The Bad Economics of a 65nm Monolithic Quad-Core

As AMD will soon release Barcelona, it will claim industry first to “native” quad core. Intel on the other hand laughs at the idea as bad economics and foolhardy for the sake of recognition. Producing a monolithic quad core doesn’t make economic sense – according to Intel. Actually even AMD once touted the advantage of a smaller die – ZDNET 2002. But who do you think is correct this time?

The only way to settle this is to try and compute the cost-per-die value of Barcelona against its main rival, the Penryn MCM quad-core. I once claimed that Barcelona costs $174/die to make while it will only cost Intel $53/die to produce a competing product. While it is impossible to prove claims that involve revealing closely guarded secrets such as Fab yield and defect density, it is still possible to demonstrate the impact die size and the native approach has on the economic feasibility of the product.

What we already know:
Barcelona QC die size: 283mm^2
Penryn DC die size: 107mm^2 (214mm^2 for a QC)

To compute gross die per wafer (GDPW), we’ll use the standard formula: {[Pi * (Wafer Diameter/2) ^ 2] / Die Area} - {[pi * Wafer diameter] / (2 * Die Area) ^ 1/2}. Edge exclusion equally ignored. This gives us the result:
AMD at 300mm GDPW = 211; AMD at 200mm GDPW = 79 (quad core dies)
Intel at 300mm GDPW = 622 (dual core dies)

I know that no Fabs are the same, but for the sake of this analysis, let’s assume that defect density at 0.2d/cm2 is equal for Intel and AMD. For Yield we’ll use the Negative Binomial yield model which is commonly used and apply a midrange alpha of 1 for both Fabs, assuming defects follows a gamma distribution. The formula for yield is as follows: [(1+ DieSize*DefectDensity)/alpha]^-alpha
This gives the result:
Wafer yield for Barcelona QC = 63.86%
Wafer yield for Penryn DC = 82.37%
This shows how random defects negatively impacts the bigger die.

Applying the yield against the GDPW this gives the result (GDPW * Yield):
AMD 300mm DPW = 134 ; AMD 200mm DPW = 50 (good QC dies per wafer)
Intel 300mm DPW = 512 (good DC dies per wafer)
Since Intel needs 2 dies to make a QC, each wafer can only yield half: 512/2 = 256 possible QC dies

Assuming once again for argument sake that AMD and Intel has equal Cost Per Wafer. This number is quite complicated as it involves Equipment costs, Depreciation, Direct Labor, materials and Fab utilization so we’ll leave the number equal for both companies to focus only on the impact die size on profitability.

Using the numbers above, in order for Intel to make a quad core MCM Penryn at $53, it has to keep its cost per wafer at $13600 per wafer ($13600/256≈$53). Now if AMD has the same $13600 cost per wafer, its cost per die becomes ≈ $100 at 300mm and ≈$267 at 200mm. Averaging the two if we assume equal loading at Fab30 and Fab36 will give us $184 per die for Barcelona.

The numbers I earlier claimed even comes off as conservative. If we even start to consider the added cost of AMD’s SOI wafer and additional metal layers, the numbers even gets worst for AMD. Again let me reiterate that the numbers above assumes a lot of equal parameters between AMD and Intel which is entirely the opposite in real life. Yet still, the message remains obvious, the impact of a 283mm^2 monolithic quad core on cost per die is massive. The big gap in cost per die between the two competing companies prove that Intel is correct in waiting for 45nm make the same "native" quad core move.

Update: Just in case anyone is wondering, Clovertown's cost is at $76. This is amazing considering Clovertown total die size (143mm^2 * 2) is roughly equal to Barcelona which is 50% more expensive.


Toshiba Goes Turion

From Reuters:
"Toshiba, the world's fourth-largest laptop PC maker, said on Tuesday it expects to put AMD processors in about 20 percent of the notebooks it sells in the United States and Europe".

Good news indeed if you're an AMD follower. Another major OEM brings AMD onboard to help increase AMD's mobile penetration. Bear in mind that Toshiba maintains a big chuck of the commercial business most of them are known to be very loyal customers since the late 90's.

