It looks like we'll see Penryn out before Barcelona.
http://www.fudzilla.com/index.php?option=com_content&task=view&id=1351&Itemid=1 I
UPDATE1:
I hope the FUDZILLA haters can forgive me now with these additional links:
The Inq
The Tech Report
Anandtech
I can only imagine serious implications for AMD if this rumour is in fact true. The key to any technology business is TIMING. Forget about Cost or Quality. If the timing is off when demand picks up, also known as the "demand curve", the business is lost. The demand curve for Quad Core kicked off 6 months ago when Intel released Kentsfield and Woodcrest. AMD has paid a heavy price in server market share by having nothing to compete in this space when the high end demand shifted to Quad Core.
Delaying Barcelona will surely result to more market share erosion in the short term. But what is more significant long term is the loss of design wins AMD was hoping for this year to sustain its server business. As to the impact of the delay to Barcelona's eco-system we've already seen an example in Cray's fiscal warning. Do not brush of Crays announcement as insignificant. Cray had an axe to grind with AMD as it hedged a significant portion of its business on this product and felt let down. Making an earnings announcement 6 months in advance just shows the frustration it has with AMD's poor execution and this is something you don't see everyday. While a lot of other businesses will surely be losing money as a result of this delay, most of them won't be as vocal as Cray. But I can assure you they are losing money and they're equally frustrated. Hard lessons are being learned here about doing business with AMD... again.
UPDATE2:
I've started writing about AMD's shipping of Radeon 2400/2600 but i fell asleep. That always happens to me with pre-fragged product releases. I only wanted to point out how AMD doesn't do soft launches. I wanted to know how AMD would describe what they just did with R600. If it wasn't a soft launch and definitely not a hard launch, then what exactly is it? I can't wait for the next gobbledygook from Advanced Maker of Doublespeak.
6.07.2007
No Barcelona in 2007
Subscribe to:
Post Comments (Atom)
41 comments:
I generally like your posts, but a Fudzilla link? There's not much worse than the INQ, but if there is, it's Fudzilla...
The article said delayed to Q4, not Q1'08 so I don't quite understand your title. It's not like Barcelona was ever going to drive AMD's revenues in 2007, so the real impact will be in early 2008 revenues unless AMD can accelerate the crossover from K8 to K10 (this of course assumes K10 is better than K8!)
Short term market share (if you mean 2007) is not and will not be impacted by Barcelona - Barcelona at best case when it was on schedule and ramping as AMD indicated was going to account for ~5-6% of AMD's total 2007 CPU output.
Don't get me wrong a delay to Barcelona (if true) will hurt AMD strategically, hurt the stock price, and hurt AMD's credibility but a delay alone will not cause short term market share erosion. Now if it were cancelled or continues to underperform that's a different story.
The market that would buy Barcelona in 2007 will not change strategy if it is simply a 1 quarter delay. If there are other issues like performance or poor yield which causes AMD to attempt to jack up prices on them - then that may change the dynamic.
For some #'s - I believe AMD was hoping to be at ~50% server K10 at end of Q4. Assuming linear ramp (which is a bit crude), this means ~12% of AMD CPU's in Q3, and ~37% in Q4. in terms of total CPU's (and crudely assuming server is 1/3 of AMD's business in terms of chip volumes) that would put Barcelona at ~4% of AMD CPU's in Q3 and ~12% in Q4.
So I guess in theory AMD could completely lose this to Intel, but in reality a chunk of that would likely divert back to K8... (Keep in mind I'm ignoring desktop K10)
As you state at the end, the real impact will be the long term strategic impact and design wins, and I think the performance will be potentially much more of an issue than a 1 quarter slip. It's one thing to delay by a quarter if you know the product is good and it will sell. It's a whole different story if Option B (Penryn) is better, has a similar timeline and there is a roadmap behind it (Nehalem, Gesher).
If I'm a server vendor in 1P and 2P space, what do I have after K10? Fusion ain't playing in this space, 45nm shrink if anything like 65nm shrink will be minimal performance -especially as AMD will not be implementing HighK& metal gate on its initial 45nm process. So as a vendor I'm left with potentially a 45nm 8 core MCM? So if if the 2 companies are roughly equivalent (best case for AMD), how do you get design wins when your competitor has a similar product, a seemingly more viable product roadmap, much better capacity, and better recent history on exceuction.
