how they separate the XE’s from the plain Janes other than JUST selling the XE’s with an unlocked multiplier!

You chip guru’s know better than I, one glitch, out of millions on the chip, will bring the whole house down. Therefore, how does INTC find the REAL good ones, the crème de la crème, Top of the heap, A number one?

Look, I’m sure the gurus at INTC selected the prime cuts, and clocked the living guts out of them during testing. I want my QX9770 to be as good as or better than Mr. Coolalar’s. That what I’m paying for!! Am I not???

I don’t care about power, 1 kW works for me. Thermals, I’ve got the pump, radiator, and hoses, you bring the H2O.

“Of course there are theoretical limits based on circuit and thermal designs which would cap the frequency at some unknown level.”

DOC, I hate to say this, but you sound like some marketing guy, or politician who knows the real deal, but doesn’t want anyone to know those “unknown levels”.

Unknown levels, horseshit! Over 40 billion in market cap, over 10 B in sales a quarter, they build 3 4.5B dollar shacks, the 1st 45nM out of the gate, and they don’t know? Gimme a BREAK!

THEY KNOW! And, I want one!

SPARKS

SPARKS said...

Enumae, they've got bigger fish to fry! Read this:

http://www.digitimes.com/systems
/a20071121PD215.html

SPARKS said...

"Even at a lower core voltage my QX9770 ended up consuming around 30W more power than the QX9650, but it's quite possible that I have a really bad QX9770 and a really good QX9650. Individual chips, even at the same frequency, can have vastly different operating voltages - which we already knew."

I hate to harp on this, Doc, but do you see what I'm talking about?

http://www.anandtech.com
/weblog/showpost.aspx?i=313

Anonymous said...

"How much does this November 1st payment end up costing AMD and what effect could/will it have on AMD's Q4 numbers?"

Not much... 6% annum is a little less than 1.5%/quarter (too lazy to do the math to account for compounding interest).

1.5% of 2.2Bil is ~$33Mil interest expense (not sure how exactly this is classified on the balance sheet) - regardless it is not a huge impact the way AMD is losing money. Though if they lose in the neighborhood of 300Mil, it represents ~10%...

Anonymous said...

"I hate to harp on this, Doc, but do you see what I'm talking about?"

While Anand lowered the voltage to the voltage Intel spec'd it is not clear if that is the same voltage as the QX9650 chip. Does anyone know the normal voltage for that?

A lot of times you will see higher voltages to get the top bin chips so even though the 9770 was lowered it still may be higher than the QX9650 chip - this could easily account for the 30W difference.

QX9770:
Vcore = 1.285V Anand (AFTER ADJUSTMENT DOWN)

QX9650
Vcore = 1.120V (legit reviews - though for some reason the CPU-Z shot had the multiplier set to 6 for a total clock of 2GHz, so this data point is dubious)

Vcore - 1.232V (Hot hardware review)

Vcore = 1.176 (X-bit lab review)

Vcore = ???? (Anand)


An older review at Anands had a QX9650 at 1.216V - not sure if this is the same chip as what he used to compare the 9770 and 9650 power levels. This is .07V lower which is not insignificant.

This could be just sloppy/incomplete reporting - if he wanted to see if something was fundamentally different between the chips he should have run them at the same Vcore's. The 30W may be just the fact that Intel needs that extra Vcore to get the clock speed for the next bin. Given the BIOS issue to begin with, the whole review makes me a little cautious about his conclusions / thoughts.

SPARKS said...

Anonymous, for your pleasure,

“Core voltages have also jumped slightly, and while the 3.0 GHz Core 2 Extreme QX9650 ran at approximately 1.2V-1.25V, the 3.2 GHz Core 2 Extreme QX9770 increases that to approximately 1.3V.”

From Sharky Extreme

http://www.sharkyextreme.com/
hardware/cpu/article.php/3712036

SPARKS

InTheKnow said...

From another blog...
Hehe, I remember having a few discussions about the same subject with him. I said "3GHz in late Q2/H2, if lucky".

Let’s change that a bit.
How about AMD got unlock with that TBL problem and can’t release anything faster.

If AMD didn’t have that problem you, mo and lex where eating your own words.

PS: You all talk like Intel never had happen such a thing, when it happened a lot of times and it took them months to years to solve them.


This stuff just continues to amaze me. With all the information out there people still seem to have problems separating architecture and process. It's really pretty simple.

As an analogy, you can build a skyscraper with a bad design and it falls down (a bad architecture). You can build a skyscraper with bad materials and it falls down (a bad process). You can build a skyscraper with a good design (good architecture) and good materials (good process) and it lasts for generations.

Since Prescott (which all the AMD fans love to point to as a bad Intel design) was made on the same process as the Pentium M, the process must be good and the Prescott design was flawed.

Trying to compare Barcelona/Phenom to Prescott is just wrong because AMD has yet to produce a great part on 65nm. Hence, it is a process problem and not some mysterious architectural bug.

Anonymous said...

Sparks - thanks for the link. I;m surprised Anand didn't mention it in the article. I'm not sure if it would account for the remaining difference, but it would certainly contribute.

The Anand article was rather curious (and in my mind sloppy). Not noticing the huge Vcore increase, than not mentioning that after it was fixed, the Vcore is still higher. I expect more from that site - perhaps they felt rushed to publish something? I will be reading their reviews a little more carefully going forward.

Anonymous said...

"With all the information out there people still seem to have problems separating architecture and process."

people will believe what they want to believe. AMD fans WANT to believe that AMD's process is competitive and Intel is not "leagues" (to borrow a description from another blog) ahead. Thus 90nm Prescott = bad 90nm process and bad 65nm K10 = bad architecture.

Let them wallow in their ignorance, it's not like folks have tried to educate them. If they are unwilling to listen they are no longer worth the effort.

P.S - is Scientia OK? Haven't seen any comments from him - he's usually good for an average of 1 out of 3 or 1 out of 4 comments in his blog. He must be having a hard time how to figure out how to sping this. I was at least expecting 'Phenom - a good start' or 'Desktop - it's not only about performance' or 'K10 - designed for 45nm' articles.

