Friday, March 31, 2006

AMD eyes 30% of commercial market

AMD states that it's on fast track to grab 30% of the commercial market. Currently, AMD holds less than 5% of the commercial client market, but owns 58% of the retail PC market (desktop and notebook combined) in US.

Dell will be in a time of hurt, as 69% of its profit came from US commerical market.

AMD is on its way to becoming the world's largest CPU vendor by the end of 2006, with Intel a close second.

INTEL, Wall Street and Media Mute on Shelf Registration

We reported that INTEL filed a shelf registration for selling common stocks, debts and warrants on March 30, 2006. Interestingly, Intel, Wall Street and the media are totally silent on this. We emailed Eric Ross @ Think Equity(sell, target $16), Hans Mosesmann @ you-know-where, Sumit Dhanda at B of A(long 38 million shares, buy), James Covello at Goldman Sachs (long 61 million shares,bullish), seeking "analyst" opinions on the level of dilution, total radio silence. We emailed David Whelan and Maya Roney at Forbes, who wrote many expert pieces on Intel, they seem have suddenly lost their strong interest in Intel also, no writings appeared on Steve Forbes the former presidential candidate's influential journal.

We thought Intel's PR people might be on vacation, but no, we heard a lot of other noise such as Intel's value PCes, so they are still there...but in need of cash?

The silence is deafening.

Thursday, March 30, 2006

Intel lures consumers with VIIV feature

The INQ digs out some details of Intel's so called VIIV here, it discovers that the VIIV Instant On/Off feature is just turning off the monitor and muting the sound - what a cheap gimmick.

INQ also reports Chartered FAB7 is ramping as Opterons are on allocation.

AMD poised to exit 2006 with 55% market share (run rate)

Various news indicate that AMD is ramping up FAB36 and Chartered FAB7 as fast as it can, furthermore, 65nm products will be shipping as early as June 2006. Clear indication that AMD is gunning to take over 50% of the x86 market by the end of 2006.

Keep in mind, as of 4Q05, AMD took 21.4% market share with one single 200mm FAB, the FAB30. What it can do with two additional 300mm FABs is staggering. Previously, it was reported that AMD was ramping FAB36 to 13000 wspm. Now, assuming Chartered FAB7 contributes 5000 wspm for AMD64, we reach a 18000 wspm for the 65nm node. The capacity estimate is done again with the Wafer application from, which computes the number of dies off a wafer.

For this, let's use the biggest known AMD chip, the Socket 1207 Rev F dual core Opteron as a reference. At 90nm, the dize size for Rev F dual core Opteron is 220mm^2, at 65nm, the die size is thus 110mm^2. Other dual core chips such as Socket AM2 Athlon 64 X2s are 25% smaller, their die sizes are around 85mm^2. To be conservative, let's use a geometry of 10 mm x 9.5mm for the Wafer program. We get 690 dual core dies off each 300mm wafer.

The annual dual core die output at 18,000 wspm is thus: 18000* 12 * 690 = 149 million. With a very conservative yield of 55%, we get 82 million working dual core CPUs.

For FAB30, let's assume AMD sticks to 90nm and continues to produce the same CPUs for the lower end there, FAB30 thus contributes 50 million units. We reach a total AMD dual core CPU output of 82+50= 132 million per year at the end of 2006.

According to IDC, global PC units will increase 10% to about 230 million (with US PC units growing to 69.5 million).

132 / 230 = 57.4%

Intel has about a dozen production FABs, however, most of Intel's FABs are oudated. Only four Intel production FABs are capable of running at 90nm or 65nm*. To be more specific, Intel has only two 65nm production FABs and two 90nm production FABs. In fact, the $10 billion Itaniums are still stuck at 130nm, while the Xscale mobile chips are still at 180nm. For some chips, Intel has to rely on Taiwanese foundry TSMC. Due to Intel's sixth generation architecture(Bob Colwell, 1995), its CPUs needs minimum of 2x2MB of cache to have acceptable performance, the large cache requires large die area. For instance, the 65nm Presler has a 162mm^2 die size, the Conroe's die size is about 155 mm^2. Intel is planning to build additional FABs. However, since Intel has only $10 billion cash but 100K people to feed, its ability to add capacity depends on making profit or selling stocks, both of which are becoming harder and harder, as its products sit in the lower end of the performance spectrum.

AMD will widen its performance lead as we progress in 2006. Dual core Turion 64 X2 will be unleashed soon at 25 watts, then Socket AM2, Socket AM3, Rev F Opteron, Quad-core Opteron and the K8L. The hounds, the K-10 and K-11 are also on schedule. AMD will soon add a new set of extensions to AMD64 to keep the clone makers busy.

* Intel has two 65nm FABs (FAB12, FAB24-2), two 90 nm FABs (FAB24, FAB11x), seven 130nm FABs, and a bunch of 0.18, 0.25,0.35,0.50,0.70,1.0 micron junkies.

*revised to use a lower estimated yield for the 65nm node.

PS: There is a great secret on Intel's capacity, if you can see it, please post your findings in the comments area.
PPS: since the secret is now exposed, I modified the main text to include Intel capacity analysis.

PPPS: In case you didn't notice, Intel did a shelf registration with SEC today, which stated "Under Intel’s second restated certificate of incorporation (the “certificate of incorporation”), Intel is authorized to issue up to 10 billion shares of common stock". Two weeks before earnings, time running out.

Wednesday, March 29, 2006

DOE orders supercomputer with 24000 quad-core Opterons

The Energy Department's Oak Ridge National Laboratory will install a massive Cray supercomputer equipped with 24,000 quad-core AMD Opterons running at 2.6GHZ. The system will have up to 400 terabytes of RAM.

Tuesday, March 28, 2006

AMD advances in China market with alarming speed

According to a report published by Chinese online media NetEase (NASDAQ: NTES), there is a hot debate on true and fake dual core throughout China. In an online survey titled "Dual Core Debate" by, 65.52% of the surveyed favour AMD as the winner, only 25.05% choose Intel. 73.98% of the surveyed support AMD's dual core challenge for Intel. AMD has conducted a 3 month dual core marketing campaign covering major Chinese cities.

Regarding the recent Conroe benchmark at the IDF, PR manager of AMD China, Ms. Yanli Liu discounted it as a future Intel chip against AMD's current offering, she questioned the integrity of the test as it was an Intel setup done in a very closed environment.

The report claims that Lenovo, Tongfang, Asus, Shenzhou, Benq and MSI are AMD's top five allies in China, while Acer, TCL and Haier are in Intel's camp.

A researcher of the Chinese Science Academy recently lashed out against the shared FSB design of Intel's NGMA/CORE, calling it a major bottleneck. AMD collaborated with Dawning in implementing China's fastest supercomputer with the Opteron processor. AMD also contributed the Geode processor design to China's Beijing University and Ministry of Science and Technology in a joint development project.

Intel's recent atttempt to block China's WAPI wireless standard in favour of its own 802.11i standard has deeply angered the Chinese. There were Chinese reports that Intel invited the voting members of the standard committee to luxurious trips. So far, only Japan, European Union and South Korea governments have raided INTEL offices.

Banc of America may lose half a billion on INTC

Banc of America pumps INTC with a target price of $25. Doing our regular exercise, we found that B of A is a bagholder of 38.215 million INTC shares . With fast dropping revenue and products that are five generations behind AMD, INTC target price is $14- $16. My projection is operating loss for INTEL from 3Q06 onward. INTC will fall substantially after its April 19, 2006 earnings report. B of A is expected to lose at least half a billion bucks on INTC.

I will watch B of A's INTC holdings for the next quarter. If it dumps INTC, then this is a securities fraud called pump and dump. And we shall report Sumit Dhanda and B of A to the SEC.