Like i said before its much easier making powerpoints than actual processors and in this case AMD forgot that it was a processor company first and a powerpoint BSer second...............
Fudzilla? You're almost as bad as abinstein and Scientia and Sharikou. Go Intel ^_^
Well, it seems those who have been skeptical of AMD's smoke and mirror tactics since OCT '06 have been quite right. AMD screwed the channel, screwed the Investors, screwed the analysts, screwed loyal fans, screwed the press, and now have screwed their partners.
Hector RUINZ should be brought up on Investor Fraud charges. Henri is selling stock like an old bitch at a fire sale.
"Barcelona."
"40 percent"
"performance per watt"
"Performance leadership"
Horseshit, I say.
They've lied to everyone, and they lied all along.
They've got nothing.
SPARKS
http://www.extremetech.com/article2/0,1697,2143301,00.asp
Article states:
That 1.6-GHz figure is true; I received that number from one of AMD's partners, along with my contribution to the story: "disappointing". And that's news, if only because three of AMD's partners used that exact word in describing the chip. One partner used this term: "chaos". And these guys had AMD boards in their booth.
hey..
guess you're partially right about K10 being DOA.
Charlie
The ones floating at the show are so broken they are not worth benchmarking. They are stable enough to finish up platforms and BIOSes, which is what this round of samples was meant to do.
They are far from full clocked, half the FP/SSE resources are broken, and the memory controller is barely functional.
On the up side, when they get debugged, things will get better/faster, presumably by a lot.
On the down side, they are not out, not imminent, and Intel has a new product coming out.
-Charlie
this might explain why Barcelona's score was so poor (POV-Ray, Cinebench...)
I generally like your posts, but a Fudzilla link? There's not much worse than the INQ, but if there is, it's Fudzilla...
my post may imply that i believe the rumour, but if you read carefully, i never said that i agreed with it. Very controversial, i know. ;)
The idea of a Barcelona delay till Q4 is not hard to believe in my opinion. Fixing a new uarch with a large freqency gap from the target speed bin like barcelona (~40%) can take 6-12 months. So Q4 really it isn't far fetched.
"my post may imply that i believe the rumour, but if you read carefully, i never said that i agreed with it. "
My aplogies.
Still I would avoid any references to Fudzilla unless that article was linked to something else that is credible. I have less issue with negative AMD articles because that site is so biased that for them to print anything very negative about AMD they must really think there is some truth to it (whereas they will just about print anything from anyone that has positive AMD info)
"The idea of a Barcelona delay till Q4 is not hard to believe in my opinion"
I agree but your title says no Barcy in 2007 which would imply it's delayed until Q1? (Perhaps you meant no Barcy in significant volume in 2007?)
As for the fix for the clockspeed it really depends on the cause of the issue. If it is process related (something varying across die/wafer) that may take significant time to fix. If it is a critical speedpath issue this could be a quick or not so quick fix depending on how many speedpath issues there are (could be anywhere from 1 major stepping revision to multiple). The nightmare case is that if there is a critical speedpath issue it may be masking other speedpath issues which will only become apparent after they fix the first one (thus multiple steppings). If it is merely one or two paths it could be fixed in a single stepping - I would imagine at this point AMD is running multiple version of steppings in parallel in F36 to identify the issue.
Sorry for the tangent but AMD fans are really starting to get loopy:
Scientia (comments in his last blog):
"On the other hand, we both know that top air systems today match water systems from 2003. So, that air claim has been fudged a bit. To claim genuine air today you can't really count higher than a midrange air system."
So apparently when you talk about an OC on air cooling with Scientia you can only use midrange air cooling (Are the highend coolers using different air?) Apparently "genuine air" only for benchmarks on OC'ing folks... On a side not you can only use DDR as well as DDR2 was not available back in 2003.
"The 1.6Ghz chip is not AMD's current best. AMD has 2.4Ghz running now. You are still thinking that AMD cherry picks its best ES chips for demo systems like Intel does."
Ahh at last... AMD CHOSE to use a chip running 33% slower than one they had... Apparently AMD cherrypicks the worst clockspeed for their demos...
"Well, I agree that that will be one fumbled launch if AMD only manages to get 2.3Ghz out the door in Q4. I'm thinking AMD is going to at least get 2.5Ghz out in Q4 but they might do better."