Ab-I'm-an-idiot also has been rather quiet and not doing more of his excellent why AMD has 50% better yield analysis.

Ho Ho said...

Well, at least Abinstein has been showing up on amdzone, though he doesn't seem to exactly succeeding in spinning the K10 to look as good as he predicted. Scientia last visited amdzone on 18'th and on the same day made his last comment on his blog. I sure hope he didn't commit suicide after seeing Phenom launch, world would loose a great comedian.

enumae said...

Thanks Anonymous and Sparks.

Peter said...

I'm worried about Scientia. He has been MIA ever since Phenom launched.

Did he jump off a cliff or something? Is there a limit to how much a fanboy can distort reality before even they realise it's a lost cause?

SPARKS said...

“The Anand article was rather curious (and in my mind sloppy).”

It was a crock of shit from get go.

I think INTC thought so, too. He was on the phone with INTC engineers immediately after the article was posted. Frankly, the initial article indiscriminately beat the piss out of the QX9770. The update clearly shows his kinder, gentler attitude towards INTC’s latest and greatest (love my life)! Not to mention pissing off INTC!

That said, as an industry veteran, he should have realized something was wrong somewhere. Perhaps, he was (hopefully?) looking for a chink in INTC’s armor, as this 800 pound gorilla chip absolutely destroys any chance of Pheromone pulling off any eleventh hour miracle from GOD.

Maybe he put too much curry on his turkey; I don’t know.

Conversely, Vince Freeman at Sharky Extreme is probably THE most unbiased reviewer on the web. My respect for this industry veteran is unparalleled, with the exception of GURU, of course.

SPARKS

Anonymous said...

ho ho and others,

abinstein and scientia are hiding.

abinstein knows his arguments are flawed and given what we know now of Phenom vs Penryn and of Barcelona, that he does not have a leg to stand on, so he is avoiding posting in Scientia's blog. He has also not posted in his blog for the same reason.

As far as Scientia's concerned, he's lost a lot of credibility and does not want to face up. He too is also posting on AMDZone.

Both of these fellas seem to want to stick to "controlled" environments.

Anonymous said...

Is it shocking that Scientia has been missing since the launch of the Phenom?

Is the news that shocking that he can't face his OWN BLOG?

I can't wait till he comes back and spreads his godly wiseness. Because obviously we are all wrong and he's always right LOL.

Anonymous said...

Isn't it obvious that despite claims to the contrary, Scientia and abinstein are a bunch of AMD fanbois? If they had no affiliation to AMD they would have come out already and derided this piss poor execution of Phenom, 65nm and, more importantly, their financial situation.

They must be having one hell of a time finding someone willing to unjam the tails from between their legs.

Fools.

Anonymous said...

Did Scientia jump off a cliff or something? It's been a week since he last posted.

I wonder if that is the end of his blog then. It will surely die off once people realise he is too cowardly to face up to the music.

Khorgano said...

For someone who claims to be platform agnostic without a vested interest in either company, it sure is curious to see Scientia completely silent and MIA. Surely he wouldn't feel threatened or scared to post on his own blog.

From a consumer standpoint, when it comes down to it, the Phenom isn't a complete failure, it will surely find a place in the low to mid-range systems of the world which is by far the largest market. AMD seems to be pricing them relatively competetively and they'll surely sell whatever they produce. However, on the business side of things they are at a HUGE disadvantage. I know I'm preaching to the choir, but God know's how long they'll be able to maintain manufacturing these chips with suffering margins, process and architectural disadvantages and no end in sight to stop the red ink from bleeding.

They may be able to keep sloughing along while issuing more convertable bonds, issuing more shares and receiving gov't subsidies, but the fact remains, the core business is nowhere near profitable and the cash IV will begin to run dry soon if nothing changes.

SPARKS said...

I noticed on this sites home page in the related new section the mention of Apple buying AMD. Who ever started the absolutely ridiculous rumor cannot be smoking the same stuff as Steve Jobs.

First, Stevie wouldn’t get it as cheap as you think. Why, you may ask? As soon as Apple EVEN TRIED to make a move the stock price would jump to over inflated levels.

Secondly, he would have to, as of last week, wrestle away a little over 8 percent of the company from Middle Eastern Investors. They speculated on this. Even if they were in it for the less obvious reasons of buying into future hi tech for the area, it wouldn’t be cheap buying it back, in any event. They would want to turn that 620 million into a nice cool billion.

Third, there are a lot of BIG institutional investors in this country who own a considerable portion of AMD, and they have taken a considerable beating as of late. They will want to recoup some those losses. They won’t let it go for cheap, either.

Fourth, Jobs took over the company when the PC platform was crushing Apple performance at the time. You see, Motorola chips were dogs. IBM chips were dogs. The Motorola group was lead by none other than WRECTOR DE LA RUINZ back in 2000. (Add that to his resume) Why would Jobs buy into this newly created mess by Wrector and his Fairchild/Motorola minions? This not to mention IBM’s dog chip which were also pee’d on daily by PC.. And I ask you, who is currently Wrectors tech partner? Yes IBM and SOI!!!! Jobs just got away from those idiots!

Fifth, debt, debt, and more debt, somebody has to eat it.

Sixth, Apple is way on top again with a harmonious relationship with INTC; why jeopardize this relationship? Further, why spend billions on weaker platform they didn’t design, only to be a competitor to INTC??? INTC is on a serious roll now, why would he spend billions to take hit on a substandard platform which needs more time and more MONEY?

Seventh, Apple offers and makes Apples, not Blue ones, or Green ones. The marketing aspect of a platform choice wouldn’t fit in their basic company strategy. At this point, he doesn’t need to spend billions internally competitive platforms, Apple (AMD) vs. INTC as a choice to consumers. He certainly wouldn’t drop INTC cold, not now, not at these performance levels. Then when would be a good time, a year, perhaps two? The money would be BURNING during that time period.