I have always believed that AMD's mobile market share has yet to reach "thermodynamic equilibrium" - i.e., continue to grab marketshare to the point where natural market competition puts it to a halt. This is the reason for the rapid increase of AMD's market share in the mobile and server space a year and a half ago simply because they provided the volume while starting of from a very low percentage. Currently AMD's mobile share is limited by OEM acceptance and this skews the way AMD and Intel competes in price. Once Toshiba is fully on board we should see a more competetive mobile market.

As Toshiba will be introducing Turion in the mid to low segment, its easy to predict market share gains for AMD. The only draw back that comes to mind is the association of the AMD brand as the low cost alternative. Its not ideal but at its current state, I'm sure AMD is willing to play this all too familiar role.


An Outlook on Scientia's AMD Outlook

If you haven't done so, please go read Scientia's blog:

Scientia said: “It doesn't seem reasonable though to assume that general demand has fallen off as sharply as AMD's numbers so there must be another reason... The only other possibility I can think of is that AMD's cpu sales fell off because of delays in delivering chipsets…”

I find this bizarre. When did AMD ever have problems with chipsets? AMD announced the reasons why CPU sales were low, why invent another one? The only reason why AMD’s Q1 sales is horrible is because they stuffed the channel in Q4 which resulted to an artificially inflated demand. Naturally, since demand didn’t pick up the next quarter, AMD’s customers stopped ordering from AMD while they clear their inventory. Anyone who wished to understand how much AMD’s demand declined, all you need to do is average out the volume sales for the Q4’06 & Q1’07 to cancel out the effect of channel stuffing. Other factors also contributd like logistics and poor product mix. But never chipsets.

Scientia said: “I've seen comment after comment about AMD's R600 delay yet no mention of Intel's 965 chipset being two months late”.

That’s because AMD’s R600 isn’t only 2 months delayed and isn’t as insignificant as Intel’s chipset. The survival of AMD’s graphics business hinges on the delivery and performance of R600. I’m sure the same thing cannot be said about Intel’s P965. And of course nobody comments about Intel’s chipset delays because nobody is really interested about chipsets.

Scientia said: “I've now seen people proclaiming that R600 is a total failure when it has already been suggested in the trades that AMD's R600 orders will max out the capacity at foundry giant TSMC.”

Normally we declare a product to be a failure when its has seen excessive delays, broken drivers, poor performance and missing features. But never mind all that. Instead let's look at how the orders have maxed out TSMC's capacity which just so happens is the ideal state of a foundry.

Scientia: Without doubt though, the good news for AMD is mini-DTX”.

AMD is creating more market segmentation in the desktop space where margins and market share have been declining year over year against mobile. How is this good news again? It looks like the only reason why AMD is pursuing this strategy is to create a niche for its uncompetitive desktop offerings. Intel isn’t interested and it appears like a lot of mobo makers are either.

Scientia said: “So, AMD is just going to have to bite the bullet until Q3 when things should improve. “

This statement is highly speculative without any real performance benchmark from Barcelona. Not to mention its very low volume shipment for the rest of 2007. AMD themselves said K10 won’t contribute to AMD’s bottom line in 2007. Anyone who suggest otherwise is fooling themselves.

Scientia said: “The chipset and graphic sales should be up by then and AMD should be fully anchored on the desktop with mini-DTX and DTX. “

This is a classic example why Scientia get’s everything wrong (see his previous predictions). He appears to have the tendency to heighten the impact of some of AMD’s insignificant business strategies and at the same time downplaying everything else that truly matters. Like Intel’s quad-core offering now becoming mainstream for example. At the moment all I see is a pending disaster with the R600 who’s fast becoming the media laughing stock and unsupported desktop form factor.

Scientia said: AMD's position does seem fairly good going into 2008 since it still insists that 45nm is on track and will be ready six months after Intel's."

Intel is ahead by a year as it will start loading 45nm next quarter to be able to ship products by Q4. AMD proclaims 45nm readiness 2H'08 (Q3-Q4'08). Now isn't that a year?

Scientia said: “So, AMD could potentially improve its volume and revenue position substantially by end of 2008."