Q4?!? Q4?!? Talk about lowering expectations, now apparently a sucess is 2.5GHz in Q4?(which was supposed to be part of the Q3 products). This was the same person talking up a 2.6GHz, 2.7GHz when doing his Penryn/Barcy comparisons. Holy spin doctor, Batman!
"No comparison. Intel was pushing its chip six months before it was released. In contrast, AMD has had the tightest security on K10 that I have ever seen."
Apparently Scientia is confusing "the tightest security he has ever seen" with the lack of functional chips. Tight security = no benchmarks, but barely working chips also = no benchmarks. It's just humorous to hear such flawed conclusions based on lack of information.
"AMD has 2.4Ghz running now."
Smell the roses, where the fuck were they demoed? I guess that's why 2.3 is on the roadmap!
"You are still thinking that AMD cherry picks its best ES chips for demo systems like Intel does"
Just like the 800MHz Conroe, the 3.33GHz Penryn, the 3GHz V8, right.
"I'm thinking AMD is going to at least get 2.5Ghz out in Q4 but they might do better."
An unsubstantiated statement with no evidence, surprising.
"No comparison. Intel was pushing its chip six months before it was released. In contrast, AMD has had the tightest security on K10 that I have ever seen."
AMD is not pushing Barcelona at all, not "blowing away Clovertown", not throwing around fuzzy "50-80% better than Opteron", not "50% better in FP", nope. I guess Scientia prefers FUD to fact.
"Apparently Scientia is confusing "the tightest security he has ever seen" with the lack of functional chips."
Scientia's problem is his desperation. AMD is badly beaten and is executing horribly that his need to defend it requires him to invent and exaggerate which ultimately leads to his inaccurate assumptions.
"Scientia's problem is his desperation. AMD is badly beaten and is executing horribly that his need to defend it requires him to invent and exaggerate which ultimately leads to his inaccurate assumptions."
I tend to agree, but he used to be a bit more objective - is it simply AMD's current situation that has changed things? (Or as AD was doing much better 6-12months ago he could afford to be seemingly more objective and still "push" AMD?
All this said I strongly believe K10 will be much better than people think - maybe not this year, but with some additional time I think it could be a relatively strong chip. Of course the question is what shape will AMD be in at that point and how much further will Intel be at that point. I do think reports of the demise of K10 is vastly exaggerated.
Why do you link to abinstein and scientia?
Why do you link to abinstein and scientia?
so its easy for everyone to go there...
Roborat, Ph. D. said...
Why do you link to abinstein and scientia?
so its easy for everyone to go there...
to see that there exist different type of fanbois, some claim to doctor, some claim to be unbiased, and some claim to be knowledgeable ... :)
They are censorers. There should be no place in the blogosphere for censorship.
Nebojsa at the Inq thinks Computex was a "debacle" for AMD.
http://www.theinquirer.net/default.aspx?article=40221
If the rumors about the B0 stepping being broken are true then it'll be another quarter for CPUs to be shipped to the OEMs and distis - let's say that is September, best case.
It looks like they will miss the xmas shopping season.
"If the rumors about the B0 stepping being broken are true then it'll be another quarter for CPUs to be shipped to the OEMs and distis - let's say that is September, best case.
It looks like they will miss the xmas shopping season."
I think folks need to calm down and wait for AMD to say something - if it really is this bad they will have to say something within the next month or so, if for nothing else then to prep Wall Street (and avoid a stock price crash). If it is on time and working well, AMD will also need to say something soon to avoid people who might jump ship shortly based on the numerous rumors.
Remember INQ is the site that published MORE THAN ONE article on reverse hyperthreading (I can forgive one article, but when you are posting something incorrect multiple times than that to me indicates laziness and the lack of desire/care to get things right)
So much for the the smoke and mirror campaign conserning Barcelona during the past nine months. Secrecy, Power Point launches,delays, et al.
The hammer drops, Ladies and Gentilemen.
http://www.fudzilla.com/index.php?option=com_content&task=view&id=1402&Itemid=1
"The hammer drops, Ladies and Gentilemen."
Not so fast. Fudzilla hasn't won any Pulitzer's yet.
