Tuesday, January 31, 2006

Real men have Opterons

SUN dominates the workstation market with Opteron. The era of cheapo and underperforming INTEL PCes as "workstations" is about to end. SUN's Opteron boxes redefines class, raw power, reliability and prestige. If your time is precious, you should get an Opteron workstation that can keep up with the speed of your brain.

According to this report, SUN owns 59% of the workstation market and is on exponential growth curve. SUN shipped 15,000 Opteron workstations in Q3 2005. In comparision, DELL has only 8% of the workstation market with INTEL Xeon inside.

In January 2006, in a single deal with the Mexico Ministry Of Education, SUN sold $400 million of Opteron powered Sun Ultra 20 workstations to Mexico's public schools. The story was that initially there were only bids with INTEL CPUs, after learning about the bid, SUN quickly moved in, and beat INTEL offerings hands down with the Opteron. SUN' President & COO Jonathan Schwartz made a quick trip to Mexico, apparently for signing this deal.

My Athlon 64 PC is no match to the SUN Opteron workstations the Mexican teachers are using. I think I will get a SUN Opteron workstation myself, just to show my time is precious and I command the computing power that matches my brain power, as I doze at my desk.

Cute little Apple boxes with 32 bit Core Duo are for ladies.

Real men have Opterons.

DELL to use Turion 64?

Some analyst stated that DELL could start selling Turion 64 notebooks as early as March. Visiting DELL.com, I found INTEL Core Duo is very hard to find, and is only being used in large 17 inch desktop replacement notebooks. Since INTEL won't have 64 bit mobile CPUs until Q4 2006, and AMD will enjoy 100% of the 64 bit mobile market, DELL using Turion 64 makes sense.

In a previous article titled "DELL will spit at INTEL and kneel to AMD soon", I pointed out that if DELL wants to go AMD, it better does it quick, anything done after 2Q06 would be too late.

The dude is certainly not dumb.

Energy Star for CPUs and servers

SUN and AMD are working together to initiate an Energy Star program for CPUs and servers. As I proposed here, SUN and AMD should ask DOE to impose power limits on microprocessors, just like it did on appliances. With an Energy Star program on CPUs and servers, people will really start thinking about total cost of ownership, including energy bills and effect on the environment.

OLPC got some folks scared

INTEL, and Microsoft. They should.

Monday, January 30, 2006

INTEL suffers from in-breeding

I took a look at INTEL top brass' bios, there are no tech guys at the Executive VP level and up. At the Senior VP level, there are two guys with chip design experience, but they only worked on old stuff such as 386, 486 and Pentium II. At the VP level, I only saw one chip design guy from Israel who worked on Pentium-M (a modification of Pentium-III).

Compared to AMD, who has a DEC Alpha architect as President & COO and a fresh IBM guy as CTO, INTEL leadership is technically very weak, no fresh blood, all in-breeding.

AMD warp speed paralyzes INTEL

Google and AMD cooking somthing big? I heard of those 5000 Opteron monsters and Google's plan to conquer the world with them. Now we know BigDaddy is ready. Anyone wants to bet what it is?

Then we heard Socket AM2 is quite ready, be out in Feburary, when I check the calendar, it's January 30 today. Socket 1207 F will be out in April 2006, sounds like a very good timing, I recall AMD introduced single core Opteron In April 2003, and dual core Opteron in April 2005, should we expect quad-core in April 2006? Hector's law says AMD is accelerating and grows at twice the speed of light.

Now, let me repeat some old formula I derived:
INTEL has 5 production 300 FABs and 7 production 200mm FABs, INTEL capacity is more than 11x AMD capacity, INTEL is running at full capacity, and can't meet chipset demand,

11 * INTEL_yield / AMD_yield = 77%/ 21.4%

AMD_yield = 3 x INTEL_yield

Heard about APM 3.0?

Sunday, January 29, 2006

IBM just realized that Power has a power problem

In this interview, Frank Soltis, a key person in designing the IBM Power processor, identified heat as a major problem for the Power. Followed the footsteps of AMD Opteron, IBM improved Power4 and added embedded memory controller to Power5, with impressive results. However, last time I checked, the Power5 generates over 150 watts of heat, in the same league as INTEL's Itanium and Xeons, much hotter than 72 watts for the 8-core UltraSparc T1 and 95 watts for the dual core Opteron 880.

With perfect hindsight, Mr. Soltis pointed out that both DEC Alpha and INTEL Itanium died or will die because of lack of backward compatibility with existing software.

Looking forward 10 years, I think there will be three major architectures left: AMD64, Power and Sparc.

AMD64 with its coming 32P glueless ccNUMA capability and multi-core DCA architecture will rule the high performance server, desktop, mobile and handheld markets.

Power will find its survival in game consoles. It will be out of the mainstream enterprise server market because Opteron will be more powerful and SUN will not port Solaris 10 over to Power. AIX is dead. Linux will not move fast enough to halt Solaris 10.

SUN will continue to shine with a Solaris 10 Sparc.

But SUN needs to focus on Solaris 10, Sparc and Opteron. There is one thing SUN can add to kill all Linux: a Solaris 10 GUI control panel that is as easy as Windows server 2003.

SUN should stop wasting money on application servers and other non profitable software. In the software industry, it's a winner take all game, the best makes 90% of the money. So if you can't be the best, you quit. BEA is very good, even JBOSS can do a better job than SUN, so why bother with a SUN J2EE server? SUN only needs some top guys writing J2EE standards and let others do it. If SUN can bundle Oracle, it can also bundle WebLogic, at a discount. Other stuff such as free SUN studio tools should also be cut, why not let Borland do it? If there is money to be made there, it's much cheaper just to buy a stake in Borland. It might be painful to reduce some workforce, but it's for the greater good of the network computing era SUN hopes to define.

I got a feeling SUN is still quite wasteful in using its money. I heard a lot of PMs just sit in their remote homes and forward emails. If you go to Microsoft, there is no such waste of resources.

An AMD-DELL deal will be good for AMD stock price

But such a deal will be bad for AMD in the long term. Let me re-iterate, AMD needs DELL to stay 100% INTEL for several reasons:

1) DELL keeps sucking INTEL blood by paying lower and lower prices for INTEL components, DELL gets more INTEL pie and becomes more dominant in INTEL space.

2) With DELL's lower prices, HPQ, Lenovo, Acer, Gateway and whitebox vendors find it impossible to make a dime competing against DELL in INTEL space and will move to more AMD. The math: DELL has only 17% of the PC market, the rest is 83%.

3) As DELL stuck with hot and slow Xeon, SUN will be more motivated to commit more to Opteron.