Captain of the Itanic jumps ship

INQ reported that an Intel fellow who was the Director of Itanium Circuits and Technology, Sam Naffziger, has jumped ship to AMD. Naffziger joined HP in 1988. Before wasting precious many years on Itanium at HP then Intel, he was a technical lead for HP's PA-RISC. His recent work was on the dual core Montecito Itanium 2 CPU, which was stuck at 130nm process thanks to the wonders of Copy-Exact. More details here.

With his credentials in 64 bit computing, Naffziger should join the ranks of Grand Masters at AMD. Finally, AMD has got top designers from all major 64 bit server camps: DEC Alpha, IBM Power, PowerPC, IBM RS/6000, Sun UltraSparc, PA-RISC and last but not the least, Itanium. Maybe AMD can even pick up some good ideas from Itanic, such as strong FP performance. No other company on earth has such an impressive line of battle.

Monday, March 27, 2006

Goldman Sachs to lose $1 billion on INTC

Goldman Sachs' James Covello started another round of pumping for INTC, after Michael Dell cast his vote of no confidence on Intel's Conroe through Dell Alienware. James Covello said his mood was a bit more bearish on AMD and a bit more bullish on INTC. As we reported here, Goldman Sachs increased their INTC holdings by 18.1% to 61 million shares during the September 2005 reporting period. Those shares were purchased around $27, a $0.4 billion paper loss already. Da baggies dumped 7.76 million INTC shares later, but there are still a lot of shares to unload to innocent investors. Previously, on January 19, 2006, Goldman baggies downgraded AMD to underperform in the wake of AMD's massive 4Q05 earnings jump reported on January 18, 2006. Since then, INTC's target price has been reduced to $15, after it warned about a massive revenue drop for 1Q06. At that price, Goldman baggies will lose close to $1 billion bucks.

Let's report the crooks to the SEC.

Crooks belong in jail.

Friday, March 24, 2006

Michael Dell on Intel's Core Achitecture - Conroe/Merom/Woodcrest

David Kirkpatrick did an interview with Mickey on Alienware, and here is what Dell had to say about Intel's NGMA chips(Conroe/Merom/Woodcrest):

"Intel [recently] showed a bunch of new parts. They're making good progress, but that only means there will be better competition between the two leading processor suppliers, which is good for their customers."

David was thinking aloud : "I would not call that a ringing endorsement of Intel from its biggest customer".

AMD's Dr. Ruiz said previously in a BusinessWeek interview that Dell "have a pretty damn good idea where we are and where we are going". (Emphasis added)

Can the Paranoid Resurge ?

David Whelan at Forbe wrote "Only the Paranoid Resurge", it was masterfully written, I couldn't help laughing at the part about Henri Richard's Ferrari and Sean Maloney's Toyota.

However, the article missed some crucial points:

0) Intel's manufacturing capacity looks impressive, but its efficiency is very much in doubt. Intel has five 300mm FABs and seven 200mm FABs, its theoretical capacity is at least 15 times of AMD's. However, AMD took 21.4% market shares as of 4Q05, Intel can't produce enough chipsets to match its CPUs. In fact, Intel has to outsource work to TSMC and has to build new FABs. My estimate is that Intel's yield is only 1/3 of AMD's yield, assuming Intel's FAB utilization is at the same level as AMD's. As I estimated, with FAB36 ramping up, AMD will take 40% of the market exiting 2006.

1) Intel Israel did not design a new archictecture, it was basically Bob Colwell's Pentium III, an Andy Grove era chip. I can only guess when the Pentium 4 folks took the throne, the Pentium III designer had to exit. When the Israelis brought Pentium III back, the Pentium 4 folks left.

2) The 31 watt Sossaman chip is just Core Duo (Yonah) dressed up as a server chip, it's 32 bit only, HP doesn't want to look at it. Running 4 Sossaman cores with only 4GB memory is a waste of power - computational and electrical.

3) Intel's price cut seems to be localized to the China market. AMD did a 3 months true-Duel Core compaign in China, Intel refused to respond to a duel there, but cut prices whereever AMD roadshow propaganda machine went. AMD prices seem to be on the rise.

4) The article did not mention the AMD anti-trust lawsuit and the almost instantaneous effect on AMD's sales: right after AMD's lawsuit, Opteron sales jumped 89% in one quarter. The lawsuit and worldwide raids on Intel offices partially cleared AMD's road to fair market access.

5) No mentioning of Intel's dramatic change in branding and strategy, even the logo has changed.

Some comments:

Intel is a sad, aging empire in fast decline. An angry Grove protégé of the 386/486 age won't do any miracles. The time of Grove has gone forever. The torch of technological leadership is firmly in AMD's hands, which is five generations ahead. We are talking about a fairly newbie Israeli regiment armed with 1995 weapons against AMD's battle hardened legions, led by grand masters such as the DEC Alpha folks, the IBM Power folks, the SUN UltraSparc folks. It's not an even match at all. There is a massive technology gap and brain power differential over there.

The NGMA will be just another dead end. It's a hack that may bump up Intel's 1P performance to a level closer to AMD, but it will lead Intel further down the wrong direction.

I think the recent news about ccHT based co-processors are of great significance. AMD is setting industry standards again. Intel ruled the PC world with the x86, now AMD64 has become the standard, and AMD fully intends to add more instructions utilizing ccHT. Once a whole new ecosystem and industry food chain is established around those AMD standards, you can imagine what will happen to those being out-innovated.

AMD is like the United States, and Intel is like the British Empire -- this analogy to the American revolution is the best I can find, given AMD's history as a second source x86 vendor. I would equate SUN to the French and Dell to the Hessian mercenaries. The breakfree was last year. The old empire still has large presence in the 3rd world.

It's just the beginning.

Wednesday, March 22, 2006

DELL bought Alienware

As we reported here, Dell found the backdoor to the free land of AMD64 computing. Alienware is heavily into Athlon 64 gaming PCes and Opteron workstations. You shouldn't be surprised if Alienware rolls out Opteron servers later.

The DELL-Alienware-AMD deal also vindicated my claim that Intel's Conroe performance claim is a hoax.

I pointed out back in January that Dell will have to go AMD no later than 2Q06, or it will be too late.

Buying Alienware shows again that the Dell dudes are still stuck in a PC mentality, while the real money is in enterprise servers.

Now, if Dell wants to make acquisition in the server business, SuperMicro or Tyan should be primary targets, these private companies should be cheaper than public ones such as Rackable.

I predict that Intel will have no choice but to elminate special treatment for Dell to avoid a quick collapse, basically, Intel has to let DELL proper go AMD.

SUN's computing grid might be a hit

SUN launched its utitility computing service, it's like using the internet as transmission lines to deliver computing power to anywhere on the net. I can imagine one day people plug onto the Sun Grid to draw vast computing cycles onto their PDAs or cellphones. At $1/CPU-hr, everyone can enjoy the service like sipping Starbucks coffee. I was thinking about buying an UltraSparc machine to port some code, but now I can spend just $5 to compile and test on the grid, beautiful.

The economics is actually great, with 5000 CPUs, assume they are sold out for a year, the annual revenue will be 5000 * 24 * 365 * $1 = $43.8 million. SUN will probably charge additional fees on storage and network bandwidth.

SUN needs to create some flash tutorials on how to use the grid. The provisioning system must be automated, otherwise, you spend 30 minutes to earn $1.

Tom's comment on 65nm Intel Presler Pentium XE 965

Finally, Tom said something of wisdom:

But even when the Intel CPU is overclocked at 4.26 GHz, the dualcore Athlon 64 FX processors still offer better results for most applications. Sure, the new Extreme Edition CPU is competitive, but it does not outperform the Athlon 64 FX-60, even at 3.73 GHz.

The Pentium Extreme Edition 965 does its job by keeping Intel in the news and by bridging the time until the first Conroe-based Core Duo E6000 (or E8000 Extreme Edition) processors become available. But at the end of the day, it is just another NetBurst processor that is inferior compared to the dual-core Athlons. did its own testing, the 65nm Presler Pentium XE 965 performed slower than Athlon 64 X2 3800+ in a lot of applications.