In fact if I were Jobs, I wouldn’t have forgotten the way Motorola treated Apple back in 2000, with spin, lies and more spin under Wrectors reign. (Sound familiar?) In fact, at this juncture, with my current success, I would sit back enjoy my profits, my share price, and gloat at Wrecktor Ruinz ultimate career blunder, and save myself billions.

I wouldn’t be there to save Wrectors ass.

SPARKS

Anonymous said...

"Since AMD has promised 2.5Ghz in Q4 this should allow AMD to match the 2.83Ghz Penryn. This means that in Q4, Intel will still have two clock speeds, 3.0 and 3.16Ghz which are faster. In Q1 08, with 3.0Ghz, AMD should be able to match up to 3.33Ghz."

"Today, in spite of the low clocks, AMD's 65nm process for K10 has excellent yields. Intel's yields, in contrast, on its brand new 45nm process will take a couple of quarters to reach maturity."

"AMD should be able to reach good volume and a 3.0Ghz clock speed by Q1 08 and if we use the same six month lead then Intel should be looking good with 45nm in Q2 08."

"However, AMD has its own bright spots such as Q1 08 when K10 gets good volume and speed on the desktop"

Source K10 - a good start, a mere 2 months ago....my how times have changed when real data and roadmaps start to become available and the compiler argument, design team observations, and Si process technology dis-information (not misinformation) is exposed for what it is.

K10 will get better, it's clock will (eventually) scale, 65nm will likely stabilize a bit, but the problem is AMD needs this to be ahead of Intel to enable them to increase ASP's, if they need to sell at a discount then they are reliant on superior manufacturing which clearly is not AMD and it is only a matter of time before Intel scales their production to feed more than 80% of the market. The only way AMD survives is though a SUPERIOR product - simply competitive and they will get ground away via Intel's manufacturing muscle. Why not just keep making K8's which are cheaper and likely yielding better (at least for dual core). Why even use ANY capacity for desktop quads - just shift it to quad server and more K8 dualies until they fix things.

Tonus said...

sparks:

"In fact if I were Jobs, I wouldn’t have forgotten the way Motorola treated Apple back in 2000, with spin, lies and more spin under Wrectors reign."

While Steve Jobs has been willing to work with companies that he has reviled (Microsoft, IBM and Intel come to mind), it's because beneath that fancy-schmancy "machead" veneer, he's a pragmatic businessman at heart.

Therefore I agree, he has no reason to help bail out AMD. It could hurt their relationship with Intel, something Jobs won't want after all the flack he took from Mac fanatics for switching in the first place. Second, switching to AMD would be seeking a solution for a problem that doesn't exist. And the solution might be worse than the problem!

I have no doubt that he is not shedding any tears over AMD's plight with Hector Ruiz at the controls, but I don't think that Jobs would avoid a deal with AMD if it were the practical thing to do. However, it's just not the practical thing to do. AMD/ATI would be an albatross for Apple in its present form.

Anonymous said...

Apple has no core chip manufacturing expertise - what exactly would Apple be gaining from this?

The only possible benefit I see is what the BLOG states - to cut OEM margin that they pay Intel, Nvdia et al for the chips, video cards and chipsets. But at the market share Apple has, is that really critical and worth taking on a company with 5Bil in debt? What's the ROI on that? How much margin reduction is needed to break even on what would likely be a 8-10Bil dollar deal (and this excludes the operating losses expected in the near future and debt).

To give you an outrageous thougt - suppose they were able to cut $80 off their component costs - it would take 100million computer sales to make back an 8Bil purchase (now pay down the 5Bil debt and finance the upcoming quarterly losses). And that's just to BREAK EVEN! Oh, this also ignores the gross margin that AMD subdivision would have earned (effectively selling chips at cost to themselves) on those chips, putting the operating loss even greater!

There's also the rest of the market to consider - is Apple/AMD going to sell chips to computer manufacturers they are in competition with? Would Apple's competitors purchase chip/chipsets/video cards from them? Or would Apple/AMD effectively have pulled an AMD/ATI? (where ATI chipsets now only compete for 25% of the market). Do you think the FTC might see some possible issues with this?

If AMD products are such a great fit as the BLOG indicates, perhaps they should just BUY THE PARTS AND NOT THE COMPANY.

I'm planning on remodeling my house and will need some lumber perhaps I should BUY A LUMBERYARD or LUMBER COMPANY so I can get the materials at cost? (Well you say that's a crazy example - well what if I remodeled 100 houses a year - should I buy a lumber company then? How about 10,000? 100,000? If I know very little about the biz, what sense does it make to buy it?)

The article's premise is just ridiculous - I'm not saying Apple won't buy AMD, but to suggest they do so because it will help their margins is ridiculous. Next thing you know Apple will be buying MOBO makers, plastics manufacturers, LCD companies, power supply companies. Sound ridiculous? Exactly... this is not a 100Mil type purchase or a small startup.

Oh and cutting your potential supply chain down to a single company (unless you think Intel would supply chips to Apple in this scenario?) would leave Apple with the possibility of having underperforming parts and no possibility of buying other parts is not what I would call a very good strategic decision. You have got to be careful about buying someone on your supply chain - the rest of your possible supply chain is now your competition, so better make sure your choice was a good one.

The logic behind the blog is RIDICULOUS, there may be other reasons for doing it, but the ones given are absurd, not well though out and rather narrow minded (tactically focused).

Anonymous said...

"Apple’s solid financial condition not only can support but also will speed up technology innovation;"

So money solves the high K integration, Moore's law, architecture design issues? It is part of the solution, nut not the only thing needed. Oh and would IBM continue to license Si process technology to Apple? Any possible issues there, maybe?

"It is not technically difficult since Apple has already worked out with Intel CPU;"

This is not a reason why they should BUY AMD, it is a possible reason to USE AMD PARTS.

"It is cheap. AMD spent $5.4 billion alone to acquire graphics chip company ATI mostly by cash in June 2006"

See previous post - AMD is cheap RELATIVE to Intel, but it is not cheap on an absolute scale. Transmeta's cheap - is that a good reason to buy them? How about Via? SIS?