I have seen nothing from the lengthy blog that suggest any concrete plans to turn things around. Barcelona is once again over emphasized. The same prediction that has led Scientia to incorrectly estimate AMD’s last two quarters is only repeated. No need to guess the out come of his new prediction.


AMD Explains the Fragging

From uberpulse:

AMD explains why 16Core Barcelona got fragged by an 8C Clovertown:

"The objective of our demo was to show performance scaling from our current dual-core processors to our upcoming quad-core processors within the same thermal envelope and drive home the point through a real-world demonstration that customers could expect to see 2x the performance without an increase in power consumption. "

Sounds like disaster control to me. Assuming that AMD is telling the truth and did intend to show performance-per-watt scaling, then I don't really see the value of the demo. The fact still remains that Intel's V8 showed better performance using less silicon. It is great that K10 scales well against K8 but if Barcelona comes out as a non-performer, who cares? Nobody is interested in comparisons to an already beaten product.

The question AMD deperately needs to answer is why is it that the best they can come up with is a crippled Barcelona while Intel's Penryn looks stable and ready to go.

AMD also said: "One thing to note, the system we showed, while it was a 4P, it was running only 6GB of memory".

I'm sorry, this is really a lame excuse. Why cripple the benchmark? I never heard of anyone running a benchmark and not optimise the setup to get the full potential of the system. Or maybe AMD can't afford the extra memory due to cost cutting.
[thanks to the anonymous poster for the link:]


Unreleased Barcelona will be DOA

The company ran a demo comparing the performance of two four-socket servers, one using the quad-core Barcelona chip and one using a dual-core Opteron chip. The demo measured the performance of the chips on an imaging benchmark called POV-Ray, ...
The quad-core chip processed about 4,000 pixels per second in rendering the image, while the dual-core chip could only hit around 2,000 pixels per second. Allen said Barcelona was not running at the fastest clock speed that will be available at launch, although he declined to specify the speeds that will be available... AMD did not test Barcelona's performance against one of Intel's quad-core Clovertown Xeon processors, but Allen said Barcelona "will be the highest-performing x86 chip out there. It will blow away Clovertown."

Remember the number: POV-Ray at 4000 for a 16-core 4S Barcelona config (4Sx4core).

To compare that number, check this Intel V8 Review: http://www.xbitlabs.com/articles/cpu/display/intel-v8.html
Intel's 8-core Clovertown scores 4677 p/s in POV-Ray.

UPDATE: From Uberpulse: "Why do you need 16 cores, when you can do better with 8. Our 8 core system is 30% faster than the 16 core machine AMD showed to the press yesterday. I just don't understand how they can claim to be 40% faster", said Francois Piednoel, an Intel engineer present at the show.

It takes Intel half the number of cores to beat Barcelona! If this doesn't spell disaster for AMD, I'm not sure what. Forget about competing with Intel. I just hope that AMD releases Barcelona soon enough to at earn at least something to pay for the bancruptcy fees.


AMD`s Hype-per-Watt Leadership

From Digitimes:
“While the Puma platform looks set to include technological features which will improve performance and battery-life for notebooks; according to AMD hardware based on the new platform is not likely to hit the market until mid-2008, i.e. one year from now. While it is nice to get a preview of upcoming products and technology, with AMD's recent track record of confirmed (R600) and speculated (Barcelona) product delays, and poorer than expected performance (HD 2900 XT), it is questionable whether, when (or dare we even say if) it is launched, Puma will meet the hype AMD is currently trying to generate though these early announcements”.

A tide of frustration and scepticism is growing against AMD. This company is getting too good in making PowerPoint presentations while becoming worst in releasing products on time with the promised performance. AMD’s “proof to claims” ratio for the last 2 years is either a very low decimal number or zero. AMD may have achieved overall success is the last few years but by the way it has been behaving, it appears like it hasn’t put down the party hat. The exact opposite can be said about Intel with its product pull-ins. As if Intel is trying to compete with itself. We’ve seen the public benchmark of Penryn and soon, Harpertown – both 2008 releases while AMD’s 2H’07 product, one that they dreadfully need to survive is nowhere to be seen. With the unflattering product reviews of the 65nm Athlon X2, QuadFX, R600 & HD2900XT, it is easy to draw a conclusion as to why AMD is afraid to show any real benchmarks.