Laying off former ATI employees has nothing to do with Barcelona per se. It has more to do with AMD taking on another company at the worst competitive time in its history, with a less than impressive product portfolio. AMD is being attacked by Intel on one side and NVDA on the other. Two battlegrounds to fight.
Good luck to them - this will not be the first such action coming out of DAAMIT. There will be more to come.
I think folks need to calm down and wait for AMD to say something - if it really is this bad they will have to say something within the next month or so, if for nothing else then to prep Wall Street (and avoid a stock price crash).
I think you have it all wrong. AMD has been talking and have been saying that everything is doing fine and right on schedule. The problem is their product samples and anxious customers are the ones to seem to be making contradicting noises.
"There will be more to come."
Your damned straight their will be. This was a corporate blunder of historical proportions. INTC beats the front door down while NVDA chops away at the back door. Well said! This tag team match will make WWF look like Sponge Bob and Patrick.
Sure, everythings fine, new platforms and on time. Looks great; as Richards, Ruinz, and Meyer rearrange the deck Chairs on the Titanic.
http://www.anandtech.com
/tradeshows
/showdoc.aspx?i=3006&p=1
"In the end, performance was absolutely terrible. We're beginning to understand why AMD didn't let us test Barcelona last month. It's not that AMD is waiting to surprise Intel; it's that the platform just isn't ready"
"None of the partners we talked to are really impressed with Barcelona"
"Continuing on the worst case scenario track, some partners don't expect to see 2.3 - 2.4GHz until Q2 next year"
"For all of us who have been crying for Barcelona benchmarks, you really don't want to see them"
This sums it up. Scaling is terrible, performance is terrible, heck, it's not even ready yet. AMD expects to launch this next month at 1.4-1.6GHz(forget the promised 2.3-2.6GHz launch)? Intel doesn't even need Penryn, K10 has already been pre-fragged by Clovertown. By the time AMD scales K10 to 2.6GHz, Intel will have Nehalem out.
Not to be outdone by Anandtech, here's overclocker's assessment
AMD: the next Cyrix?
http://www.overclockers.com/tips01165/
Someone please make Scientia come out of his self-imposed exile in denial
Someone please make Scientia come out of his self-imposed exile in denial
Don't worry about him. He'll come around. It's just that he's a bit slow to catch on. It took him up until May 2007 to post his 2007 AMD Outlook to predict that AMD is indeed in trouble. Nevermind that he furiously debated us about it late last year. "AMD in better position in 2007" - yeah right.
But i'm not the sort that gloats. lol ;p
"Someone please make Scientia come out of his self-imposed exile in denial
Don't worry about him. He'll come around."
Come on - he's still of the opinion that the share loss is nothing more than temporary (only 1 quarter doesn't count?).
This is also the same person that ASSUMED when Intel started picking up market share last year (think it was Q3) and Via shrank by 2-3% , it was simply Intel picking up Via's market - he failed to even examine the possibility that it was AMD picking up Via's share and the first sign of Intel starting to eat into AMD's higher end...
Though hindsight is 20/20 which theory regarding where Via's market share went now seems more plausible?
It is Scientia's recent refusal to accept any plausible explanation other than the one that puts AMD in the best possible light, which tells me he is slowly degenerating into Sharikou territory (I said into not AT).
The newest one is Abinstein - have you read his blog... he seems to have a decent background on architecture, but his latest analysis on yield and capacity is utterly laughable. He calculates AMD must have effectively double the yield of Intel due to his capacity analysis as he estimates Intel has an 8:1 capacity advantage (which he then drops to only 50% better in case some of his assumptions are bad). Of course the 8:1 ratio is based on his inability to even consider fab size as a variable, real 90mn to 65nm scaling, aggregate die size, dual core/single core mix to name just a few things....
If you read the comments his knowledge of capacity, yield, and market are hilarious! (Just don;t question his "conclusion"!)
There are really two issues here
1) The Q4 disaster where AMD oversupplied parts that no one in the end wanted - it's plausible that this was indeed a blip; however, the real issue is with
2) K10's revenue implications [or lack thereof] (meaning Q2 through Q4 and beyond). With AMD having to be forced to cut prices ahead of Intel's scheduled July price cuts, it just does not seem as if there is much hope for increased margin revenue from K10 SKUs (when I say K10 I mean Server and Desktop variants).