"The government will be glad and the deal won’t face challenge by government regulation of monopoly and competition policy"

See previous post - Apple competes DIRECTLY with people they would presumable be selling chips to? Apple still does make laptops, workstation and desktops, right?

"Intel (INTC) will gain a more balanced competitor, which is healthy for the entire semiconductor industry, and even Intel may benefit from it"

What specific expertise does Apple have that will fix the current problems that exist at AMD? Is Apple going to help with the K10 design, put some SW engineers on that nagging high K problem? They can throw money at the problem but other than that what else, specifically, can apple do tpo make AMD "healthy"?

"Apple’s best computer system architecture, operation system, user interface, and applications will combine with AMD’s most advanced and genuine multi-core CPUs and energy saving technology"

And this directly contradicts the 'well they already integrated Intel argument' - if their architecture works on Intel what ADVANTAGE do they get with AMD (other than slower parts). What problem does AMD solve on the computer architecture front? Yes there are megin reasons (which are bogus on the cost scales we are talking about), but what technical reasons are there?

Or is the blog's logic that they should they buy AMD simply to get to a point where they are already at today in integrating x86 into their computers?

I just don't see the technical advantages and there are a TON of potential technical shortfalls.

Then there's the strategic gamble
- what if Intel and/or maintains or opens up a significant technical lead and Apple is left using their own internal parts which can't compete?
- what if AMD's cost explode which wipe out any margihn advantage, or actually makes it worse than the current situation of Apple buying the chips (for example if they are not able to cotinue at the same pace as Intel on tech nodes, or if yields are horrendous). Do you think at this point Intel or Nvidia would sell them chips? :)
- what if their computer competition (HP,Dell, ACER, Lenovo, etc...) decide it is not in their best interest to buy chips from a company they are competing with? This would kill AMD's market share and make their chip manufacturing that much less profitable.
- What if IBM decides to no longer renew their technology agreement with AMD and tells them, you're on your own as we no longer wish to help make a company that we compete with make faster chips.

I get really concerned that a blog like Seeking Alpha starts getting some news play, which drives the stock price (temporarily) in the near future and someone mysteriously profits from it. (pump and dump?)

InTheKnow said...

Ok, so AMD still has an edge in 4-way servers and memory bandwidth limited applications. They will have this lead for another 3-4 quarters until Nehalem comes out, and then that niche is lost.

K-10 is a dog (should have stuck with calling it K-9 after all) on 65nm. Eventually AMD will get the speeds up, but by then it is unlikely that it will be competitive with Intel's newest offerings.

In addition to the product issues, they have all the recent debt they have taken on to deal with.

So if Hector gets the boot (and a huge severance package I'm sure) tomorrow. What does his replacement do?

I really only see one path for them to take.

I think they should focus on their strength in the 4-way systems immediately, but that is only a niche and not a long term solution with Nehalem on the way. This would buy them time to try and fix K-10.

With K-10 fixed, it should have pretty good performance per watt. I would tweak it as hard as I could to favor power savings to make a scaled down 2-core version suitable for laptops.

This is projected to be the next big growth area. If they really want to get out of the hole they have dug, they need to get in early on the next big growth opportunity. Short term I think that is laptops.

In the mean time, I would start looking hard at the designs for the Silverthorn competitor. These chips are going to be dirt cheap. Intel is going to get ~2500 of them on a wafer. I really believe that this thing is going to open a lot of doors to the x86 architecture. AMD can't afford not to get a foot in the door early on this one. They need to get 65nm fixed fast and get this thing on track to compete with Silverthorne.

It isn't a pretty picture any way you look at it, but I think this at least gives AMD some hope and maybe a way out of their current disaster.

Now I'm just a simple process guy, so I know there are bound to be gapping holes in this plan. What are they and what are the alternatives?

SPARKS said...

In The Know,

K9 !!!!!

LOL LOL!

SPARKS

SPARKS said...

“K10 will get better, its clock will (eventually) scale, 65nM will likely stabilize a bit”

Maybe, but from what the tech whizzes on this site have stated previously, wouldn’t bet the farm on it. The thermals, even at the currently released speeds, suck. GURU has me sold on the fact this thing and its process is leaking like a bloody pig. The fact still remains at 65nM, in its present form; this is still is an unproven process. The evidence is the thermals got worse going from 90 to 65, not better. 135W at 2.4 is a nose bleed. What will it burn at 2.8 or 3.0?

I’m no expert but there also seems to be timing issues, as the DOC mentioned months ago. They pulled 2.4 because it was unstable.

Even if by some miracle they get the thing up to 3 Gig they face a number of dilemmas, pricing, a failed launch, bad timing (late), volume, margins and clock for clock less performance than CONROE!

This is not including many AMD enthusiasts who waited and have become disenchanted with the company. An enthusiast based platform change now will be a fatal blow to AMD’s sales.
This will happen. Further, if Intel cuts prices again when they ramp 45nM, its game over. Presently AMD can only compete with INTC’s cheapest 65nM quad, barely.

To add insult to injury, my Q6600 (SLACR GO), as I type, is easily overclocking at three Gig stable. (3006 MHz@1336 FSB, 1.40V) Pheromones will never even catch this chip. This is gospel.

Pheromones is just not enough and simply too late.

SPARKS

Anonymous said...

"This is projected to be the next big growth area. "

No offense but what rock have you been living under? Projected? How about IS? For both Intel and AMD mobile has been the largest growth area for some time now and the market between laptop and desktop is projected to crossover sometime in late 08 or 2009. The problem with mobile is K10 adds little to that space - the independent clocking of cores is useful but it becomes more useful with 4 or more cores (the effect with 2 cores is marginal). As mobile will not be going quad core anytime soon due to thermal/power limitations their current K8 design is about as good as K10 - which is why AMD is working on a mobile architecture change.

The problem for AMD is that they been growing very fast, but that is somewhat related to the relatively small market share they started with in mobile. It is easy to get a 50% increase going from 8% to 12% but a whole different story getting another 50% going from 16% to 24%. AMD has eaten away at the low price notebook segment with Turion, but they will need performance to grow further.