Eventually, DELL on INTEL will be like SGI on Itanium, DELL is just an expendable screw driver company which our civilization does not need. Any Chinese farmer can assemble a PC that frags Dell XPS by using AMD64, NF4, SLI and other quality components.

Once AMD starts selling cheap to DELL, the above advantages AMD gets from INTEL-DELL axis will be gone. Furthermore, with Dell's low quality and bad service, AMD's brand will be in serious danger of being damaged by DELL.

Joe Osha sings for AMD

INTEL CEO pounds his chest and vows to gain shares in Q1 2006 -- without telling us how, and INTEL forms an axis to pump $10 billion dollars into Itanium--with DELL, IBM, Microsoft and SUN notably missing from its list of friends.

Merrill Lynch analyst Joe Osha hymns for AMD in a report: "Nothing that we heard from Intel this week changes our belief that AMD is going to take more share in server processors this year".

Of course, this was after Joe downgraded AMD to sell on January 17, 2006 and then bought $500 AMD shares at rock bottom prices.

I admit that I underestimated Joe. Maybe the only real fool is Goldman, but who knows?

Saturday, January 28, 2006

Core Duo notebooks are literally hot

Some suspected that the elephant in the room for the Core Duo (Yonah) is battery life. Tomshardware measured the battery life of a Turion 64 notebook, a Pentium-M 32 bit notebook, and a 65nm 32 bit Core Duo notebook, and found that Core Duo uses almost twice the power as the Turion 64. When viewing DVDs, the Core Duo T2500 (2GHZ) consumes 34.1 watts, the Turion 64 ML-40 (2.2GHZ) consumes 19.1 watts. When reading, the Core Duo T2500 consumes 31.3 watts, the Turion 64 ML-40 consumes 15.9 watts. While using office software, the Core Duo T2500 consumes 33.1 watts, and the ML-40 notebook consumes 16.9 watts. The Core Duo notebook consumes almost twice the power than the HP ML-40 notebook.

Please note, in all these tests (DVD playing, reading and office), only one CPU core is needed, therefore one of the Core Duo's two cores was shut down during the tests, yet the system power consumption was as high as 34 watts. In a previous article, based on an AnandTech measurement, I estimated that the 65nm 32 bit Core Duo (Yonah) 2GHZ is at least 53 watts when its two cores are running at full speed. In comparison, AMD's 2GHZ dual core 64 bit server chip the Opteron 870HE is 55 watt max.

INTEL claimed that if 64 bit capability was added to Core Duo, battery life would have been even shorter. My guess is that the amateurish INTEL design team in the middle-east haven't figured out how to copy the AMD64 instruction sets invented by grand masters at AMD.

No wonder the 32 bit Core Duo notebook tested uses a massive 71wh battery, while the HP nx6125 equipped with a Turion ML-40 only uses a 48wh battery.

Again, why HP is using the Turion ML instead of the low power Turion MT.

AMD Live! soon

According to this report AMD Live! PCes with dual core 64 bit processor will run Windows Vista and enable simultaneous pervasive 64 bit entertainment through out the home for kids and parents. INTEL's 32 bit V//V with 32 bit Core Duo will be fragged to death.

INTEL's $10 billion Itanium is still at 130nm

Intel hopes to move its $10 billion IA64 chip to 90nm soon. Jack Uldrich the fool should bash INTEL a bit now, as its process for the enterprise server chip is still so ancient, 130nm on bulk silicon. Mr. Uldrich, don't you agree that is too primitive?

INTEL has 5 production 300mm FABs and 7 production 200mm FABs, and INTEL is having problem producing chipsets for months and months. INTEL's $10 billion Itanium is still at 130nm. Such absurd situation shows the primitiveness of INTEL's copy exact methodology. And the coming 90nm Itanium will have to run at 1.6GHZ instead of the planned 2GHZ to stay below 100 watts and avoid silicon melt down.

AMD has one 200mm FAB called FAB30 (FAB36 CPUs are about to flood the market), with this one 200mm FAB, AMD is making Opteron Socket 940, Opteron Socket 939,Athlon 64 FX, Athlon 64 X2, Athlon 64 socket 939, Turion 64, Sempron socket 754, Mobile Sempron, Socket AM2, Socket F 1207, AMD 8131, 8111 HyperTransport chipset, Geode NX, GeodeGX2, Alchemy, and each of them has dozens of models.... And AMD has 21.4% market share in x86. AMD is light years ahead of INTEL in processing technology.

In fact, AMD's APM 3.0 is years ahead of the whole FAB industry, it can perform on the fly adjustments and corrections at the die level.

Friday, January 27, 2006

Chartered FAB7 and AMD FAB36 in high gear

FABTECH reports here that AMD will kick in Chartered FAB7 production earlier than forecasted. Chartered has increased its capacity to 15000 wafer starts per month. It may start cranking out AMD64 in Q2.

FABTECH also reported that AMD's recent offering of $500 million stock is for the ramping up of its own FAB36.

To me, this is proof that AMD got some serious demand (<-GOOG) and I am quite confident on this analysis.

I predicted 40% maket share for AMD exiting 2006 and gave more details here.


This fool is an idiot

Motley Fool Jack Uldrich is hyping up INTEL's future 45nm process here, on the news that INTEL made SRAM on it. You may ask why he is not hyping INTEL's flasgship 65nm that much. INTEL barely started doing 65nm.

The sad truth is, even though INTEL has a lead in 65nm process, AMD's 90nm is far better than INTEL's 65nm in terms of performance and power consumption. The 65nm Pentium D Presler and Yonah showed near zero improvement on power consumption and clockspeed over INTEL's 90nm. The 65nm Pentium D Presler is still spewing out 170 watts and running at 3.4GHZ, and the Yonah is at 1.7GHZ. Both are no match to AMD's current processors built on 90nm with DSLSSOI.

If we look further into 4Q06, the INTEL 65nm Merom is having heat problems and can't be faster than 2.33GHZ. So much for the faster switching speed hyped by Jack Uldrich the fool.

INTEL has started doing 65nm for some products but will remain mostly 90nm for months to come. With the primitive copy-exact methodology, it takes a long time for INTEL to adjust all those FABs. AMD's 65nm starts in 2H06. Who will switch to 100% 65nm first? It's hard to say, because AMD CEO said with its advanced APM3.0, when AMD starts 65nm, it will be at mature yield and switch can happen very fast.

INTEL's 65nm is smaller on feature size, but its process is the simple and cheap Strained Silicon. AMD's expensive and complex DSLSSOI (Dual Stress Linear Strained Silicon On Insulator) provides 30% performance advantage over SS.