As AMD is desperately trying to come off as the innovator in the industry, they had no choice but to pre-empt Intel’s Santa Rosa product release with their very own “me-too!-Centrino-copied” mobile platform. The big difference is Santa Rosa was announced sitting inside an OEM laptop, Puma was announced on a PowerPoint slide. Luckily for AMD, they won’t hear Intel shooting its mouth about someone copying their strategy. It’s not because AMD copying Intel is status quo, it’s just that everyone knows being a year behind is embarrassing enough.


Otellini Hints Why In The End, There Will Be Only One

Otellini: "Chip manufacturing gets harder and more expensive all the time. It's basically the laws of physics and the laws of economics at play. To build a modern semiconductor plant which uses 300-millimeter wafers, and state-of-the-art lithography with designs measuring 45 nanometers, costs close to $4 billion. Then there are the tools inside it, which amount to close to a billion dollars, per generation of technology. That means you need to generate $4 billion to $5 billion a year in revenue out of it. There are not many semiconductor companies that are $5 billion or more in revenue. So economics leads many companies to collaborate. "

AMD currently generates roughly $5B annual revenue but runs 2 Fabs. It is clear that the reason why AMD needed to capture 30% marketshare was to generate enough revenue to support 2 Fabs. Having failed that goal, it will be tough for AMD to return to profitability without significant sacrifices. We've seen a few adjustments, one of which was to halt CAPEX on 300mm upgrades for Fab30. This will affect the company's long-term competitiveness in cost and still this isn't enough to turn the company around.

Otellini: "No. 2: The laws of physics mean chip making gets harder and harder. We think we have some breakthrough technology at 45 nm. We think it will be harder for people to do that than it was in the past at, say, 65 nm or 90 nm. Our lead over the competition may extend with this generation and probably extend a lot after that. This is not a macho thing. It's all based on sheer economics. "

AMD's reliance on Foundries to support volume needs at the low end is not a viable long term solution to compete with Intel. The diverse set of customers Foundries have forces it to focus development on process shrink and less on transistor performance for logic devices. As Intel rolls over its volume runners to 45nm, AMD's foundry made processors will look 1-2 generations behind. These single core 90nm, foundry made CPUs will never command a high enough price to make any business sense. AMD can go on and continue the price war with useless processors but AMD needs to understand that the quicker they commoditize CPUs, the quicker their demise will be. We've already seen what the lastest price war has done to AMD's profits - or lack of.

A Talk With Paul Otellini.


Why the AMD Stock Upgrades?

"ThinkEquity analyst Eric Ross, who boosted the stock from "sell" to "buy." Ross lifted his rating based on what he said appear to be strong orders from Dell ..."

AMD is receiving upgrades from different analysts in the past few days, mostly from a “Sell” into a “Buy" rating. This should be pleasant to hear for a few AMD fans, particularly the hypocrites who have the tendency to believe analysts only when they have something good to say. The positive news this time focuses on AMD’s inventory which appears to be on a recovery. While this may be one good step away from AMD’s quagmire, it isn’t out of the woods yet. A stock upgrade can make sense considering how much it was beaten down from its 52-week high.

The issues AMD reported last quarter was primarily about lower ASPs and significantly lower volume sales. With such problems, easing of inventory does not necessarily point to a recovery. In fact the opposite occurs whenever inventories are cleared. ASPs fall further to move products quickly and shipped volume decline to allow time for market consumption. In fact, one can assume that this is precisely what we saw last quarter as AMD ‘allegedly’ stuffed the channel in Q4’06 and suffered for it the next quarter. AMD may be genuinely on the path of inventory recovery this quarter as its last guidance points to a flat to slightly better Q2. AMD should see more stability in pricing and customer orders from here on. This is definitely good news for AMD but the nagging reality of low ASPs and low volume remains. We all know what AMD needs to do to correct this and rumours of delays, bugs and lower speed bins doesn’t sound very encouraging for a possible upside.