Anonymous said...

"So if Hector gets the boot (and a huge severance package I'm sure) tomorrow. What does his replacement do?"

1) Raise (or at least maintain) chip prices and give up this tilting at windmill, ]all we need to do is get to 30% market share and everything will be OK' strategy.

They likely will not lose a lot of market short term as Intel is close to its capacity limit and is relatively lean inventory-wise(long term is a different story) This should in theory get them closer to in the black quickly or fund other development efforts. Long term to do this they will either need to improve the competitiveness of their products or greatly reduce their cost structure to justify lowering prices back down and remaining competitive that way - neither of which will be easy (but this will only get harder to do if their debt load continues to grow)

2. Give up (short term) on the quad core desktop market - focus all of their quad capacity on server where there is a real demand for quad, better margins, and ability to sell lower clocked parts (especially in 4P+ environments). Either get dual core K10 desktops working and out the door quickly or scrap it and focus on K8 (L3 cache on a 2core design seems useless to me - though admittedly I'm not an architecture guy. It also seems much of the K10 improvements are designed for 4 cores or scalability.

Yeah this is will be a huge PR hit but it's not like they are pressuring Intel in the high end of this market anyway and the folks buying in this area tend to look at performance over cost. Not to mention this still remains a niche market (likely thru 2008). How much revenue (and more importantly profit) does AMD expect to gain in this area in the next year vs the amount of revenue (and profit) of focusing that capacity on server and dual core desktop (or even more mobile). Let Intel drive the market and attempt to pull the SW market to catch up on things - by the time the majority of SW actually takes advantage of quads (1+year) AMD should be in better shape either thru fixing K10 clocks and/or migrating it to 45nm or just waiting for bulldozer for the quad desktop space (but keep producing K10 server quads)

3) Give up on the 'we can keep up with Intel on the process technology front' and stabilize the processes a bit before trying to ramp them. This should allow a reduction in the # of CTI steps and reduce the churn in manufacturing. Do an ROI on SOI and reevaluate whether the technical benefits still are a good tradeoff for the cost (and complexity?) increase for using SOI. If not, swallow your pride and do the right thing.

Keep in mind while the typical 30-40% die size reduction is significant, the added wafer cost of the new technology and poor yield and/or bin splits can quickly wipe out/offset any of these gains if the technology is not ramped properly. Throw in immersion litho (which, surprisingly to most, is more expensive than a 2 pass process) and any perceived 45nm gains can quickly evaporate. Throw in the fact that there will be little/no gate oxide scaling and you are likely not to even get significant performance gains (the only other reason to do a shrink)

4) Immediately start a crash course on bringing the graphics and chipset production in house (or just start with chipset which should be fairly easy to do) while maintaining enough volume to be able to use foundries as burst/upside capacity. This will allow better loading in the fabs and smooth out fab utilization when market slows. (this of course assumes they can ramp foundry production if they need to produce more CPU's in house). The cost of doing this can be subsidized short term with #1 (raising CPU prices) until the economies of scale of doing this in house kicks in.

In addition to keeping the fab(s?) better utilized, they will no longer have the foundry margin which should give them a bit of a cost advantage over Nvidia.

5) Kill fusion (or stick to only MCM approach). AMD does not have the market clout to influence the industry to optimize the market for this unless Intel also does a similar implementation. If AMD bets the farm on the next design on this and Intel goes in a different approach, AMD will be left holding the bag and won't have enough internal resources to prove out even a marginally better solution. (unless of course Fusion is VASTLY superior). Unless AMD has some internal simulation or data suggesting this is the next big thing and is 10X better performance, than this seems like a large bet for a company in AMD's shoes to be making.

These are just a few things - none will help immediately (except maybe #1, but there are no quick fixes in this industry. The argument of what will change immediately/short term if Hector is removed is not the right one - the questions is how much longer (and harder) will the road to recovery be if Hector remains?

InTheKnow said...

"This is projected to be the next big growth area. "

No offense but what rock have you been living under? Projected? How about IS?


Yes, that was poorly worded. It would have been more accurate to say that mobile will remain the next big growth area for the next several years.

InTheKnow said...

“K10 will get better, its clock will (eventually) scale, 65nM will likely stabilize a bit”

Maybe, but from what the tech whizzes on this site have stated previously, wouldn’t bet the farm on it. ....The fact still remains at 65nM, in its present form; this is still is an unproven process. The evidence is the thermals got worse going from 90 to 65, not better. 135W at 2.4 is a nose bleed. What will it burn at 2.8 or 3.0?


Don't get sucked into confusing the K-10 architecture problems with the 65nm process issues. K-8 on 65nm is starting to look respectable, but still isn't as good as 90nm. I don't see any reason to believe that K-10 can't at least match K-8's thermal performance eventually.

If you cut K-10 down to a dual core design and pitch it at mobile, I think you might have a decent product. It's not there yet, but I think that AMD could get there eventually. The big question in my mind is can they get there before K-10 no longer matters at all.

Of course all this depends on what the K-10 architecture was optimized for. If it was optimized for performance, then there is room for improvement in the power area. If it was designed to provide a balance of performance and power, there is still some room for improvement in the power area. If the initial design was optimized for power, then it is just toast and it should go the way of Intel's canceled Tejas design.

Anonymous said...

"If you cut K-10 down to a dual core design and pitch it at mobile, I think you might have a decent product."

What architectural features over K8 really matter in dual core and mobile space? It would seem many of the features were targeted toward server and scalability. From the reviews it does not appear the IPC is that much better than K10 - then you have to figure out what to do with the L3 cache in desktop (is it needed?) and mobile (is it need? How much additional power consumption does this cost you?)

I think when you cut K10 down to a dual core you more or less may end up with something remarkably similar to a K8 dual core with faster interconnects and the ability to independently clock each of the cores. Nice features, but are they worth the effort and additional die size?

Ho Ho said...