I do agree with the fool that a smaller INTEL is better for everyone.

AMD to rule 2006 full year unchallenged

1) Desktop: AMD Rules

The Pentium D 9xx Presler is the last of the Pentium 4 family, it got beaten up by old Athlon 64 X2s. Moreover, AMD is soon to release the Socket AM2 version of Athlon 64, which will widen its lead. INTEL's next desktop chip is the Conroe, which is not expected until 4Q06 and won't have a big impact for that quarter.

AMD will gain big in corporate desktop market, due to its performance and performance/watt advantages over Pentium 4. (Funny that INTEL can't create new chips, so they started to rename CPUs)

AMD has another 12 months to run free and FAB36 is cranking. I expect AMD to exit 2006 with 50% of the desktop market, slightly more than double its current 24.3% market share.

2) Mobile: AMD gaining fast

First of all, AMD will have 100% of the 64 bit notebook market for at least three quarters, and people are getting to know the importance of 64 bit more as Windows Vista approaching. INTEL's amaturish middle-east design team is having trouble adding AMD64 extensions into the Pentium M core.

Also, as JFTC recommendations implemented, Japanese vendors will start doing Turion 64 machines.

I expect AMD to exit 2006 with 25% mobile market share, a modest increase from 15.1%.

3) Server: AMD rules

Opteron simply has no competition, only uninformed people will choose Xeon.

INTEL is pretty much capacity constrained, it can't solve the chipset shortage problem for months. With so many FABs*, the only possible reason is that INTEL's yield is too low.

I expect AMD to exit 2006 with 45% server market share, almost triple the current 16.4% market share.

Overall exiting market share about 40%.

* California: D2 - 200mm - 90nm; Oregon:F20 - 200mm, D1C - 300mm 90nm, D1D - 300mm 65nm/45nm; Arizona: F12 - 300mm 65nm, F22 - 200mm, F32 - 300mm 45nm; New Mexico F11 - 200mm, F21- 300mm 90nm; Colorado: F23 - 200mm; Mass: F17 - 200mm; Ireland: F14 - 200mm, F24 - 300mm 65nm; Isreal: F18 - 200mm 90nm, F28 - 300mm

INTEL's next generation Merom may have no legs

INTEL renewed its pledge to the sinking Itanium by piling up $10 billion more on the dead horse. With AMD ready to introduce mainframe class Socket F 1207 Opteron capable of up to 32P glueless ccNUMA, INTEL is getting pushed into the ultra low end in the server space. The coming INTEL Dempsey processor will be no more than an imitation of the old Athlon MP, and will be only capable of 2P operation. Since INTEL's FSB can't scale up to more than 4 cores, and a 2P Bensley paltform with Dempsey or Woodcrest already has 4 cores, INTEL is running into an architectural limitation. With Dempsey, INTEL may get closer to AMD at 2P. However, once AMD introduces quad-core, even assuming INTEL introduces quad-core the same time, INTEL will still be inferior at every level, from 1P and up. The Merom 4 issue core won't change this picture at all, due to the same oudated FSB architecture. Itanium is also FSB based, but its EPIC instruction set requires little branching, and its massive cache may be able to help a bit. Clearly, INTEL's further reliance on Itanium is a sign that the Merom core based server technology won't do any miracles.

Thursday, January 26, 2006

AMD stock is under valued

Reporting $194.8 million profit or 50 cents per diluted share, Broadcom Corp (BRCM) went up 24.49% after hours to $73.10, leading to a market cap of $25 billion. Based on after hour price, BRCM's forward P/E is 73.1/1.74 = 42 .

AMD reported a 4Q04 profit of $206 million, AMD market cap is only $16.3 billion. AMD's forward P/E is 23.

Wednesday, January 25, 2006

64 bit computing power + GeForce 6100 for $349

All four emachines desktops are AMD64. It reads: "64-bit computing power from an AMD Sempron™ processor". You can also choose from eight Gateway Athlon 64 X2 models from CircuitCity, or choose from four Gateway models from BestBuy.

Now, that's pervasive 64 bit computing in action.

SUN's UltraSparc T1 is doing well

SUN sold about 20000 Opteron servers the last quarter, with about 8000 of the Galaxy line. SUN's progress in the Opteron space is kind of slow, maybe because SUN has little experience handling large volume of inexpensive gear?

I think SUN desperately needs something to compete against the HP DL385, which is a small business work horse with 6 hard drives. Currently, none of SUN's Opteron servers have enough internal storage.

However, the newly introduced T2000 with one 8 core CPU did quite well. SUN has already sold thousands of it. According to David Yen, most customers bought only a couple of the machines to test them out, but one telecomm company placed an order for 75 T2000 machines yesterday.

Overall, I think SUN needs to speed up things. SUN started the Galaxy project almost two years ago. My hair is turning grey waiting to see more of them.

AMD's current generation is more advanced than INTEL's next generation Conroe

We all know this already, INTEL is 5 years behind. But here is more detailed look at the core. The problems with INTEL's Conroe/Merom are discussed here.

Compared to Nebojsa Novakovic's speculations, Josh Walrath has offered us true insights.

Nebojsa Novakovic failed to mention the most fundamental design choice in both Alpha EV7 and Opteron: the embedded memory controller. The Alpha designers had long realzied that memory latency and bandwidth were the foremost limiting factors for system performance (there was a seminal paper which I wasn't able to find a link). Once you choose the route of embedded memory controller, something like ccHT is a must have in order to connect multiple memory controllers together, cache coherently. Nebojsa Novakovic also claimed that INTEL is using HyperTransport in their chipsets, which is probably misunderstanding of the basics. INTEL is not a member of the HyperTransport consortium, which AMD has a lot of control. Furthermore, AMD has only opened up HyperTransport for I/O, it has not yet licensed the true crown jewel, the ccHT to anyone yet, though it's planning on doing it.

All modern CPUs, such as Opteron, Power5, UltraSparc T1 and Alpha EV7 have embedded memory controller(s). The T1 has four memory controllers, and I believe Socket F opteron will have at least two memory controllers.

Oudated designs such as Itanium, Xeon and Conroe use shared FSB.

Nebojsa Novakovic suggests that INTEL should adopt HyperTransport since CSI is delayed until 2009. But things are not so simple. Non coherent HyperTransport is an open standard. But AMD kept to itself the ccHT protocol for inter processor communications. ccHT is what INTEL needs, and there is no way AMD give it away for free.

Michael calling Paul

Mike: Paul, what da heck is going on? I saw Apple's INTEL ad, it showed some fancy INTEL equipment, and the voice over was saying "dull little boxes".