Making Sense of Intel's July Price Drop

Intel’s recent announcement of another round of price drop in July must feel like another stake through the heart of AMD’s struggling business. If AMD lost a massive amount of market share last quarter due to Intel’s dominant product portfolio, for example a C2 Quad Core at $500, one can easily expect the hurt it will do to AMD’s if that same product drops half price. Without an existing competitive product when Intel’s price cut arrive, it’s easy to predict another decline in revenue and ASPs for AMD for the 2nd half of the year.

Without decent competition and a healthy demand for C2’s, Intel’s change in pricing can only be due to a healthy process resulting to an increased supply of top-end CPUs. Always remember that Intel prices it products at the point of ‘maximum revenue potential’. It is ridiculous to even consider that Intel is reducing prices across the board in response to a low-volume competing product that is only going to ship at the late end of the year. This reasoning would only be valid if, 1) Intel reduced prices only on directly competing products, 2) if Intel is dumb enough to reduce prices months in advance of competition and lose revenue.

There is also an increased chatter of a possible 45nm ramp pull-in from Intel. 45nm material suppliers have already increased guidance for Q2 based on orders from an ‘undisclosed’ customer. It doesn’t require a lot of thought who’s the only Fab running 45nm this year. Pulling in the price drop from the original schedule of October to July further reinforces that Intel is executing well ahead of schedule. If Intel has started loading 45nm this quarter, counting 15 weeks from wafer starts to customer built, I’ll leave it to your imagination what can be done at the start of Q4’07.


AMD's 3 Steps Forward and 4 Steps Back

I've decided to post the 2005-2007 Semiconductor Leaderboard for future reference. Quick note: Intel appears on track to get greater than $36B in 2007 while AMD slides down back to 17th place cosidering Q1'07's figure includes ATI sales.


Why ATI Sold Itself


"It is slightly off tradition that the GPU company's flagship product sails off not to meet the flagship of it's competitor, but one target lower".

One has to acknowledge ATI's brilliant move to sell while they realised first hand that they were going the way of the graphics dodo -3dfx. Dwindling sales, massive losses, and now a delayed product launch not even close to competition. It's hard to imagine how AMD can still say they are the smarter choice. Clearly the worst you can do right now is couple an AMD CPU with an ATI graphics.

An Erroneous Outsider's View of the uArchitechture Design Process

This a commentary reaction to Scientia's latest blog:
How we Got Here...
"Since the separate Penium M line has now been abandoned in favor of the C2D Merom, we would seem to need one less design team. However, it is hard to believe that Intel which has always had three design teams is now down to just two: the Itanium and the C2D line. "

The history lesson is fine. The mistake is trying to tie-in R&D teams to released products.
The whole idea of trying to determine intel's design teams and design life cycle based on product releases is overly simplistic and completely wrong. Again, this is a common mistake of an outside observer.

There are multiple levels and teams of R&D affecting a product during its discovery, design and manufacturing phase. To say for example that Intel has "3 microprocessor design teams" is inaccurate. One can accurately say that intel has 3 microprocessor design integrators - which will be comprised of several design teams working on different architectural enhancements.

uArch enhancements done by pockets of R&D teams aren't product specific. uArch/ucode discovery which occurs 5-7 years before the technology is applied isn't product specific. For example, there are teams working on improving the ALU or the Scheduler and there are teams working on improving the integration of both. . There are different design teams working of different technology paths - and again, it isn't product specific. The same is true in process R&D. There are several R&D teams finding out solutions to problems we'll encounter 15 years from now. Even then discoveries are handed over to another team for development and then another team for deployment if feasible.

What is visible to an outsider like Scientia is the R&D team that responds to market conditions and tries to release a product based on demand and competition. This is just a tiny fraction of Intel's (and I'm sure AMD too) overall R&D headcount. This development/deployment team is defined based on what features/enhancement the CPU will have and is in fact very late in the R&D phase.

As for Intel's tick-tock strategy, what it means is that the number of product teams - design integrators are doubled. Instead of waiting and lumping all R&D discovered improvement in 4 year cycles, the idea is to leverage on R&D quicker and implement discoveries in an overlapping manner. 2 product teams in a 4 year cycle will result in a product family released every 2 years. So every year intel will either release a new product family or will move to a new process node.