What's AMD to do? AMD does not do 'soft launches' according to Henri, so what do you do instead of a 'soft launch'? Well, I guess you delay it or launch sans features. Oh wait: didn't they just do that on the GFX side?
AMD earnings warning window is less than 3 weeks away.
AMD Fellow employee resumes are making the rounds and ex ATIers are fleeing to NVDA and Intel.
Barcelona is AMD's Itanic
"Barcelona is AMD's Itanic"
Come on folks, I like to bash AMD as much as the next person, but at worst (if/when AMD addresses clockspeed and/or yield issues) it will be a K8 with better power. Granted this is not exactly AMD's aspiration for K10 but it will be (somewhat) competitive especially in dual core space- I can't see these having the same yield/clockspeed issues as quad unless there is a monster design issue. Yes this case would really hurt in server space but lack of a quad would have virtually no impact in mobile space and mnimal impact in desktop space.
If quad turns out to be non-viable AMD could also try to swallow some pride and "glue" (MCM) the dual cores... not sure about how fast this could be slapped together given IMC complications, though. Or they could just wait to 45nm like Intel. Either backup plan is not an attractive one, but I can't see a scenario where AMD gives up on K10 short of AMD going belly-up, it's just a question of how painful it will be.
"AMD Fellow employee resumes are making the rounds and ex ATIers are fleeing to NVDA and Intel."
One corroborating data point
http://www.theinquirer.net/default.aspx?article=40257
Nvidia is finishing its complete triumph over results of internal AMD-ATI struggle, which is resulting in many of ATI people leaving or being forced to leave.
In response to
"lack of a quad would have virtually no impact in mobile space and mnimal impact in desktop space."
AMD needs the higher ASPs that it could presumably charge for the quads to offset the prices it will charge for duals. You can argue that they would make the diff in volumes but that is predicated on how much they can charge and with Intel price cuts to the bone to put pressure on AMD it really boggles the mind how AMD will make a decent profit out of this. Wal-Mart and the low end will just not be enough to carry them forward.
http://www.hardspell.com/english/doc/showcont.asp?news_id=785
Why bother supporting a dying platform?
You should consider linking overclockers.com. Good stuff.
"The demand curve for Quad Core kicked off 6 months ago when Intel released Kentsfield and Woodcrest. AMD has paid a heavy price in server market share by having nothing to compete in this space when the high end demand shifted to Quad Core."
I think this point also indirectly points out the non-trivial nature of an MCM approach. There were many fans (including Sharikook) that said well AMD should just slap together their own MCM with 2 K8's if they wanted to. Intel had experience on this (even prior to the P4 MCM days...) and while they bad mouth it as a glue job there is actually a bit more to it than that.
If it was so straightforward don't you think AMD would have done it to hold things over until K10? Especially as this would give them early learning as they indicated they will pursue this in the future.
With IMC, the MCM is not a quick "glue-job", if AMD (presumably) kept the same socket they would have to figure out how to handle memory requests between the 2 cores. While this can be engineered it is not a slap it together in a couple of months type effort. So while AMD was bad mouthing Intel's non-native quad design, I think they are clearly regretting it (didn't some AMD director say something like this in an interview?)
Just think how much better off AMD would be if they had a K8 MCM, were able to release the K10 dual cores (which shouldn't have near the issues of the quad parts), and then let those hold them over while they debug/fix the native quad design. And if yield and /or clockspeed turned out abysmally on quad, they could even shelf the design until 45nm...
By putting quad first (which in all fairness they almost had to do as Intel was eating their lunch in this space) they are holding back dual core K10's and a desktop refresh which is only compounding their problems... I would not be terribly surprised to see AMD rethink the K10 launch timing and start cranking out some dual core server/desktop parts while they work on the quads. The only problem is that AMD has been touting native quad so much this will be a pit oo a PR hit short term.
Either way the common theme is someone in management should roll for the lack of contingency planning... Can anyone else see the mgmt discussions:
OK, so we're going to acquire ATI by taking on debt and potentially take on more debt later (to the tune of 2Bil) to fund expansion and we will get all this back by delivering this killer K10 product.
In fact we will bank on this so much we will slash K8 pricing to get market share (at the cost of profitability and margins) so when K10 comes out we will be able to then just switch this market share over to a higher priced killer product and totally squash Intel!