Paul: Yeah, I know, I thought the ad is cool, they taped the newest tools we bought from Applied Materials. I assure you that Apple won't compete against you. Their OS is different.

Mike: But a lot of folks who saw the ad told me they heard "Dell little boxes".

Paul: Yes. Dull little boxes. I do agree Steve Jobs have some artistic sense, at least a lot of ladies think so. Of course, we men don't care, we only care about stuff like clockspeed.

Mike: You don't get it. That ad is delivering a subliminal message: we are dull.

Paul: You are dell?

Mike: No. Dell is dull!

Paul: Dell is dell? I don't get it, of course dell is dell.

Mike: He meant to say: We, Dell, D E L L, is, D U L L.

Paul: I see what you mean. Steve is smart, isn't he? But, I heard you told people before that he should shut down Apple and go home , he is probably returning the favour. Steve has bigger market cap than you now.

Mike: Paul, I take this seriously. Our brand is the most important asset we have. I talked to my lawyers. They think we can sue them on trademark infringement, unfair competition and false advertising and other grounds.

Paul: Trademark infringment?

Mike: Yes. Lanhan Act. Their use of the word "dull" is likely to cause confusion, and cause dilution of our trademark.

Paul: Indeed. When do you become so good at law? Did you study it in high school?

Mike: Paul, let's get back to the main issue.

Paul: One thing I don't understand, SUN had this ad about "DELL HELL", and you are not suing them?

Mike: I am not stupid. Suing them would be free advertising for them, I am not going to take McNealy's bait.

Paul: But you are upset about Steve's ad?

Mike: Most of my business is in PC, he is ruinning my bread and butter.

Paul: I see.

Mike: Do you have a part in this ad? Why are they putting it under "intel/ads/" URL?

Paul: Of course not. They came to our fab and shoot a video, the dudes in the bunny suits are not INTEL engineers, they look too good. You can see that right?

Mike: Okay. Is Jobs going to sell Windows PCes in their fancy boxes?

Paul: Well, I know nothing of such plans, but now that you mentioned it, it might be a good idea. We need some fashionable boxes to maintain market share.

Mike: Paul, let me remind you, we have 17% of the PC market, if Jobs gets a better deal, I am going AMD.

Paul: Mike, be honest with you, we can't cut prices any further, our unit share is dropping and our ASP is also dropping, we can't withstand more revenue falls.

Mike: But AMD is cutting prices like there is no tommorrow.

Paul: Sorry, I can't cut you more discounts, AMD's capacity is limited...

Mike: Good and bye now...

AMD readies compatible quad-core Opteron for socket 940

Quad-core Opterons will be demoed mid 2006, on compatible socket 940 infrastructure. According to Marty Seyer, senior VP of AMD's CBPCMSS unit: "To go from single-core to dual-core to quad-core on the same platform, that has never been done in the industry."

I bet this will be on 65nm process. Also, I wonder when we can see the socket F with 1207 pins. A couple of years ago, when I showed a socket 940 to some dudes only worked with INTEL CPUs, they were awed by the massive number of pins. More pins, more expensive, more features, more performance, it's that simple. :-)

AMD's message is crystal clear: buy a server with socket 940 Opterons today, you can do single core or dual core right now; if you need more CPU power juice later, swap out the CPUs with quad-cores, and sell the existing CPUs on ebay. That's so no brainer.

If you go the INTEL route, it will be hot and slow, and you need to change horses. INTEL relies on external chipsets soldered on the motherboard that must be upgraded to handle more shouting matches on the shared FSB. Those bought Paxville won't be able to use Dempsey, those bought Bensley won't be able to do Tigerton, no wonder INTEL is also making billions on chipsets and motherboards.... And the Paxville story told us that you may even need to drill bigger holes to support heavier heatsinks....

I wrote here that if AMD releases quad-cores in 2006, INTEL will be very dead, no way for INTEL to scale beyond 4 cores with shared FSB.

Tuesday, January 24, 2006

Hector's Legions Crossed Rubicon

AMD's market share has risen to 21.4% in 4Q05, up from 17.7% in 3Q05.

Like Joe Osha who downgraded AMD to sell, Goldman Sachs downgraded AMD to "underperform", because the "sleeping giant has awoken". Joe Osha's downgrade allowed Merill Lynch to grab $496 million of AMD shares at rock bottom price of $35.2. What about Goldman?

Checking with Nasdaq.com, we found Goldman Sachs a major bag holder of the fast declining INTC stock. Goldman held 61 million share of INTC, which lost a couple hundred millions of value since January 18, 2006. Moreover, Goldman increased its INTC holdings by 18.1% in the last reporting period ending 09/30/2005.

Yes, the INTEL the giant has awoken. Like Pompey the Great, INTEL is scared, it is fleeing and hiding. Under the cover of the VIIV curtain, INTEL is dividing its businesses into the Digital Home Group, Digital Health Group and Channel Platforms Group and various other groups. INTEL is abandoning the Pentium, the Intel Inside and the Intel logo. It has fled Rome in total disarray.

And the Caeser is striking hard and allowing the people of the world to enjoy the power of Opterons, X2s and Turion 64s.

Hector promised a Pax AMD Era to the world, under the customer-centric innovation directive.

INTEL is finished, five years of technology gap is impossible to overcome.

My advice to Goldman the bag holder: dump it now, or lose more. Mark my word.

SUN and AMD should merge

SUN still got $4 billions of cash, which AMD can put into good use and build two more FABs. SUN has an OS and Java, AMD has x86 everywhere. Both are looking at the enterprise market.... Most importantly, the two share some common values.

Of course, one guy is going to give up the CEO post. I think Dr. Ruiz has proven himself.

Monday, January 23, 2006

OMG! INTEL is fragged!!


If this is true, INTEL is fragged, finished, dead!
Opteron 265 dual core for $341, Opteron 270 for $476, 275 for $700?
This is insane. One can build a 4 way server for less than $1500!

I think the recent drop in INTEL's ASP (Average Selling Price) was partly due to further discount to DELL to stop DELL from going AMD. Now that AMD has cut prices so drasticlly, INTEL will have to match up for DELL, or the dude will jump ship. But, it's a futile effort, INTEL is between a rock and a hard place, DELL will keep sucking INTEL's blood, while INTEL is losing market share, as other INTEL vendors will leave and go AMD because they can't compete against DELL in the INTEL space.

At this point, AMD must play this game I proposed here. Wait until HP/Compaq takes a good portion of the US Corporate market, wait till DELL begs for mercy. DELL should not be given any favoured treatment. Let DELL grab more and more of the INTEL pie, at lower and lower ASPs.