Centrino Pro - Miles away from competition

Intel released its newest mobile platform this week codenamed Santa Rosa. Quite a lot of IT proffesionals and some of them I know personally are waiting for this release and should spur a small jump in corporate mobile sales.

Santa Clara (CA) – Intel today launched four new Core 2 Duo processors as part of the much anticipated “Santa Rosa” platform, which not only enhances the Centrino platform core components, but offers completely new features such as optional flash cache as well as wireless broadband capability.

Is it me, or does it looks like Intel is miles away not just in promoting the Centrino brand but in mobile innovation as well. All AMD could do in their press release last week is claim to have similar capabilities.


Scientia's Poor Attempt To Appear Unbiased

I suggest you go over Scientia's blog and have a good read:
Scientia said:
"I constantly get accused of being slanted/biased/pro/fan in favor of AMD. So, I'm going to try to see the world through the eyes of an unabashed Intel enthusiast. This is not an easy point of view to maintain though because Common Sense keeps getting in the way."

As you can see from his first paragraph, he doesn't get even past the 2nd sentence while quickly abandoning the attempt.

While his common sense tends to get in the way with his poor attempt to argue for Intel, mine was laughing at the trivial arguments raised. If anyone wishes to make a case as to why AMD may be doomed, all you have to do is look at AMD's conference call and listen to the questions that were raised. Read the guidance report from major investment banks. Read the industry and market reports from Gartner or Mercury and watch where the trends lead to.

AMD is back to below 80% market share. AMD's entire product portfolio is outclassed and outsold by Intel's product. AMD is making huge losses due to lower volume shipment and ASPs. AMD's $900M inventory continues to grow with a disgruntled channel and a wrong product mix. AMD was forced to cut costs and stop critical Fab upgrades. AMD and was forced to take in more debt and restructure its broken business model. The real kicker now is that AMD doesn't seem to have a concrete plan to turn things around.

Is AMD Doomed? By any normal standard; revenue losses, borrowing money, restructuring, its easy to claim that AMD has already lost. If we even apply Intel's standards on AMD, it may appear that AMD has been DOOMED for quite some time now. Do not be influenced by the low standards the AMD fanbase apply to their beloved company. AMD is struggling like never before and it is only a heartbeat from closing shop.


NVIDIA and Intel Cooperate Against AMD

"Chicago (IL) – Nvidia apparently is looking for a closer relationship with Intel and release some of its graphics technologies in exchange for an improved cooperation with the world’s largest processor, graphics and chipset manufacturer."

Common sense requires that NVIDIA work with Intel to negate any advantage AMD has with its shared CPU and & GPU technologies. Intel on the other hand is quite open to helping out anybody that can hurt its main rival AMD. While it is known that Intel is looking at including graphics capabilities in future multi-core CPU's, the high-end graphics solution is a market segment intel does not wish to enter. The margins and return on investment to fighting a low volume segment does not appeal to Intel. NVIDIA will be more willing to share some of its secrets to get free reign in the discrete market.

Intel and NVIDIA shouldn't find it difficult to work out a win-win strategy as they both take market share from the struggling AMD instead of each other. Maybe they can both teach AMD a lesson about getting something without necessarily breaking the bank.


Intel Business to Get Better

From the Wall Street Journal:
"According to Mr. Otellini, while Intel expects to have revenue growth in 2007 and 2008, he expects the bottom line to grow faster than the top line."

This doesn't bode very well with Sharikou's Q2'08 Intel Bankruptcy.

$5 Billion savings in 2 years is quite an accomplishment for an already lean company such as Intel. One big question that needs to be answered is whether the $3B savings next year involves selling Intel's flash business. Otellini may be in for a big fight with the Intel's "grey men" which includes the company founders, Andy Grove and Gordon Moore. If you know anything about the history of the semiconductor and integrated circuits you would realise how Intel holds the Flash business close to its heart. For the old Intel, there is just too much history and pride involved with the money losing division. But then again, Otellini runs the show while Grove, Moore and Barrett knows exactly change is necessary.

For AMD, this is another $5B that Intel can throw just to keep competition at bay. Not good.