We expect AMD64 to get a 10-40% performance boost by running 64 bit Windows Vista. However, Microsoft delayed Windows Vista for consumers because Microsoft will "get credit for listening to partners and ensuring that we're thinking about their businesses, not just ours or some portion of their businesses". You guessed right, Wintel is the greatest partnership, Windows Vista is waiting for the Israeli amateurs to get AMD64 working.

AMD published its 64 bit programming guide 5 years ago, Opteron was launched 3 years ago. Linux 64 bit has matured. If Windows Vista consumer edition gets delayed yet again to 2008, don't be surprised, our Israeli friends are still learning the AMD64 basics.

Tuesday, March 21, 2006

AMD's CEO reached his maximum rewards already

Accoding to this SEC filing, the number of restricted shares Dr. Ruiz can receive during the 2005-2007 period for meeting his target is 95,000 shares, the maximum he can receive is 210,000 shares. Then, according to this SEC filing on March 16, 2006, Dr. Ruiz received 210,000 restricted shares. Other AMD execs received similar rewards. Remember these incentive plans are like contracts, you don't get rewards unless you satisfy the conditions, or you get class action lawsuits from your own share holders.

You ask me what this means?

The answer: 25-30%.

Here is another hint.

Windows Vista delay indicates Conroe delay

Microsoft announced that it will delay consumer Windows Vista to January 2007. This is another indication that Intel's Israeli amateurs are still working on cloning AMD64 instruction set. Previously, I stated that Intel's failure to demo any 64 bit capability of the so called NGMA during the IDF was an indication that current NGMA implementations are IA32 only.

On the other hand, businesses will be able to enjoy business version Windows Vista on AMD64 according to original schedule of November, in coordination with AMD's big push into commercial client space. On the server end, we know Microsoft has standardized on Opteron.

AMD should move to NASDAQ

Too many crooks on NYSE, read this.

Intel Presler Pentium D 920, Pentium D 930 slower than Athlon 64 X2 3800+

Xbitlabs did some tests, and found the 65nm Presler Pentium D 930 slower than the slowest AMD dual core, the Athlon 64 X2 3800+. An overclocked Presler Pentium D at 4GHZ is about 20% slower than the AMD's flasgship FX60 in some tests. It seems AMD's price premium over Intel is well justified.

ccHT is becoming another AMD standard

AMD has opened cache coherent HyperTransport, now various co-processors are ready. Read the news here (Celoxica, HP, Cray), and here(Clearspeed vector FP). On non-coherent HyperTransport front, PathScale's infiniband can do 10 million messages per second.

Implications? Go figure, especially if you are inside Intel.

Monday, March 20, 2006

AMD capacity check again

INQ has reported that AMD is ramping FAB36 to 13000 wspm, so let's redo the capacity check.

This time, let's compute the die count for dual core Opterons at 90nm using the Wafer program. Approximately, we can get 300 dual core Opteron dies out of one 300mm wafer. This translates to 300 * 13000 * 3 = 11.7 million dual core Opteron dies per quarter.

Of course, if you go with smaller Turion X2s and Athlon 64 X2s, the number will be higher.

So, if AMD can sell all it makes, 40% market share is even achievable at 90nm.

1Q06 notebook sales to decline 10%

IA32 doesn't sell, I told you.

The Israeli amateurs claimed that people don't need 64 bits, a cheapo excuse for their inability to copy AMD64. Now the market has spoken, only uninformed people will choose IA32 today, but more and more people are getting to know about AMD64.

Prudential cut INTEL 2006 EPS to $0.86 and target price to $15. My projection for INTEL's 2006 EPS is -$0.15. Intel shall see operating losses from 3Q06 onward.

Saturday, March 18, 2006

Random news and random thoughts

It is MOE again, instead of the Mexican one buying 27000 Opteron workstations, this time it's Malaysia Ministry of Education buying 15000 AMD notebooks.

Hector Ruiz testified about the destruction of monopoly again.

Since AMD filed its lawsuit against INTEL last summer, we saw almost immediate positive effects for AMD. Opteron sales jumped 90% right away, a lot of timid companies went out to use AMD processors, Intel started to crumble with less and less sales. In this case, ten laywers are better than 10000 sales people.

AMD should dispatch its legal team around the world to drag the monopolist into the Courts. We have seen dawn raids in Japan, Eruope and Korea, more the better, PRC may be the next raider, then the Russians, then the US.

JFTC is going to release the incriminating evidence soon, let's wait and see.

Easiest way to get WiFi working under Linux

I bought this Athlon 64 3500+ socket 939+MB combo and installed Windows XP and CentOS 4.2 for AMD64 onto it. CentOS is just a build off the RedHat source code. I have an old D-Link DWL-G122 USB wireless dongle for connecting to the WiFi router downstairs. It works under Windows XP, as D-Link supplied the Windows 32 bit driver. Under Linux, it was a struggle, I tried to use the RT2500USB drivers, but found actually the USB dongle was a Prism54 chip. I also tried to use ndiswrapper, which allows you to use Windows drivers under Linux, unfortunately, with that approach, I have to use 64 bit Windows drivers under 64 bit Linux too, but there is none available. Going back to the dark, primitive, uncivilized age of IA32 computing is not an option, AMD64 is here today, and I must use it to its full potential.

I was at Frys today, and I found this AirLink101 Super G Model AP421W Wireless Access Point for $49.It was exactly what I need, as it can operate in AP Client and wireless bridge mode. To setup, I connect an ethernet cable from the Linux PC to the AP421W, point my browser to the AP's address at, set the device in AP Client mode, enter stuff like WEP key, etc, and viola, the 64 bit Linux PC is online, no drivers, no hassle, faster WiFi speed (USB dongle too small) --- I am writing this 64 bit article right now off the CentOS 4.2 box. Linux looks cooler and runs faster than Windows XP. I even run my 32 bit Windows apps under Wine.

I have tried the 64 bit version of OpenOffice, it opens much faster than under 32 bit Windows, FireFox is also a lot faster. The machine has only 512MB of RAM, but I feel no need to add more at this point. Folks who have AMD64 today will be pleasantly surprised that their PC gets free 10-40% performance boost by just upgrading to a new 64 bit OS, Intel EM64T users will actually find their PCes run slower in 64 bit. This I call the AMD64 effect.

Friday, March 17, 2006

Compare Broadcom and AMD again

Broadcom: 4Q05 profit $194.8 million, current market cap $23.59 billion

AMD: 4Q05 profit $206 million, current market cap $16.33 billion

AMD current stock price (March 17, 2006) : $33.95

AMD valued at Broadcom's valuations: 23.59/16.33 * 206/194.8 * $33.95 = $51.86

What does "destruction of the monopoly" mean?

Dr. Hector Ruiz, CEO of Advanced Micro Devices, recently used the phrase "destruction of the monoply" to describe the progress in the PC market. Previously, he always said stuff like AMD facing a "brutal monopolist".

What does it mean? In market share terms, it will be more like 2:1 ratio, instead of a 4:1 ratio. Go figure.

SuperMicro should have some balls

Stop being such timid little dudes, show your Opteron boards on your home page.

You see, INTEL CTO was doing free advertising for Tyan S4880 Opteron board at the IDF. If you dudes had some guts, your Opteron boards could have been there at the IDF too. Learn from the master at Tyan, be courageous, be free, this is America, this is not Soviet Union.

BTW: I think SuperMicro boards look ugly. I personally only use Tyan boards. Look at this (16 DIMM slots, no capacitors), and this (4 Opterons, 4 PCI-e 16 slots).

I predict a massive AMD64 effect in 2007

I installed both 32 bit Windows XP and Windows XP x64 on my AMD64 PCes, Windows x64 can run 32 bit applications without any problem. But I often have to work on the 32 bit OS because there are 64 bit drivers missing for some devices.