"I think when you cut K10 down to a dual core you more or less may end up with something remarkably similar to a K8 dual core with faster interconnects and the ability to independently clock each of the cores."


... and higher memory latency thanks to L3. Of course L3 will also make it much more expensive to produce than K8 based dualcore. I guess this is one of the reasons why they didn't want to use K10 for mobile CPUs.

Chuckula said...

Good news guys.. Scientia didn't commit suicide, he's back to censoring posts that don't boot-lick AMD's upper management.

Here's one of mine preserved here in the interests of free speech:

Hey Scientia: I got bad news for you, censorship doesn't work so great when you don't have Stalinist control over the whole Internet:

Instead of being a coward with a delete button, I challenge you to refute a single point I made (you can't). I further challenge you to say something remotely honest about K10 now that all of the cards are on the table and the marketing bullshit is just a bad memory. And without further ado:

Well today was quite the revelation. Considering everything we've seen for the last 6 months the fact that Phenom is an unmitigated disaster. I mean seriously, the 2.4+ Ghz chips can't come out because there is a freakin' race condition that is so common that the review sites were able to set it off just doing normal testing?

Then there's the whole "launch" fiasco where no honest websites were even given the chips ahead of time to do independent benchmarks. The only "approved" reviewers were effectively bribed by AMD with all sorts of goodies, and then given hand-picked machines that had chips in them that AMD will no even be selling for another 4 months! This has gone from being pathetic, to being unethical and AMD (starting with that asshat Hector-molester Ruiz) should be ashamed of itself. "Blow Away" Clovertown in every dimension? Shit, even if AMD could clock this pig to 3Ghz it would have a chip that is competive with Clovertown (although probably 5-10% slower on 95% of workloads), but beating it by 40%??? Show me one fucking task other than a rigged Spec_fp which had to be withdrawn from the Spec website where AMD actually beats Intel by 40%.... good fucking luck.

Since unlike Sharikou, Abinstein, one_molester, pezal (who still has the stupidest post of all time), Jeach & Jong, and whatever alternate logins Sharikou uses, I am not a blind fanboy, let's at least use this experience to learn about warning signs to look for when ANY company is bullshitting us:

1. Massive powerpoint hyperbole: Just look at anything Sharikou has parroted. Notice how Intel never claimed the moon & the stars for Penryn, just realistic, achievable numbers that it has more than met. Overpromise & underdeliver is a bad combination AMD.

2. Naked Hubris: Confidence is a good thing, but literally believing that you are God and can ignore the rules everyone else plays by is stupid and leads to a fall. Intel suffered from a smaller-scale version of this disease during the Netburst era but has since recovered and allows the engineers and not the marketing people to run the company.
AMD on the other hand took some success with the Athlon64 and X2, and instead of being smart & humble got arrogant. AMD thought that it no longer needed to worry about Intel ever again and completely squandered an opportunity to make a great follow up to the X2. The stupid insistence on "native" quad core which we now see is one of the stupidest buzzwords ever conceived, and the arrogance in taking a 65nm process that AMD's engineers knew damn well was substandard and making gigantic chips on it are just 2 small examples of AMD's hubris. AMD's shareholders and unfortunately the actually good employees who were not able to stop the idiot Motorola management are going to be the true victims of this.

3. Insane Secrecy: Every company tries for secrecy on new products, but note the differences. With Intel, even though it didn't want to give away the farm, it was confident enough in its chips to get Penryns out to customers to do actual tests with to make sure everything is going forward. AMD was so completely unable to manufacture anything that even its partners could not get in chips.
This leads to all sorts of idiotic rumors. Remember the "miracle stepping" and "dancing in the aisles"? What a fucking joke. The same goes for "30,000 3Dmarks". The stupid over-protectiveness lead to the hyperbole from #1 to replace any rational facts.

4. Becoming insanely controlling & paranoid: AMD has turned into that psycho stalker from a Lifetime TV movie. Instead of being open and honest when there is a problem they resort to outright lies like "blowing X away in every dimension" and when they cannot actively lie they seek to prevent anyone else from telling the truth.
Witness the NDA's that AMD wants to prevent any honest journalist from telling the truth about how Phenom's perform. Witness the lie of a "launch" event where AMD gets to control all the benchmarks and parameters. Witness the insanely tight control in not giving any review sites the ability to have a Phenom chip for more than a few days. (Well, that might also be because AMD was literally incapable of MAKING the chips so there were none available... never ascribe to malice that which can better be ascribed to abject incompetence).

The above points apply to ANY company, this time it was AMD. In the future watch for the same warning signs. Companies that don't overpromise and have nothing to hide are the ones you want to buy from and invest in, not the ones with the the most air-headed bimbo marketers and the psychotic complexes.

Axel said...

Lol Scientia just deleted my post that wasn't a troll in the least. All I said was that he's probably busy preparing his next essay and his outlook on AMD for 2008 is sure to be negative because there's no way he can spin the simple facts:

- K10 is significantly slower than Kentsfield clock-for-clock.
- K10 will not see 2.4 GHz until Q1 2008 at earliest.
- AMD have nothing on their roadmap for the next 18 months that will pose a threat to Intel on the high end of the market.

The bottom line is that AMD in their current form are screwed and even the fanbois are starting to come around to their senses.

Ho Ho said...

He did not only delete yours but many others. At least around 15 posts are gone and he has turned on moderation.

Anyone wants to bet his excuse will be that the TLB bug is lowering K10 IPC by roughly 10-15% and things will improve once it gets sorted out?

Anonymous said...

"Anyone wants to bet his excuse will be that the TLB bug is lowering K10 IPC by roughly 10-15% and things will improve once it gets sorted out?"

Actually I'm leaning toward the "we're listening to our customers and they want us to focus on energy efficiency and mainstream performance" excuse that they used on Barcelona. It will be a bit tougher card to play on Phenom as the folks buying quad core and looking to a spider platform with up to 4 graphic cards, tend to be a little less focused on power and well the TDP's of Phenom really aren't that great. That said, with a couple of links to some PhD dissertation on SOI, throw in a few "actual power consumption" metrics and it could play! (with some effective use of comment moderation to eliminate any counter arguments)

My second choice would be the good old "the compilers are optimized for Intel" excuse - it has been a little while since he's rolled that dog and pony show out!