AMD's risk factors: INTEL may discipline AMD customers

Quote from AMD's filing with SEC:

Intel’s significant financial resources enable it to market its products aggressively, to target our customers and our channel partners with special incentives, and to discipline customers who do business with us. These aggressive activities can result in lower unit sales and average selling prices for our products and adversely affect our margins and profitability.

AMD offers $500 million common stock, impact on EPS

AMD announced that it will sell $500 million common stock, and it would use $226 million to pay back 35% of the 7.75 percent senior notes due 2012. Let's compute the effect on AMD's 2006 EPS.

1) Additional shares, based on AMD's current stock price of $35 is:
500/35 = 14.3

2) Total diluted shares: 453+ 14.3 = 467.3 million, where 453 million is the fully diluted shares before the offering.

3) Reduction in interest payments: $226 * 0.0775 = 17.515 mil

4) Interest earned on the $274 million cash on assumed interest rate of 3%: $274 * 0.03 = 8.22 million

5) Assuming original EPS is E, we compute the new EPS due to the stock offering:

Net income = E * 453 + 17.5 + 8.2

EPS = (E*453+ 23.7)/467.3 = E * 0.9694 + 0.0507

Now assume original EPS of $1.8 (AMD's forward P/E is 20.6), plugging into the above formula, we get new EPS

1.8 * 0.9694 + 0.0507 = $1.7956

Clearly, if original EPS was lower, then this stock offering will actually increase EPS.

Earlier, AMD announced that it would force conversion of $500 million 4.75% Convertible Senior Debentures Due 2022 (dilution from these convertibles were already added), adding today's offering, AMD will reduce its long term debt by a massive $0.726 billion. From AMD's 4Q04 balance sheet, AMD has long term debt of $1.327 billion, it will be reduced to $601 million. AMD is getting to its 20% debt/equity goal in a snap.

Also keep in mind that the conversion of the 4.75% debentures will also reduce interest payment by 500 * 0.0475 = $23.75 million, or adding EPS of 23.75/467.3 = 0.0508.

So, considering the $500 million conversion and the $500 million stock offering, our adjusted EPS formula should be:

(E*453 + 17.5 + 8.2 + 23.75) /467.3 = E * 0.9694 + $0.1058

Considering the fact that Merill Lynch is the underwriter of this $500 million offering, Joe Osha's downgrade of AMD to "sell" on January 17, 2006 becomes very disturbing.

AMD entrusted Merill Lynch to sell $500 million of stocks. Merill Lynch then deliberately pushed down AMD's stock price.

What kind of treachery is this?

INTEL hit with a slew of downgrades

As I predicted, INTEL fell more after AMD's super report, it's now at near $21. As of now, it has been downgraded by 9 analysts. Finally, people have realized INTEL is in trouble. My prediction is AMD will exit 2006 with about 40% market share (run rate), as FAb36 ramps up and Chartered enters production. AMD CEO Dr. Ruiz indicated that by the end of 2006, FAB36 and FAB30 will have the same level of production. INTEL will start to see losses by the end of 2007. I expect large layoffs in INTEL in 2007. In 2008, INTEL may have to sell FABs to pay AMD damages for anti-trust violations.

The wishful analysts believe as INTEL is getting better, it will regain market. This is 100% wrong. Most people of the world, especially in the 3rd world such as India, still don't know AMD, more and more people are discovering AMD and its lead every day, the message is spreading via word of mouth. Unless INTEL can really leap ahead and overtake AMD in a significant way, AMD's market acceptance will keep growing exponentially. From forseeable INTEL roadmap, we don't see any INTEL product that can surpass AMD in 2 years. Instead, we expect AMD to substantially widen its lead and put INTEL further behind:

1) The new process developed with IBM can further improve performance by 40%. I expect all chips from FAB36 are made with this process. AMD is cutting prices across the board now, indicating release of higher speed CPUs are imminent.

2) AMD is working to use the Z-RAM technology to increase cache size by up to 5 times on the same die area. This technology is for SOI only, a process AMD spent years doing and INTEL has not yet mastered.

3) AMD's socket AM2 processors will be soon available, this is an improved core with support for DDR2 and virtualization.

4) Socket F Opterons will soon be available, this 1207 pin chip will feature multiple memory controllers and quad-core support.

5) Further architecture enhancement will allow Opteron to do glueless 32 way SMP.

6) Dual core Turion 64 will be available in mid 2006.

Sunday, January 22, 2006

HP to attack SUN with Solaris 10?

This is ironical. Why are SUN customers not buying the SUN Galaxy servers but choose to run Soalris 10 on HP BL25p, BL35P, BL45P, DL145, DL385 and DL585 Opteron servers?

Is Andy Bechtolsheim getting too old or too rich to roll out other Galaxy models? He had spent years at Kealia since he found it in 2001, and SUN acquired Kealia in Feburary 2004, almost two years ago, yet we have only seen three Galaxy server models, the X2100, X4100 and the X4200 -- see I know more about SUN's model numbers than SUN's CEO. The X2100 is obviously a Tyan S2865 box. I am not sure about that other two, but compare to HP's fleet of Opteron battle ships, cruisers, destroyers and frigates, SUN's product line is thin and incomplete, there is no replacement for the Newisys V40z, and there is nothing to match the HP DL385 small business workhorse which has 6 hard drives.

Well, SUN is going to say HP has no CPU, no OS and is becoming an assembler of other's technology such as Solaris 10. But talk is cheap, money works, HP is making billions with Opterons+Linux/Windows. Opteron is SUN's last chance, DELL is still 100% INTEL, if SUN blew up this oppurtunity, I don't see how it can survive. At that point, HP won't even need to pay a dime to acquire SUN -- Solaris 10 is open source and free for all. HP will happily stamp a HP OpenSolaris 10 label and merrily sell millions of Solaris servers. Support and services? No problem, HP will hire the Solaris 10 folks using the money made by selling Solaris 10 boxes.

DELL will spit at INTEL and kneel to AMD soon

DELL dude is good at bad mouthing others, the arrogant high school diploma holder who made billions with screw drivers insulted AMD, insulted Apple, and he insulted SUN and others. The insult on AMD has won him quite some favours from INTEL. But the dude is in a time of hurt.

Look at DELL's latest quarterly report, and look for the segment information, 52.6% of DELL's revenue ($7.324 billion) came from corporate accounts in Americas, however, if you look at the profits, 68.3% of the profit ($817 million) came from corporate accounts in Americas. In comparision, only 13.6% of DELL's revenue came from US consumer market and that only generated $80 million profit (6.68%).