From my experience, Windows XP runs much much faster in 64 bit mode. This is because Athlon 64 is a native 64 bit chip that can process data in 64 bit chunks and have twice the number of registers in 64 bit mode. Microsoft's tests show AMD64 CPUs run 10-40% faster in 64 bit mode.

Intel's EM64T Pentium 4 CPUs runs 10-20% slower in 64 bit mode, and Intel's Centrino (Core Solo, Core Duo, Pentium M) is 32 bit only.

In 2007, when Windows Vista gets popular, those who bought AMD64 notebooks and desktops will be delightfully surprised: with the dormant 64 bit potential unlocked, their Athlon 64 and Turion 64 can run Windows Vista x64 at up to 40% faster speeds.

Those who bought INTEL will find their PCes having next to zero residual value: can't run 64 bit games, can't run Vista, only old 32 bit stuff.

The effect?

Guess what those Intel buyers will do with their next PC purchase.

Orders for Intel Mobile Chips Slahsed by 30-50%

According to DigiTimes report today:

"Leading notebook vendors, including Dell, Hewlett-Packard (HP), Acer and Asustek Computer, reportedly have slashed their orders for CPUs and chipsets already placed with Intel by 30-50% due to an unexpected rise in notebook inventory in the first quarter of this year, according to sources at Taiwan notebook makers."

The causes are believed to be Obsborne effect and Intel's lack of AMD64:

"Vendors suspect that consumers are waiting for the launch of dual-core Napa-based notebooks, resulting in slow sales for the single-core Sonoma-based notebooks, the sources indicated. Although some Napa-based notebooks are available now, consumers also have shunned them due to their higher prices, the sources noted.

In addition, the planned year-end launch of Microsoft’s Vista OS is also a concern as consumers are worried that the notebooks they buy now might not be able to run with the 64-bit enabled Vista OS, the sources stated."

A couple of days ago, IDG reported that notebook sales are on the rise, if that report was true, then AMD64 is making massive gains in mobile. Yes, Turion 64 X2 is ready and FAB36 is ramping.

I projected 3Q06 operating loss for Intel. I may have to move the time frame forward to 2Q06, depending on Intel's strategy.

Thursday, March 16, 2006

Finding the Nash Equilibrium between INTEL and AMD

From Wikipedia: "In game theory, the Nash equilibrium (named after John Nash, who proposed it) is a kind of optimal collective strategy in a game involving two or more players, where no player has anything to gain by changing only his or her own strategy. If each player has chosen a strategy and no player can benefit by changing his or her strategy while the other players keep theirs unchanged, then the current set of strategy choices and the corresponding payoffs constitute a Nash equilibrium."

Now, because of AMD's capacity contraint and Intel's technology lag, the Nash equilibrium is reached when INTEL reduces production and hikes prices to keep revenue constant. Basically Intel must allow AMD to take the additional market share.

If Intel tries to cut prices to a point lower than AMD's, it may gain back 1-2% of market share, but it will inflict massive revenue loss on itself.

AMD already runs into capacity contraint, lowering prices only lead to lower revenue, as it can't increase the units any more.

Bill Gates is very afraid

Bill Gates tried to downplay the OLPC $100 laptop running Linux again. What Gates said has zero merit, the OLPC has 512MB memory, much larger than those $399 PocketPCes with 64MB, the OLPC also has a much larger screen than PocketPCes.

Let me make a statement:

With OLPC, the world's next generation can grow up connected with each other and understanding each other, that will be a great breakthrough of the human civilization.

With OLPC running Linux, the world's next generation will grow up with open systems and open standards, they will use OpenOffice, Open Document Format and other open technologies. And they will contribute into that open environment as they grow up. The open software and the collaborative software development for the OLPC will lead to advances of technologies at no cost.

Mr. Gates, every empire falls.

Wednesday, March 15, 2006

Rattner: 2004 AMD64 is Intel's envy, ideal and future

The beauty of the 4 way Opteron board is a reflection of the underlying topology and symmetry of the Direct Connect Architecture -- Sharikou

The sharp eyes at INQ suspected the ideal platform envisioned by Intel CTO Justin Rattner was an AMD64 based system. Looking at Rattner's slide 34 at the Spring 06 IDF, I immediately recognized it as an old Tyan S4880, which is probably out of production, because it was an old design with only 10 DDR1 slots, the newer Tyan S4882-D has 16 DDR1 slots and supports dual core. The Tyan S4880 envied by Rattner was introduced in April 2004.

Intel Senior Fellow, CTO, Justin Rattner on Intel's future
Spring IDF, 2006
( In the background, Tyan S4880, Quad Opteron Board)

Tuesday, March 14, 2006

OMG, OMG, I have to admit SUN is smarter

Than everyone else....

2P, 32 core with 128 FB-DIMMs, or 8P, 128 core with 512 FB-DIMMs, man, now SUN needs to get their lazy sales people to work.

Rahul Sood asks Eric Ross to use his brain

Eric Ross @ Think Equity heard a rumor, horrified, he cut AMD from $50 to $30, panicked, those who trusted him unloaded their shares like newbies in the market.

Previously, Mark Edelstone @ Morgan Stanley reckoned:

a) AMD ASP $100, INTEL ASP $150
b) to gain 1% share, INTEL has to lose 35% revenue
c) that means INTEL will report loss
d) that means no bonuses, and stock options worth 0
e) INTEL thinks
f) So, there will be no price war
g) The logic is perfect

Now, Mr. Sood of VooDooPC lectures Eric Ross on some basics of logical thinking. Please read it, Eric, and thank Mr. Sood for his kindness.

DELL to follow Gateway

Rahul Sood, the guru CTO at VoodooPC predicted that DELL would buy Alienware, now the rumor is 45% news. Buying Alienware could be a move for DELL to escape to the freedom of AMD64 computing, just like what Gateway did with buying emachines.

Dell probably found that Intel doesn't sell at any price, people aren't excited about the $243 P4D 820 PowerEdge, instead they become suspicious- people don't want to spend less money to live with pain. They want to spend more money to enjoy the assured future of cool AMD64 computing.

Kevin Rollins is growing increasingly impatient with Intel amateurs, he is calling for making bold moves, yes, Kevin, make the jump, be courageous, be free.

What's next? RACK might be a good choice for anyone wishes to enter Opteron server space.

Then the Dell dudes can exclaim: free at last! free at last! yeah!

PS: Some sharp eye noticed a big change on Alienware's web site, it looks like a replication of . OMG, Aliens are invading the home offices, government and education systems....

Monday, March 13, 2006

INTEL wants a price war?

There are analyst reports saying that Intel is about to start a price war against AMD. If Intel does it, it will run into operating losses as soon as 2Q06, instead of the 3Q06 I previously projected.

Right now, Intel CPUs' average selling price (ASP) is $150, while AMD's ASP is around $100. Intel could sell their CPUs at a premium because of their brand. But things have changed, most people start to know that AMD commands a 20-40% performance advantage over Intel and AMD CPUs use about 70 watts less power. Therefore, Intel CPUs are a tough sell.

For Intel to effectively compete against AMD and stop market share loss, Intel needs to bring its ASP to about 15% below AMD, or $80. Even at an ASP of $80, AMD will still have some advantage when you consider electricity cost over the life of the CPU.

With an ASP of $80, Intel revenue will be at 80/150 = 53% of current levels. If I use the current revenue number of 8.7 to 9.1 billion, this means a quarterly revenue of $4.35 to $4.55 billion, or roughly $2.5 billion operating loss per quarter, or EPS of (-$0.40) .

Intel is basically in an inventory clearance mode for all its products.

AMD should not get involved in this price war, instead it should shift production to more high end server and performance chips.