It is rather amusing that despite his constant claims of not being a fan and just analyzing things, how there really has not been a negative article on AMD on his blog - and if you look real carefully any negative point about AMD is almost immediately followed up with well here's a better roadmap or here's another area where AMD is better.

InTheKnow said...

What architectural features over K8 really matter in dual core and mobile space?

Why, SSE5 of course! What planet have you been on? Don't you know that is one of the top developments of the year? LOL.

Seriously, I seem to recall reading something about dual core K10 being on the roadmap. I'm assuming that This includes a reasonable reduction in die size. Of course given the way the rest of the K10 thing has played out, that is an assumption that is probably way to generous.

Anonymous said...

"Why, SSE5 of course"

SSE5 is not part of K10 - that will be either the shrink (unlikely or a small subset) or the next architecture. Most important development of 2007 doesn't mean on products in 2007!

"Seriously, I seem to recall reading something about dual core K10 being on the roadmap."

It was orgininally Q4/Q1'08, it is now Q2'08 (plans subject to change of course).

I hear AMD is also preparing to take advantage of the "native" tri-core design and will sell a "unique" dual core design with one of the tr-cores disabled! Oh wait you mean the tricores are just aquad with 1 core already diabled - well this new dual core design should help yield/bin splits even more!

BTW - can someone help me out, when AMD said the most anticipated launch of 2007...was that Barcelona or Phenom (ummm....I mean the Spider platform as there is a need to de-emphasize the processor)?

How to botch a launch:
1) Promise the moon...40% better... CHECK!

2) Release parts at anywhere from 3-6 speed bins below expectations... Barcy 1.9 (or 2.0 if you count vaporware) vs 2.6 expected, Phenom 2.3 vs 2.6+(?) expected.... CHECK

3) Price the parts below your competitors existing parts in order to maximize the de-valuing of your new processor brand... CHECK

4) Completely botch the thermals and release the low speed bins in the upper bin TDP's.... CHECK

5) Fly 'journalists' to a remote location and have them test hand picked parts that you don't intend to launch. Oh and then don't tell them that until just prior to the launch and they've written all their reviews...CHECK

6) Make up creative excuses for underperforming parts (Barcy - we're focusing on energy efficiency, Phenom - we were surprised at the last minute(?) by a TLB bug)....CHECK

7) Have VP of marketing leave days before launch to steal some of the press coverage away...insist the timing has nothing to do with the launch....CHECK

8) Release no benchmarks of your own to show people how good (or bad) the product is.... CHECK

9) Release some (pretty good, BTW) overclocking SW as you know the top bins are nowhere in sight. This will incentivize the enthusiasts to buy the initial, cheap, parts rather than waiting for the higher price parts in order to minimize the profits.... CHECK

Anonymous said...

Dementia has enabled comment moderation. Damn chicken.

Anonymous said...

"It is rather amusing that despite his constant claims of not being a fan and just analyzing things, how there really has not been a negative article on AMD on his blog - and if you look real carefully any negative point about AMD is almost immediately followed up with well here's a better roadmap or here's another area where AMD is better."

exactly. I am still waiting for 2009 where Intel will be "painted into a corner" and have to overhaul Nehalem to incorporate SSE5, which as we all know is the "top development of 2007".

Also, I am eagerly waiting on AMD in 2008 since "It is also clear that AMD will be able to convert to 45nm in about half the time that it takes Intel."

Of course, AMD can then "choose" to do High-K or not, just like how AMD "chooses" to supply the bulk of the market with low priced low performance parts (because that is where market demand is).

ah, AMD released overclocking software for their Phenom, I see they've just "chosen" to recapture the enthusiast's mindset again.

SPARKS said...

In The Know,

“If you cut K-10 down to a dual core design and pitch it at mobile, I think you might have a decent product. It's not there yet, but I think that AMD could get there eventually. The big question in my mind is can they get there before K-10 no longer matters at all.”

Man they jumped all over you. I must apologize; however, I tend to agree with all the previously mentioned responses to your post. They took the words right out of my mouth and then some!

It seems, not only has AMD backed themselves into very ugly financial corner, but they backed themselves into a hideous technological one, as well.

In fact, finding what do with this broken process/architecture to make it a viable, sellable product, against an older product, has put them in a quagmire. Can a dual core K9 @ 2.6 Gig out perform a dual core K8 at 3.2 Gig? It doesn't mater, really, as they are now competing with themselves as opposed to competing with INTC.

“a decent product”, in your words, is the ONLY thing they have, presently. And, that still remains to be seen.

SPARKS

SPARKS said...

“Keep in mind while the typical 30-40% die size reduction is significant, the added wafer cost of the new technology and poor yield and/or bin splits can quickly wipe out/offset any of these gains if the technology is not ramped properly. Throw in immersion litho (which, surprisingly to most, is more expensive than a 2 pass process) and any perceived 45nm gains can quickly evaporate. Throw in the fact that there will be little/no gate oxide scaling and you are likely not to even get significant performance gains (the only other reason to do a shrink)”

GURU, you never cease to amaze me how your technical prowess, taken in a well structured, organized way, all boil downs dollars and cents (sense?), made palatable for the simple minions such as myself.

Brilliant, simply brilliant!

SPARKS

JumpingJack said...

Take a look at Scott Wasson's commentary on the 'Spider' launch:

http://www.techreport.com/
discussions.x/13677

Wow... I hope AMD is reading. Very much in line with Anand's commentary in his Phenom review.

InTheKnow said...

Man they jumped all over you. I must apologize; however, I tend to agree with all the previously mentioned responses to your post. They took the words right out of my mouth and then some!

No worries. If I had wanted weak sycophantic agreement with my every statement I could make pro AMD posts on another blog.