You do see a slight decline of DELL's US consumer business from 2004 to 2005. Clearly, AMD's gain in consumer market had a tiny impact on DELL. AMD has grabbed substantial market share in high end server business, but AMD has not entered corporate PC business in a big way. In other words, DELL has not felt AMD's impact on its core profit engine.

But that's going to change. As corporations increasingly adopting the superior AMD64 platforms on the server front, it becomes more and more obvious that they should adopt the same architecture for desktop PCes. After all, just as Xeons are Pentium 4s, Opterons and Athlon 64s are almost identical chips, with the same architecture, same instruction set, the same performance and performance per watt advanatges. In fact, the socket 939 Opteron 1xx are the same as Athlon 64s.

AMD claimed in their 4Q05 conference call that the quarter's success was partly due to penetration into some corporate desktop business. As I look at www.hp.com, I saw more AMD business PC solutions, it even has this Athlon 64 blade PC weighing only 2 lbs. HP is the #1 x86 server vendor with almost 3x the x86 server revenue as DELL, it should be very convenient for HP to sell AMD64 desktops and notebooks along with Opteron servers to the same corporate customers for exactly the same reasons -- better performance, lower power consumption, AMD64 compatibility.

It really doesn't make sense for a corporation to choose powerful and cool Opteron 64 servers but uses slow and hot Pentium 4 desktops and 32 bit Pentium M or Core Duo notebooks. It is especially irrational for companies to buy 32 bit centrino and Core Duo notebooks, which will be obsolete in just a few months.

AMD promised that 2006 will be the year that it enters the corporate PC business in a major way. AMD is going to do it with HP Compaq, by bundling Opterons and Athlon 64s in deals.

Now, DELL will have to do something sooner or later. But anything done after Q2 will be too late.

AMD does not need DELL.

AMD's 2005 EPS adjusted by excluding Spansion

Since Spansion has been IPOed, it's useful to trace back 2005 and look how AMD did if Spansion was excluded. According to AMD's 4Q05 report, excluding the results from Spansion, AMD had operating income of $543 million for 2005, or $1.49 per share.

However, to compute net income, we need to substract the interest costs and taxes. A better way is to use AMD's financial statement filed in December 2005 which adjusted for Spansion. From Q1 2005 to Q3 2005, if Spansion is excluded, then AMD's net income would have been $225 million. Then we add the net income for AMD prime from Q4 2005, which had operating income of $268 million, substract interest cost of about $20 million, 4Q05 net income should be around $240 million. Therefore, we reach a net income of $225+$240 = $465 million for AMD prime in 2005(AMD's paid almost zero taxes).

In summary, net income for AMD prime in 2005 was about $465 million, with 409 million dilute shares, this is a EPS of $1.14. Based on AMD's Jan 20, 2005 price of $35.50, AMD prime's trailing P/E is 35.5/1.14 = 31.

Please note I did not consider the non-cash charge due to AMD's reduction of holdings on Spansion from 60% to 37+%. Our goal was to assume that there was no Spansion in AMD in the first place and compute its 2005 EPS.

INTEL is being driven out of the retail desktop market

Go to CircuitCity.com and BestBuy.com, and check the number 0f INTEL and AMD machines.

Best Buy:
INTEL desktops: 3 Pentium D, 2 Pentium 4, 0 Celeron.
AMD desktops: 4 X2, 8 Athlon 64, 4 Sempron.

INTE notebooks: 6 Celeron, 25 Centrino
AMD Notebooks: 11 Turion 64, 1 Sempron, 3 Athlon 64

Circuit City:

Desktops: 18 AMD, 9 INTEL
Notebooks: 15 AMD, 24 INTEL

Dekstop processors count for about 2/3 of INTEL's processor revenue.

Saturday, January 21, 2006

The hidden messages in INTEL's 4Q05 report

You can uncover more truth from INTELs 4Q05 report: INTEL CPU business is in a mode of abosolute decline in a growing PC/Semi global environment.

In Q3 2005, INTEL's total revenue from microprocessors was $4.936 + $2.331 = $7.267 billion.

In Q4 2005, INTEL's total revenue from microprocessors was $4.929 + $2.400 = $7.329 billion.

INTEL's Q/Q growth was near 0 (less than 1%), in a very strong holiday season. IDC reported 17% global increase in PC shipments in 4Q05.

One should keep in mind that the revenue from the mobility group included CPUs such as the INTEL StrongArm processors used in most PocketPCs. The PDA market is growing 20% year over year. INTEL signed up PalmOne and Research In Motion to uses its StrongArm processors; all DELL Axim PDAs and most of HP's iPAQs uses INTEL CPUs. So this part of the revenue (hidden in the mobility group result) must have grown substantially, and the conclusion must be that INTEL's PC processor business is not just slowing down but in decline.

INTEL expects a drop of up to $1.1 billion from 4Q05 to 1Q06, therefore the CPU revenue for 1Q06 could be as low as $7.329 - $1.1 = $6.229 billion.

In Q1 2005, INTEL CPU revenue was $4.944 + $1.917 = $6.861 billion.

In other words, in 1Q06, INTEL's CPU revenue could drop as much as (6.861-6.229)/6.861 = 9.2%, on a year/year basis.

Again, IDC projected over 10% year/year PC growth for 2006.

Where will the missing money and units be? Three letters: A M D.

AMD projected 70% growth for 1Q06 on a year/year basis.

Injured, Joe Osha started bashing INTEL

Accoding to this:

"In posting the weakest Q4 revenue result and Q1 outlook since the bursting of the bubble, Intel has made it clear how much competitive ground the company has lost to AMD. The extent of the losses has exceeded our expectations," Merrill Lynch analyst Joe Osha wrote in a research note on Wednesday.

Joe Osha made himself a fool by downgrading AMD to sell on the eve of INTEL Q4 earnings report. Joe Osha has a buy rating on INTC. I bet many investors lost money thanks to Joe.

As the Q4 2005 data shows, AMD is winning the hearts and wallets of American consumers, who always get the newest and best computer technology. The people in the third world provinces are still sticking to the INTEL brand, but the trend will propagate soon, in less than a year, we will see a global shift to AMD64.

As we can see from INTEL's Q4 2005 report, INTEL sales in the Americas dropped 10% y/y to only $1.8 billion. We can estimate that 85% of INTEL's sales were overseas. INTEL's 4Q05 sales in Asia-Pacific(excluding Japan) increased 16% over Q4 2004 to $5.1 billion, but was flat compared to Q3 2005, indicating a halt of its growth in the region, probably due to AMD's advances in China and India. Right now, AMD simply does not have the resources to have a presence in more provincial regions of the 3rd world. But those markets, though small individually, add up to a large percentage of nearly 40% (I estimate China was 10%, 50%-10% = 40%), those regions are INTEL's last safe haven and hideout.