Sunday, March 12, 2006

Henri Richard on Intel's future

From this interview, he seems to understand more than Mooly Eden on key issues. Intel folks are in denial, when the world+dog knows NGMA is Bob Colwell's P3. But Henri Richard said something that surprised me:

"I saw recently a comparison of what's been disclosed of the NGMA, compared to what's been disclosed of the AMD K8 architecture, in the sense of predictive branching, buffering, pipelining and so on. And when you look at their NGMA, it looks like the K8 more than anything else. As they move away from Prescott and into their new products, it's really a mirror, a much closer mirror image of the K8 than previously, with the exception of course of the integrated memory controller."

So, Intel amateurs also "borrowed" a lot of stuff from K8, according to AMD's grand masters.

Going forward, Mr. Richard promised higher IPC for integer and floating point performance, higher clocks, larger cache, more scalability and what customers want:

"We don't need something new, we just need to continue down the path of improving memory bandwidth through new technologies, improving speed, through faster clock rates, larger caches, and, in certain areas that are driven by customers who demand improvement, absolute performance. It's clear that we will improve both the integer and floating-point performance of our cores, independently of improvements in transistor technology, manufacturing technology and so forth."

Saturday, March 11, 2006

IDF summary report

NGMA is Pentium 3

I stated repeatedly that Intel NGMA is Bob Colwell's Pentium 3. A lot of Intel fans strongly opposed such a notion and insisted that NGMA would be all new from ground up.

I was 100% positive, because I know this: it is impossible for the Israeli amateurs to do some radical new stuff out of no where, they simply don't have the track record for ground breaking work.

Now it's all confirmed, the Core architecture (NGMA, Merom/Conroe/Woodcrest) is 95.5% Bob Colwell Pentium 3, plus 4% Bob Colwell list of todos when he left Intel, and probably 0.5% Israeli patchwork.

Look at those minute improvements: smarter cache, larger cache, better chipset, faster bus, grouping a couple of adjacent instructions when doable, Intel deeper sleep, tweaking a few SSE instructions, issuing 4 instructions per clock....I am trying to put all I remember here, but none of them is mildly interesting stuff. AMD would not even mention such minor improvements with their core revisions.

One thing that shocked me was that Anand reported that with the NGMA core, all SSE instructions are executed in one clock cycle instead of 14 clock cycles. I thought how could that be possible? Did Intel Israel invent a special 1 stage SSE unit? It turned out the cheerleader was too eager to cheer. It was all hoax.

It's bad that Intel Israel bragged so much for so little. And it's not good that they willfully failed to mention Bob Colwell the American, but instead took every credit of the Pentium 3 Israeli edition as all new.

Conroe Performance is a hoax

There is simply no way a 2.66GHZ Conroe can be 20% faster than a 2.8GHZ AMD64. Woodcrest is 10-20% slower than Opteron 280. INTEL had to use a 3GHZ Woodcrest to pit against a 2.4GHZ Opteron 280. Woodcrest is pretty much identical to Conroe, both are based on Merom.

NGMA/Core/Merom/Conroe/Woodcrest is 32 bit

One thing I observed was that there was no demonstration or even indication of AMD64 compability of the NGMA chips during IDF. No Windows Vista demos, no Linux 64 bit demos, no CPU-Z screen capture indicating the presence of x86_64. In HP's demo of the DL 380 with two Woodcrest processors, there was only 2GB of memory, odd for a server of this calibre.

If the Cores were 64 bit capable, it would be unthinkable for Intel to not demonstrate this capability, which is critical for Windows Vista and enterprise computing. The fact there was zero sign of AMD64 capability means all NGMA chips as of today are 32 bit only.

While AMD's grand masters grew up with 64 bit, the Israeli amateurs don't even understand there is a need for it.

The dire situation for Intel

I predict that Merom/Conroe/Woodcrest will be delayed. Intel will struggle to get AMD64 compatibility figured out. One way is resort to emulation, but there will be 10-20% performance degradation, as we can see from current EM64T benchmarks. AMD64 runs 10-40% faster in 64 bit mode.

There will be no IMC, no CSI, no true multi-core for INTEL in the forseeable future.

On desktop and mobile, AMD's preemptive strikes in May and June will be many steps and many months ahead of Intel's.

On servers, Woodcrest/Bensley will be stuck at 2P level for a long long time. In the end, Intel may pay big money to IBM to get the Hurricane chipset so they can get 4P working, but the cost will be very high. On cost alone, there is no way for Intel to compete against AMD64, which can do glueless ccNUMA up to 16P.

The biggest problem for Intel is, time has already run out. With FAB36 converted into 65nm, AMD will be able to supply 50% of the global PC market by 1Q07.

Intel is and will be five generations behind AMD64, IDF reconfirmed that.

Game over!

Google and SUN merger

I don't like the idea of Google buying SUN, but the merger of the two will definitely create long lasting effects.

With a single trick (ads on search), Google's growth will soon run into a wall, at that time, its stock will be valued on trailing P/E. Once Microsoft integrates MSN search into every Windows application, watch Google goes downhill fast.

SUN has everything Google doesn't. SUN has chip design, server, storage (not STK, but disk stuff), OS, Java, enterprise server stack, OpenOffie, etc, etc, plus the kitchen sink.

Once SUN and Google are merged, the possibilities are infinite. I don't want to talk about them here -- just use your wildest imagination of world domination.

DELL will rescue Intel

Wow! Dell PowerEdge SC430 server with a dual die Pentium D 820 CPU, 256 MB DDR2, 80GB HD, GbE, Cd-ROm, all for only $243. Yes, not $599, not $499, not $399, not $299, but $243, two-four-three.

Did Intel give Dell the CPUs for free plus some $20 bills? I hope Dell stay 100% Intel as long as it can. Dell can drive HP, Acer, Lenovo and whitebox vendors out of business for sure, out of Intel business.

I predict that Dell may acquire Intel in two years. It works like this, Intel gives Dell CPUs for free, so Dell makes a lot of profit, times a big P/E ratio, stock goes through roof, then it buys Intel in a white knight rescue effort.

AMD Live! will be for you!

I pointed out Intel's VIIV reflects a Intel-centric thinking, which is essentially this: give us your money or else.

Now, we understand more about AMD Live!

While Intel is planning a massive invasion of people's bedrooms and living rooms with monsters like this, AMD is working with set top box vendors so people can access the content in their Cool N' Quiet AMD64 PC on their TVs, over TCP/IP -- ethernet or wireless.

AMD is going to open source the technology, so it can work on different OSes, and even Intel PCes. As long as you have an AMD Live! certified set top box, you can freely enable the AMD Live! technology on your PC, be it AMD64 or IA32, hook up the network, and viola, you can view the goodies on your TV in the privacy of your bedroom!

That's what we need and that's customer-centric innovation!

Friday, March 10, 2006

AMD the most innovative semiconductor company

Among FORTUNE America's Most Admired Companies 2006, within the semiconductor industry, AMD was ranked #1 in innovation and #1 in quality of products/services. AMD ranked #2 in quality of management.

Intel ranked #4 in innovation, #4 in product/service quality and #4 in quality of management.

Intel Fellow Choked on question

The Register reported here:

"Why were we last?" Crawford said, repeating our question. "Why were we last? Why were we last?"

This mantra started to make everyone uncomfortable.

Looks like INTEL is stuck in a time loop of last century.

"At some point in the future, we will have an integrated memory controller," Crawford confessed. "That's something we are wrestling with."

"Intel is working really hard on that," Crawford said. "NEC has an FB-DIMM that is down to 5 watts. Maybe then we can get it down to 4."

"Our challenge there is to maybe catch up on some of the frequency angles in that product line," Crawford said. Although, he did hold out hope that Intel could eventually play in the high-end chip market. "I suppose that some day we will need a 128-bit architecture and then maybe the game could open up again.

What? 128-bit architecture? When? Is that part of the IDF benchmarketing per Moore's Law? Intel Senior Fellows need to go back to college and take a crash course on computer architecture 101. No wonder INTEL is still five generation behind and IDF changed none of that.

Thursday, March 09, 2006

Post IDF news summary

Conroe/FX60 generated much of the fun for IDF.