That said, I still think AMD has to get into the mobile space in a bigger way to get their ASPs up long term. 4P+ will only buy them a few quarters before Nehalem kicks them out of that niche too. So it isn't a matter of what I think AMD should do. It is more a matter of what I think AMD has to find a way to do.

They have to find a way to be more than a bargain player in mobile and they have to do it in less than a year. K-10 is their only shot, because Turion will never be more than a low end product.

Anonymous said...

"4P+ will only buy them a few quarters before Nehalem kicks them out of that niche too."

I've read (I'll try to find a link) that AMD ships about 5% unit volume as server (and that's all server segments). Factor in 1P and 2P are still the vast majority of server shipments (despite folks seeming infatuation with 4P+ performance) and this market is truly a niche market - sure the margins are good but with low volumes...who cares if that is the only thing you can control? This is not a big enough market where you should spend years focusing you're design and then try to waterfall, in an unoptimized fashion, to the other 99% of the x86 market as opposed to targeting your design at the main volume of the market and adapting from there (like Intel did with Core2)

Also with increasing performance in the 1P and 2P space and the move to multi-core, is the market for 4P+ likely to grow very fast?

You are dead on about mobile - that is where the cash is made these days (and will be in the future). Unfortunately AMD has a decent low end solution which is gaining traction, but they don't seem to have anything in the mid and high range segment. I don't think K10 is the answer - they need a new solution - I don't know if their upcoming mobile solution (forget the codename) will be more of a tweak or truly a change.

I still think AMD's best strategy to raise ASP (for them, not the consumers).... is to simply raise ASP on their current products. Suppose for example they raised the price of EVERY CPU by $10? They ship what,~13mil CPU's/quarter - that would be an additional 130Mil going directly to the bottom line. Would a $10 price hike impact their maret share? Now what if they did the price hikes a bit more intelligently (as a percent of price instead of a flat increase)?

This would leave Intel in the decision of whether to hold at current prices and ramp production in an effort to take more market share or raise ASP's themselves to get more margin. I think Intel would likely slow down/halt all price cuts and fatten margins to make the stockholders happier (and in theory be able to put more money back onto R&D). Again this is not best for the consumers, but is a $10 price hike on a part you buy what, once every 2-4 years, a big deal? (if it was done intelligently, it would have even less of an impact to the average Joe). And if you think this is screwing the consumer is this better or worse than having a permanently wounded AMD struggling along?

The strategic question for AMD is: is it best to keep battling for market share and continue to take losses (or at best eke out a small profit) in the hopes that one of their future products pulls another K8? Or is it better to end the price war, get "healthy" (relatively) and then re-focus on their core expertise which is design? This is why I think Hector should go - it is clear what his choice is.

AMD at least has a shot at winning or competing in the design game, I don't think they have any shot at winning in the manufacturing game. I really think they are deluding themselves if they think they can compete with Intel on cost and manufacturing. They can continue to spin APM, SOI, immersion, etc. but really these things have nothing to do with manufacturing prowess.

Giant said...

Well, well, well. $5.6bn and falling. $200M to go until they're worth the same as what they paid for ATI.

Ho Ho said...

Don't you just love it when people make fools out of themselves? It is as if they are not even trying any more.

Anonymous said...

abinstein is a coward as is scientia.

Hiding from the inevitable.

They are PWND

Anonymous said...

Some say the a person's true character is revealed when times are the toughest.

It is easy to be magnanimous and appear even-handed when things are going well, it's how you handle things when things aren't going well that tell you a lot about someone.

Just a thought...

Anonymous said...

Oh man this is too funny.
AMDZone crashes just in time....oh what a coincident. What idiot doesn't keep backups of their website? LOL.

They claim the mobo went out, well get a new mobo, reinstall the oS and put your back up on and voila, the site is up.

oh and that Sci character.... is no where to be seen. LOL. I knew he was gonna take a dump somewhere and never show up to clean up the mess.

What a douche.

Anonymous said...

This blog is dead. Long live Intel

Anonymous said...

Oops. I meant Scientia's blog is dead :-)

Anonymous said...

I hate to contribute useless post on here... but this could serve as an entertainment...

http://www.amdzone.com/index.php/Forums?func=view&id=129&catid=6&limit=10&start=20

Look at the bottom of the page, when he made his remarks on Intel's quads...

...not to mention he's a moderator :D...

SPARKS said...

“That said, I still think AMD has to get into the mobile space in a bigger way to get their ASPs up long term. 4P+ will only buy them a few quarters before Nehalem kicks them out of that niche too.”

Hmmm, this reminds me of the wild African savanna, where the Lions, obviously, get the Lions share of the bounty, and the “Scrappy Little Jackals” eek a living on the remnants left by the Lions. The Lions are either they’re too full or don’t waste the energy to protect the scraps.

4 processors servers and low end laptops could be the pituitary gland and the anus, respectively, of the computing carcass. From a market perspective, it is, basically, much ado about nothing. As an enthusiast, I couldn’t give a flying frig what they do with this stuff. You will never see either one of these sitting on my desktop or in my gang box. There are millions more like me.

However, the’ Scrappy Little Jackals’ can be an annoyance when the Lions territory is encroached. I’m guessing, but I suspect, Nehalem in 4P and HPC will scale extraordinarily well. 2500 chips per wafer, so cheap fast laptops will be in the future, along with any other device that can support x86 architecture.

So ends another chapter of AMD’s free reign on the computing plains, further, it gets very, very ugly when a Lion manages to grab a ‘Scrappy Little Jackal’.

Yeah, new lows, new downgrades plus 9.80 a share, this is very, very UGLY. It will get uglier. But, we all knew this, didn’t we?


SPARKS

Anonymous said...

hahah Ghost has no credibility what so ever, infact I think he has negative credibility.

He's such a JOKE, I'm sure his old ass shat himself for every phenom review he read.

I love reading his posts just to get a laugh, NOTHING more.

According to him, if you run INtel, you're evil and a troll LOL

Tonus said...

Someone needs to point out to people that just because they say that having a bit of server share is enough to keep you going financially, that doesn't make it true.