Looking foward, AMD will grow share in Japan due to the implementation of the JFTC recommendations that banned INTEL deals excluding AMD. AMD will continue to grow substantially in China, the world's second largest PC market, as all top 3 Chinese PC makers have signed up with AMD due to customer demand for AMD64. The Internet is speading the message that INTEL is outdated. And the crumbling of INTEL stock is free advertisement for AMD -- I bet many CIOs across the world own INTEL stocks, and are busy dumping INTC to reduce losses.

Execution is the key for SUN

It was interesting to see the four founders of SUN get together and indulge themselves in the fond memories of the glorious old days. They are all showing their ages, not as sharp as 20 years ago (of course). One thing struck me was that Scott McNealy couldn't remember the model numbers of the X4200 and X4100 Opteron servers. This is not good, not good at all. SUN spent a lot of money on the Galaxy line, and there is serious money to be made in the Opteron business -- HP is silently counting the billions it got from x86. HP introduced 3 blade and 3 rack servers for Opteron. SUN is slow in delivering and upgrading the Galaxy line, SUN lacks something to compete against the HP DL385, which has 6 hard drives. And let me tell you this, when DELL gets on the Opteron business, SUN's hope of penetrating x86 volume market will be pretty much finished and the millions SUN invested would be a waste.

As I pointed out , SUN has enough technology, it all depends on execution. SUN can't expect to buy some company and hope they can grow revenue, SUN needs to focus and act fast.

INTEL's defective gene

Inside Intel : Andrew Grove and the Rise of the World's Most Powerful Chip Company (Hardcover) , ISBN: 052594141X .

Jackson, a columnist for the Financial Times, points out in his excellent book, Grove was also largely responsible for Intel's arrogance toward customers, aggression toward competitors, and pettiness toward employees.

Greed and exploitation.

One Amazon reader who claimed to be a former INTEL employee commented:
INTEL "resembles the former Soviet Union"

INTEL 65nm dual die Pentium D-950 Presler re-fragged

Another set of benchmarks, the INTEL Presler fighting against AMD's second team, the old Athlon 64 X2 4800+, the INTEL Pentium D 950 3.4GHZ with 4MB of cache got hammered on the head again and again -- no surprise here. What is interesting about this test is the Apache Bench which measures Apache HTTPD server performance in serving web pages. In a simulation of 50,000 users, the old dog Athlon 64 X2 4800+ handled 1453 page requests per second, the spanking new 65nm INTEL Pentium D 950 3.4GHZ Presler dual die with 4MB cache handled 986 requests per second, roughly half the speed of the AMD X2.

An Athlon 64 X2 3800+ did 1070 requests per second in Apache bench, 10% faster than the 65nm Pentium D 950. Try check the prices.

So if you are a web master, don't be fooled by the massive 4MB cache and 3.4GHZ clockspeed of INTEL, its obsolete architecture is a limiting factor. DELL is selling PowerEdge 850 rackservers equipped with INTEL Pentium D for $4000. It's your money.

Friday, January 20, 2006

AMD: INTEL is too far behind, not in AMD's rear view mirror

TG Daily had an interview with AMD's VP of server products group, Randy Allen. Mr. Allen basically said that nothing in INTEL's forseeable future roadmap (as far as 2008) can compete against AMD's Opteron of 2003. Mr. Allen mentioned that AMD is offering Opterons from 68 watts to 120 watts, it seemed that he forgot the Opteron HE dual core version, which is 55 watts, or 27.5 watts per core.

HP does not support 64 bit Solaris 10 on INTEL Xeon

Look at this Solaris 10 support matrix for HP's servers, you can see the BL25P, BL35P, BL45P, DL145, DL385, Dl585 servers support both 64 bit and 32 bit Solaris 10. The HP DL140, DL360, DL380, DL580, ML370, ML530, ML570 only support 32 bit Solaris 10.

Now the quiz: what is the difference between these two groups, why is one group 32 bit only?

Hint: the model numbers of one group end with 5, indicating they use powerful 64 bit processor, the other group's model numbers end with 0.

Hint: It seems one processor vendor is enjoying 100% of the 64 bit Solaris 10 x64 market.

Thursday, January 19, 2006

AMD may put 5MB of cache into CPUs

According to EE Times, AMD has licensed Z-RAM technology from a start up company. This technology can increase density of cache memory by five times. In other words, using the same die area, an Opteron can have 5MB of cache. Z-RAM only works on SOI, a process technology that INTEL has not yet mastered, INTEL is only doing Strained Silicon. All AMD CPUs are made on complex DSLSSOI (Dual Stress Linear Strained Silicon On Insulator) process to achieve higher performance and lower power consumption.

According to the startup which developed the Z-Ram technique: "Innovative Silicon has developed a true capacitor-less, single transistor DRAM - named Z-RAM for Zero Capacitor DRAM – by harnessing the floating body effect of Silicon on Insulator (SOI) devices. This technology is capable of achieving twice the memory density of existing embedded DRAM technology and five times that of SRAM yet requires no special materials or extra mask/process steps. "

INTEL Core Duo powered iMac performs poorly

Macworld tests if Steve Jobs is lying about the claimed 2X performance advantages of the Core Duo iMac over PowerPC G5. The tests show Core Duo was only slightly faster with two CPU cores. Whatever, INTEL's 32 bit technology is only 50% the speed of AMD64.

INTEL CPU revenue may go down 15% in Q1 06

INTEL's revenue for desktop and servers CPUs went down from $5.256 billion in Q4 2004 to $4.929 billion in Q4 2005, a 6.2% year over year drop. AMD's CPG revenue jumped from $0.73 billion in Q4 2004 to $1.3 billion in Q4 2005, a 78% increase.

INTEL projected $9.1 to $9.7 billion total revenue for Q1 2006, down from the $10.2 billion of Q4 2005, below the previously expected $10 billion. Moreover, since INTEL revenue from CPUs was $7.3 billion, INTEL's CPU revenue may go down as much as 1.1/7.3 = 15%, q/q.

However, the $7.3 billion figure used above also included revenue from INTEL's other CPU businesses, such as those for PDAs and cellphones. If we only look at the CPUs for PCes and servers, the revenue is only about $6.1 billion. Using this number, INTEL's Q1 2006 down fall for PC processor revenue can be as much as 1.1/6.1 = 18%.