Time for the post IDF realtime news:

*) Intel throws good money after a bad chip, I said HP is killing Intel by having it hooked on the Itanic while flooding the cabinets with Opteron. On the other hand, without Itanium, INTEL would be reduced to an AMD64 clone maker.
*) VeriSign goes Opteron.
*) Continental AG Taps a Full Range of AMD Solutions for a Competitive Edge, this 81,000 employee company goes AMD on servers, workstations and desktops. They are complaing though: "The only downside is that the cluster is so fast, we don't even have time for a coffee break!"
*) Supermicro Debuts 4/8-Way 1U/4U Servers at CeBIT Based on Dual-Core AMD Opteron(TM) Processors, this former INTEL only house now found the courage (from GOOG) to issue its own Opteron news.
*) AMD in growing partnership with China's Lenovo, Lenovo is now 50% AMD.
*) NVIDIA Introduces Industry's First High-Definition Integrated Graphics Processor for Notebooks, fine print: for AMD64 only.
*) NVIDIA to Demonstrate Next-Generation NVIDIA(R) nForce(R) 500 Family of Core-Logic Solutions at CeBIT, fine print: AMD64 only.
*) AMD Turion 64 mobile technology available in Fujitsu Siemens, Turion 64 X2 is coming to town five months ahead of Merom.
*) AMD expects 60% LatAm revenue growth in 2006, in more developed regions AMD64's share of retail market has risen to 81.5%.
*) New thin clients from Fujitsu Siemens computers feature Geode ...
*) German labour office to buy 70,000 Fujitsu-Siemens PCs, AMD64 Inside (no Lanham Act please, INTEL lawyers, the Court is swamped)
*) Nasdaq, INTC, $19.74 (1.30%), Volume 69,752,665
*) One old news, Intel Lowers Revenue Forecast on Share Loss, up to 20% Q/Q revenue fall for 1Q06, estimated EPS falls sharply year over year and quarter over quarter.
*) One observation, DELL didn't cheer for NGMA? Rollins chided INTEL: "Is Intel going to meet the technology needs-–server performance and thermals–-where AMD does have a lead? That will answer the question. If they don't, that will also answer the question.”

INTEL has the burden of proof

After the public outcry about the Conroe/FX60 benchmark we discussed here, Intel offered some defensive explanations on why the BIOS on the AMD box was old, etc. But, that's not even the key issue here.

The key issue is, INTEL needs to prove its performance claims. Intel must prove that it didn't dope the Conroe machine.

Whoever wants to claim something to be true and wants to benefit from its being true has the burden of proof.

INTEL claimed its Conroe chip is 20% faster than AMD64, INTEL made this claim with the aim to profit from the claim.

So, prove it.

Tanya Harding can't claim she is the champ, unless she goes through a contest in a public arena, then subjects herself to doping tests. The same applies to any other athlete.

INTEL doped Skype with the aim to force consumers to not buy AMD64 but Intel's double die. INTEL claimed that only Intel CPUs are good enough for Skype's 10-way call, but it was found to be a mere CPUID check to exclude AMD CPUs.

INTEL can dope Skype, it sure can dope its own horse.

INTEL can't have the machine in an environment of its own control and claim anything. Intel can't add credibility by merely having Anand relay the data of this machine arranged by INTEL itself (without any anti-doping tests etc), the Anand cheerleading afterwards doesn't add credibility either.

So my challenge to INTEL is simple: don't be shy, don't hide, lend two Conroe boxes to two 3rd parties and let them do a full test.

One thing we also want to know is whether the Intel has done copying the AMD64 instruction set---no news about that in the IDF.

By the way, I have clear and convincing proof that Woodcrest will be 10-20% slower than Opteron 280 here, based on the following credible evidence:

1) The independent benchmarks done on Paxville DP 2.8GHZ and Opteron 280 are credible.

2) INTEL's claim of Woodcrest's 80% performance increase over Paxville DP 2.8GHZ is presumed to be credible.

Woodcrest is the server chip based on the Conroe, just like FX60 is pretty much the same as Opteron 280 in most aspects. Since Woodcrest is slower, Conroe must also be slower. Yet, the Conroe setup by Intel was claimed to be faster. I think this is clear and convincing proof that there was doping acitivity there.

Intel throws good money after the Itanium chip, because so called INTEL NGMA won't do wonders. NMGA is just an enhancement to the Pentium 3 and is five generations behind today's AMD64 .

Bill Gates running out of ideas?

From 4 pound Tablet PC to 2 pound Origami? There are notebooks weighing only two pounds.

Wednesday, March 08, 2006

Woodcrest will be 10-20% slower than Opteron 280

In October 2005, INTEL announced the Paxville Xeon processors, INTEL claimed a 50% performance increase. That has become a subject of ridicule in the industry, as it had only half the speed of Opterons and INTEL refused to benchmark them.

It's Spring IDF time again, and INTEL sales are plummeting, so again, INTEL promised by 3Q of 2006, we will get a new server chip called Woodcrest on the Bensley platform. INTEL reckoned that Woodcrest will be 80% faster than the Paxville DP 2.8GHZ, it will also get rid of the useless hyperthreading stuff invented by INTEL America.

Wow! 80% on top of 50%, impressive!

So let's check some existing benchmarks for the referenced Paxville DP 2.8GHZ and see how this Woodcrest wonder chip will do for us in 3Q06. The Paxville DP 2.8GHZ and Opteron 280 results are here.

First, let's look at web server performance:

For 10,000 users, the Paxville handles 701.7 HTTP request/second, not bad, times a factor of 1.8, we get 1263.06 res/sec for Woodcrest, impressive for INTEL. Opteron 280 does 1505.8 reqs/sec, Opteron 280 is only 19.2% faster than Woodcrest.

Next, let's look at media encoding, this benchmark might be relevant for workstation users:

Windows Media Encoder 9.0: Paxville 2.8GHZ finished in 100 seconds, quite fast, devide that by 1.8 for Woodcrest estimate, we get 55.55 seconds; Opteron 280 does it in 51 seconds, Opteron 280 is only 8.9% faster than Woodcrest.

Next, let's look at Scientific stuff, scientists need computing power too:

Sciencemark 2.0 - Primordia: Paxville 2.8GHZ finished in 584.3 seconds, divide that by 1.8, we get 324.6 seconds; Opteron 280 does it in 290 seconds, Opteron 280 is merely 12% faster than Woodcrest.

During the IDF, HP had to use a HP DL380 armed with a 3GHZ Woodcrest to pit against a 2.4GHZ Opteron.

AMD just shipped Opteron 285 a couple of days ago, you can get it FedExed to you in 1-2 days, it plugs in nicely onto current Socket 940 boards.

In conclusion, AMD's Opteron 280 only has a 10-20% performance advantage over Woodcrest, we don't have scores for Opteron 285 yet, it's too new, maybe AMD will do better by 3Q06 when Woodcrest comes out to take a frag.

Tuesday, March 07, 2006

Was Anand duped by INTEL?

I warned that INTEL will try all sorts of dirty tricks, this is INTEL's character. From the book "Inside Intel" by Tim Jackson, INTEL even tampered documents to defraud the Court and the jury. From AMD's lawsuit, INTEL was accused of rigging benchmarks (remember BAPCo?) and tampering compilers. More recently, we have INTEL caught playing dirty with Skype. The list could go on.

Today, at IDF, the smart INTEL folks left two machines for people to play with, one equipped with an AMD CPU and one with a future INTEL Conroe CPU, both with a bunch of gaming and other benchmarks installed. Curious as our Anand, he hang around them with a lot of interest, an INTEL guy recognized him as the Anand, and granted him 60 minutes to play with the machines. Anand pushed through the crowd, hurriedly yet professionally ran a set of benchmarks, then quickly wrote: "Intel Regains the Performance Crown". He did all that in little less than 59 minutes and 59 seconds, amazing.