Lone engineer: Err... we don;t really have functional Si yet, what happens if something goes wrong or if the product gets delayed due to technical issues.
(Laughter ensues)
Hey who's writing the native quad core for dummies book!
And the whip comes down.
http://www.xbitlabs.com/news/cpu/display/20070610234912.html
http://scientiasblog.blogspot.com/
The Eye Of The Storm
"Back in January, Randy Allen at AMD stated that K10 would be 40% faster than Intel's best Clovertown"
We all know what happened to that story.
"it will take a K10 more like 2.8Ghz to match a 3.33Ghz Penryn."
I love how Scientia pulls out random numbers.
I'm guessing you gave AMD too much credit (i.e. 2.5GHz K10 has 40% performance advantage over 2.66GHz Clovertown). I love how you're just making up random numbers to sound unbiased.
I love how Scientia pulls out random numbers.
<333
it will take a K10 more like 2.8Ghz to match a 3.33Ghz Penryn.
And where is your evidence to back this up Scientia?
Using what we know so far,
http://www.dailytech.com
/Quick+and+Dirty+AMD+K10+Cinebench
/article7574.htm
Cinebench completed the default benchmark in 27 seconds for the 1.6 GHz K10; 17 seconds for the Intel Xeon X3220. The Kentsfield Xeon was 58% faster with a 50% higher clock frequency for Cinebench.
So simplified, assuming the Cinebench is correct, a 2.4GHz K10 would be the equivalent of a 2.3 GHz Kentsfield.
This much more accurate than Scientia's "estimated" (or should I say random) guess.
The "UPDATE #2" is nice. Here's a brief summary over what happened:
November 2006, Nvidia launches Geforce 8800. Hard launch, immediate availability.
People wonder when R600 will be launched.
In January Henri Richard tells people (at the earnings announcement) that R600 will launch in Q1'07.
Q1'07 ends, Q2'07 rolls on. No R600. David Orton says that R600 was done but they were waiting a few weeks to launch the whole product line at once.
In April Henri Richard claims "We don't do soft launches!".
Then, in May, AMD finally launches R600. But you can only buy the HD2900XT!
Now AMD announces that they have shipped the boards to their partners. It'll still be weeks before you can buy one though.
Meanwhile Nvidia did a hard launch of both the 8800 and 85/8600 series and will hard launch the 8400 series this month.
In other news.......
http://www.tomshardware.com
/2007/06/12/vigor_force_recon_qx4
/index.html
AMD's 4x4 with dual FX-74 space heaters gets fragged in and out by a Dell XPS720 equipped with a QX6800.
Games,
http://tinyurl.com/2myrs2
http://tinyurl.com/3xaszn
Fragged by 67%
Audio,
http://tinyurl.com/2r88pc
Fragged by average 57%
Video,
http://tinyurl.com/2jukud
Fragged by average 57%
Their conclusion,
"Vigor Gaming's Force Recon QX4 won a single benchmark, PC Mark 2005's hard drive test."
Hardly a relevant benchmark and it was a very close call. On average the quadfather gets a whooping of 54% in all the benchmarks collectively.....
Which excuse will it be today.....
A) improper benchmarks.....
B) numa aware OS(scientia's favorite)......
C) improper system config......
D) old arch versus new arch.......
E) intel paid pumper site.....
Actually..
F) Intel copied us big time, so we're going to post some deceptive benchmarks to trick Intel into believing we're weak, so Intel won't have enough time to prepare for Barcelona's major fragging.
yomamafor2 said...
Actually..
F) Intel copied us big time, so we're going to post some deceptive benchmarks to trick Intel into believing we're weak, so Intel won't have enough time to prepare for Barcelona's major fragging.
wow, the last AMD hopeful... we don't see a lot of your kind around anymore.
"wow, the last AMD hopeful... we don't see a lot of your kind around anymore."
I could be wrong but the comment could be sarcasm?
I propose choices
G) AMD's designs are so advanced that today's SW can take advantge of them. Just buy it now as it will be "future-proofed"
H) Bad SW compiling (another AMD fanboy fav)
Was it PGI? Was it? That AIN'T FAIR, WRONG COMPILER, WRONG COMPILER!.... sorry I was channeling there...
Post a Comment