Readers of AnandTech were awed by INTEL's performance: 10-30% faster than 2.8GHZ AMD on gaming! Many vowed to hold off buying AMD. I bet INTEL folks are giggling.

But, wait a minute, the Anand! First, INTEL hasn't gained the crown yet, Conroe is still months away, AMD hasn't shown its cards yet. So Anand, be precise, a title of "Intel May Regain the Performanc Crown In Six Months" might be more proper, that is how professional journalists write, you should learn.

Then, pay attention, Anand, INTEL painstakingly arranged the AMD box for a purpose. There might be traps there. Looking at the BIOS screen, it says "Main Processor: AMD Processor Model Unknown", hmm, something fishy already from the beginning, it indicates that the BIOS didn't recognize the CPU at all.

So, let's look closer, "Phoenix - AwardBIOS v6.00PC, Copyright (C) 1984-2003, Phoenix Technologies, LTD". Let's dig deeper.

After googling for a while, we found that AwardBIOS v6.00PC was used on old Athlon XPs.

For an Athlon 64, DFI has updated BIOS here dated 2005 which fixes a lot of problems and added support for FX60. I doubt DFI is shipping 2006 motherboards with 2003 BIOS. Anand, please call DFI and see if they shipped any 2006 motherboards with a 2003 BIOS, that's what professional journalists do: get the truth out and earn credibility.

If not, then INTEL was rigging the BIOS so the dual core AMD64 CPU would act like a unknown CPU without the super duper AMD64 features. Anand, you claimed that the CPU was an FX60 running at 2.8GHZ, but how did you know? Did you remove the heatsink and have a peek? What if INTEL put an Athlon 64 x2 3800+ there?

I never thought Anand could be so gullible, Rahul Sood, CTO of Voodoo PC, was surprised by Anand's apparent naiveness too.

PS: I sent this message to INTEL execs, AMD execs, Anand and media folks, plus a request for INTEL to let Anand take the Conroe machine home for a test. We will see if INTEL can meet that small challenge.

AMD should say something to clarify the situation too, such as the effect of using an old BIOS on a new chip.

PPS: After public outcry, INTEL changed the BIOS on the AMD box, and Anand did the same tests again. This indicates INTEL has complete control of the system and knows what they were doing, Anand was simply reduced to a pawn in an INTEL controlled environment to push the right buttons INTEL pre-arranged. Given the fishy stuff going on with F.E.A.R, and INTEL's Skype CPUID cheat, such INTEL controlled tests cannot be trusted.

Again, I challenge INTEL to lend the Conroe box to Anand for a full test. According to my estimate based on INTEL's PR, Woodcrest will be 10-20% slower than Opteron 280, there is simply no way that Conroe can be faster than the FX60.

INTEL's self-defeating prophecy

INTEL announced that their Woodcrest server CPU will be 80% faster than the Paxville DP 2.8GHZ. However, currently, the Paxville 2P has a SpecFp_rate score of 40.3, and Opteron 280 2P has a score of 74.4, the Opteron 280 is already 85% faster than Paxville DP. Furthermore, AMD just released Opteron 285 yesterday, with another performance boost of up to 14%. Actually, Opteron 890 is already out there, and AMD readied hounds to blood Intel's next gen hares.

INTEL's bogus cliams of performance using vaporware no one has seen will definitely cause a massive Osborne effect, people will wait for the new wonder chips for sure, so expect the sales of Paxville to halt. INTEL execs see that.

What INTEL is trying to do here is not to help itself, but to hurt AMD, it is a feeble attempt, just like its Skype deal. INTEL hope that people will stop buying AMD but wait for INTEL.

It won't happen. As we have seen from the case of Google, businesses depend on computing won't wait for INTEL even if INTEL CEO sits on their board. Diehard INTEL fans will wait for INTEL, so they will stop buying current INTEL junkies.

So the situation will be this: INTEL revenue will drop sharply, some going for AMD, which promises an upgrade path, some wait for INTEL, then when INTEL comes out with something 6 months later, AMD already launched their newer generation. AMD will always be two steps ahead. For instance, the dual core Tuion 64 X2 will be released in May 2006, several months ahead of Merom, which is struggling with heat problems. Before INTEL releases Woodcrest to match Opteron 285, AMD will have Rev F opterons for sale. Not to mention that Woodcrest is only good for low end 2-way computing.

Monday, March 06, 2006

INTEL to suffer severely from Osborne effect

Tech companies try to avoid pre-announcing products as much as possible, especially if current products will be end of lifed when the new ones come out. The Osborne effect gave a good lesson, purchases of current products may stop cold as customers wait for the next generation.

INTEL is hyping its Pentium-III based NGMA CPUs, including Merom mobile chip, Conroe desktop chip and Woodcrest server chip like crazy. What we have learnt so far:

*) The NGMA chips are designed for high performance and low power consumption instead of clockspeed

*) The NGMA chips perform 50% better than current IA32 chips (Pentium 4, Xeon, Core Duo)

*) The NGMA chips consumes maximum of 60 watts instead of the current 130 watts.

*) The NGMA chips are AMD64 and Windows Vista compatible, unlike the current IA32 chips

*) The NGMA chips require new chipsets and even new VRMs, they will not be compatible with current motherboards

*) The NGMA chips run at bus frequency of up to 1333MHZ, current IA32 CPUs can't run on NGMA motherboards

*) Current IA32 CPUs already hit the frequency ceiling, there won't be any upgrade path for current chips.

*) NGMA chips are expected in 3Q06, July 2006 the earliest, only 4 months away.

*) INTEL plans to convert 20% of its CPU production to NGMA by the end of 2006.

Needlessly to say, an informed INTEL customer will incline to hold off purchases as long as possible for the super duper NGMA, consequently, in the next four months before NGMA shows up in store, INTEL sales will definitely suffer a slow down.

Then at the time when NGMA arrives in stores, we run into an even bigger problem: because of limited availability, INTEL can't meet the demand, but nobody wants IA32 chips any more. It's just like Microsoft shipped a handful of Xbox360s ahead of PS3, then the thing quickly ran out of stock, and few want to buy the old Xbox with a Celeron inside. Microsoft doesn't care, because Xbox wasn't making money any way. But for INTEL, the situation is quite different.

Another major problem is that AMD has already prepared the next round of frags for INTEL. I bet both INTEL and AMD showed its roadmaps and next generation chips, yet Google decided to go AMD, that confirms my point that INTEL's NGMA is over promising. So when people can't get hold of NGMA due to availability, they will have no choice but choose AMD's Rev F.

AMD platform choices provide stability and continuity. Socket 754, 939, 940, AM2 platforms can all be upgraded, AMD keeps providing core revisions and clockspeed upgrades. For now, there is no motivation to wait for Socket AM2, as the Tomshardware tests showed that Socket AM2 version of the Athlon 64 X2 4800+ had identical performance as current Socket 939 ones -- AMD is careful to avoid the Osborne effect.

Once AMD and Intel move to quad-core in 1Q07, any enhancement made by Intel at the dual core level will be wiped out in a flash. Four Conroe cores sharing a 1333MHZ FSB will be worse than two Presler cores sharing a 800MHZ FSB. The 1333MHZ bus (10.6 GB/s) is not even fast enough to handle dual channel 800MHZ DDR2 (12.8GB/s), while the Rev F AM2 has a dedicated memory interface for 800MHZ DDR2 plus up to 3 hypertransport links with 8GB/s each. Furthermore, I expect AMD to increase AM2's clockspeed to up to 4GHZ using stress technology, leading to another round of GHZ war in 2007.

It will be deja vu all over again.

AMD capacity sold out for 2006

At a conference hosted by Morgan Stanley, AMD CEO Dr. Hector Ruiz was tight lipped about how much share AMD has gained from INTEL. I think he had a good reason, 1Q06 still has 3 weeks left. When asked about whether AMD sees a PC market slow down, Dr. Ruiz stated that AMD does not have any concerns. Previously, in an interview with BusinessWeek, Dr. Ruiz stated that AMD is doing well while INTEL is not and AMD intends to keep it that way.

Dr. Ruiz did provide some overall view of AMD's operations and the micro processor industry landscape.

For 90nm process, FAB36 is executing extremely well at mature yields. Dr. Ruiz had a hard time finding the super duper words to praise the superb work done by the German team at Dresden. FAB36 is ramping as fast as possible and finished products are shipping. FAB36's transition to 65nm is on schedule (2H06) and will be pretty converted to 65nm by mid-2007. Finished 65nm products are coming out of FAB36 right now, and Dr. Ruiz used the extreme word to describe them.

As for Chartered FAB7, it's now at product qualification stage, finished products will hit the shelves in 2H06.

On margins, Dr. Ruiz stated that AMD will achieve 60+%, while still allowing PC makers to profit.

On multi-core, Dr. Ruiz pointed out that AMD's architecture shines more as one goes to more cores. AMD will go beyond quad-core. There is no technological barrier to do more cores, it's up to customer demand.

On capacity, Dr. Ruiz reckoned that AMD may be capacity challenged in consumer space, AMD will put higher priority on server and commercial client. Dr. Ruiz pointed that it is much more difficult to sign up server customers than desktop. Getting a company to use a server chip involves talks at the highest levels, while commercial desktop is more like dollar per pound business decisions. Right now, AMD has less than 10% market share in the commercial client space, AMD's focus this year will be changing that.

On mobility, Dr. Ruiz admitted that INTEL had a head start on mobile, but AMD is going to bring some major innovation in the mobile space within a year.

Dr. Ruiz observed that there are only three companies in the world that can provide total micro processor solutions, IBM, INTEL and AMD. AMD is going to break INTEL's monopoly, but it won't become another monopoly itself. Unlike INTEL, who kept all profits to itself, AMD will create an ecosystem that will allow everyone else to profit.

Based on Dr. Ruiz presentation, I conclude that AMD's capacity is sold out for 2006 and INTEL is left to grab whatever AMD can't fill.

Moreover, the hyping of the Merom/Conroe vaporware at the coming IDF will create an Osborne effect that will sharply reduce customer demand for IA32 technology, including Netburst based P4 and Xeon, and Pentium-M based Core Duo.

Sunday, March 05, 2006

I predicted 15-20% INTEL 1Q06 revenue decline back in Feburary

Back on Feburary 20, 2006, I projected that INTEL will suffer a 15-20% revenue decline in 1Q06, now INTEL's 8-K filing admits a 10% - 20% decline. For me, this was no surprise, and I predicted even larger INTEL revenue drop for 2Q06.

While smarter folks such as Prudential, Morgan Stanley, Joe Osha have figured it out, the real retards are still scratching their heads.

Dan Niles, CEO of Neuberger Berman Technology Management, said that AMD couldn't possibly have gained enough share to force Intel to drop revenue by $500 million. Why? He explained: "The math doesn't work, since Intel's share is nearly 80%".

This kind of statement makes me sick, such retarded words off the mouth of some CEO? We have to improve our education system. My suggestion to folks such as Dan Niles is this: go back to elementary school and re-learn the basic logic.

Joe Osha, though often gave in to the dark force, is probably one of the smartest, back in October 2005, he said that the situation between AMD and INTEL was like a head-on collision of two high-speed trains. I say Joe Osha is smart, because I found that he was the only Wall Street analyst who saw the problem, or disaster for INTEL.

One word: FAB36.

As of 4Q05, AMD captured 21.4% market share with the 200mm FAB30 running at 120 MPH. Now, it's a matter of simple math to compute the output of FAB36, which started production in Oct 2005. You can go read this whole journal, I analysed the production potential of FAB36 numerous times. Some people don't believe my projection of AMD capturing 40% of x86 market exiting 2006, so I did the die count here.

If I assume that FAB36 is in a slow cruise mode and only does 4000 wspm at 90nm, that's enough for an extra 5% of the PC market. As of 4Q05, INTEL market share dropped to 77%, a 5% drop, it's 72%.

But, as I pointed out in January 2006, FAB36 is in high gear (notice the GOOG symbol there?).

The math computes, because INTEL was not only losing share, it foolishly lowered prices trying to stop the inevitable share loss. I proposed the ONLY way for INTEL to survive: kick out DELL and raise prices, but INTEL folks won't listen.

The collapse of INTEL will be sooner than I previously expected. Increasing decline of market share coupled with lower ASPs will result in even more drastic drop in revenue. Unless INTEL follows my advice, I project it to suffer operating loss in 3Q06.

Friday, March 03, 2006

AMD readies more frags for INTEL

As expected, our friends at INQ reported that AMD readies hounds to blood Intel's next gen hares. I wouldn't say hares, hamsters might be a better word. The amateurs will be ridiculed. AMD64 architecture has built-in so much potential, the technology gap is widening.

IBM should be very afraid of Opteron now.

INTEL has warned that its 1Q06 revenue will be between 8.7 and 9.1 billion due to lack of demand for IA32 technology. This is a 13% drop from the $10.2 billion of 4Q05, the situation is even worse than the dotcom bust, but in a growing PC market. AMD64's share of the US desktop retail market has risen to 81.5%. Analysts are saying that INTEL's trouble has just begun and have lost confidence in any forward earnings projections. I predict that INTEL may start reporting operating losses sometime in 2006 instead of previously projected 2007.

AMD is expected to pick up INTEL's lost revenue. Dr. Ruiz has stated AMD intends to keep INTEL down. It seems that my earlier projection of $1.7 billion revenue for AMD's 1Q06 might be a little bit conservative. But I will stay my projection that AMD will exit 2006 with 40% market share.

Thursday, March 02, 2006

Google's Economics

Forbes re-reported the story that Google is going AMD, even though INTEL CEO Paul Otellini sits on Google's board. Surprised? Not really, it's a simple matter of economics.

Google has 200,000 servers right now, and it's adding more and more every day. How much can Google save by going AMD?

Marc Andreessen has done a study for his own dotcom startup's data center choices, he found by using SUN's Opteron servers instead of Intel Xeon, he could save $97.5 - $39.0 = $58.5 per server per month on energy and rack space. Since Marc is using his hard earned money, I bet he is pretty adamant on minimizing the total cost.

Now, let's apply the Marc Andreessen study to Google. With 200,000 servers, the total savings per year is:

$58.5 * 12 * 200,000 = $140,400,000 = $140.4 million

Google has 295.5 million outstanding shares, the savings above translate nicely into $0.475 EPS.

With a P/E of 75, that translates into $35 higher stock price and $10.5 billion of added market cap. Even though INTEL is quite rich, I doubt it can pay Google that much money to keep it.

Now, consider that Google must grow and add more servers....

Wednesday, March 01, 2006

AMD to keep INTEL down

In an interview with BusinessWeek, AMD's CEO Hector Ruiz reckoned that in the past, when AMD did well, INTEL also did well, now, the situation is different, AMD is doing well, and INTEL is not, and he intends to keep it that way. On the Dell issue, Dr. Ruiz stated that AMD's business plans are based on the assumption that DELL doesn't sign up with AMD. Dr. Ruiz pointed out that without AMD64 technology, DELL products are mediocre at best. More details on the BusinessWeek interview are here.

Morgan Stanley reported that most of Google's new server purchases will be AMD Opteron, despite the fact that INTEL CEO Paul Otellini sits on Google's board.

Dr. Ruiz was appointed to the President's Council of Advisors on Science and Technology of the Bush Administration today.

Merom/Conroe performance claim is a hoax

INTEL claimed 20% performance advantage of Merom/Conroe over AMD's Athlon 64 X2 is a hoax. Upon information, clock for clock, Merom/Conroe will not be any faster than AMD64 for real applications.

This paper is a good analysis of the so called INTEL NGMA -- Bob Colwell's 1995 P6 from an architectural point of view.

Israeli amateurs will be discredited when real numbers show up.