AMD projected over 70% year/year growth for Q1 2006. AMD had CPG revenue of $0.78 billion in Q1 2005, 70% growth would put Q1 2006 at $1.33 billion.

Always remember this: the x86 market is a two horse race, a win-lose situation.

According to WSJ: "The shift to AMD is apparent in a sample of fourth-quarter sales by U.S. computer retailers by Current Analysis, a market-research firm. Nearly 69% of the notebook computers sold by Hewlett-Packard Co., the No. 2 maker, used AMD chips, up from 30% for the same period in 2004, the firm said. The share held by Intel, which has long led AMD in the portable-computer market, declined to 42% from nearly 70% over the same period."

So, AMD owned 58% of the US retail notebook market in Q4 2005. Soon, this trend will propagate to the third world.

AMD saves the day for global stock market

The global stock market was on a crash course yesterday after INTEL reported worst earnings for Q4 2005 and worse outllok for Q1 2006. Investors don't seem to believe Paul Otellini's claim that INTEL will gain share in 2006: people just don't see how.

However, after AMD reported 35% Q/Q growth and projected 70% Y/Y growth for Q1 2006, and IDC reported 17% increase in global PC sales, the market rebounded in a lot of joy, as people found the missing money. AMD's stock rose more than 11% and its target price was raised from $40 to $55.

AMD now has 1/4 of INTEL's market share and 1/9 of INTEL's market cap, and AMD is on the way up with its 5 year lead in technology. AMD's fair value should be around $50 as of today.

"AMD is seen as a bellwether and its outlook was enough to make investors overlook a dimmer forecast from Apple Computer Inc.", AP wrote.

Meanwhile, Prudential cut INTEL's target price to $17 and urged INTEL management to take a better look at its product lines and correct its miscalculations in financial projections in the wake of an industry wide shift to AMD64 technology.

Morgan Stanley raised AMD's target price to $55- $70 based on projected price/sales ratios.

Merrill Lynch, who downgraded AMD to sell just two days ago, admitted that they "have been surprised – blown away is more like it – by the extent of the company's gains".

Wednesday, January 18, 2006

AMD's share of x86 market reached 15% in terms of revenue

For Q4, 2005, INTEL reported $4.9 billion revenue in desktop and server CPUs, and $2.4 billion revenue in mobile processors. Total INTEL CPU revenue $7.3 billion. AMD reported 1.3 billion rvenue in CPU products. AMD's revenue share is 1.3/8.6 = 15%.

This is a two horse race. The missing 400 million in INTEL's revenue projection went directly into AMD's pockets.

AMD promised double the industry growth rate. Some analysts said what AMD is doing is far more than that. AMD's CEO Dr. Hector Ruiz said he likes to under promise and over perform.

According to IDc, global PC units grew 17% in Q4 2005. Noticebly, DELL units grew 20%. Since DELL is 100% INTEL and INTEL only grew 2% revenue, the data indicates that DELL is eating more INTEL pie and others like HP and Lenovo are pushed out of INTEL space and switching to AMD. Just as I analysed previously, keeping DELL 100% INTEL is good for everyone, except INTEL.

AMD fragged INTEL

AMD reported super earnings for Q4 2005. $0.45/share. In particular, its CPG revenue grew 35% over Q3, which is beyond anyone's wildest imaginations. Even I was only expecting 25% growth, 11% above AMD's projected 7-14% growth. (But I also put a 40% potential growth there in the "recorded" Hector phone call).

For Q1 2006, AMD guided flat to slightly down, in comparison to INTEL's 8% down.

Shares of AMD went up 10% after hours.

Both Joe Osha and Ben Lynch were at AMD's conference call congratulating AMD. If you recall, both downgraded AMD to sell just a or two ago. Joe Osha asked some idiotic questions. I felt that Ben Lynch sounded smarter, his firm dumped 22% of INTC already as of September 2005 and started to buy heavily into AMD, but Joe Osha's Merrill Lynch increased INTC by 15%. Maybe Merrill Lynch should fire Joe Osha and hire Ben Lynch.

INTEL stocks will fall more after AMD earnings

Some folks still don't get it, AMD is going to frag INTEL all over the place, you aint seen notin yet. You are still looking at 90nm products from FAB30, you haven't seen the flood from FAB36. Paul Otellini claimed that INTEL will gain share in 2006. Do you really believe this? Did he mean AMD won't be able to sell FAB36 chips?

On December 8 2005, INTEL gave a mid-Q guidance of 10.4 to 10.6 billion revenue, that was only 2 weeks away from the Q4 end. Was INTEL telling the truth?

More people will see the light when they compare AMD's result today with INTEL's. This is a two horse race, a win-lose situation. AMD tripled its wafer orders. FAB36 can easily triple AMD's production.

AMD will end 2006 with 40% of the x86 market share. Mark my word.

BTW: INTEL share prices plunged more than 10%.

Tuesday, January 17, 2006

Joe Osha fragged by INTC

Just like his buddy Ben Lynch at DB.com, Joe Osha at Merrill Lynch downgraded AMD to sell, right before INTEL earning's release, citing strong INTEL 32 bit product roadmap in 2006 (AMD almost exited all pure 32 bit market and is in 64/32 only). Joe Osha had a buy rating on INTC. However, INTEL reported a bad quarter with revenue of 10.2 billion, only 2% above Q3, 2005, far below seasonal trend of 8%, and below its own December 2005 mid-quarter projection of 10.4 to 10.6 billion. INTEL's Q1 2006 projection is 9.1 to 9.7 billion, down 8% from Q4 2005, 3% more negative than seasonal trend of 5% down. INTEL claimed that the 3% extra was because of inventory build up. Year over year, INTEL's digital enterprise group went down 5%.

So, INTEL is doing double down here. First, it went below seasonal Q4/Q3 growth, then it will go below seasonal Q1/Q4 from the already downed Q4.

INTC crumbled after hours.

Like Ben Lynch, Joe Osha's AMD downgrade was actually an INTEL upgrade. Unfortunately, unlike his smarter German friend, Joe Osha said "we are buyers", his firm increased its INTC holdings by 15% to 22 million shares --- expect this foolish bag holder to lose quite some money. We all know INTEL is in deep dodo.

AMD is expected to report 13% higher revenue tommorrow.

Saturday, January 14, 2006

INTEL Pentium brand end of lifed

Details are here. My comment: running out of ideas.

INQ folks need to learn reading SEC FORM 4

INQ reported that AMD's Chief Sales & Marketing Officer Henri Richard raked in $469,346 by selling 15,010 shares at $31.27, see http://www.theinquirer.net/?article=28891. However, as I review the SEC FORM 4 filing here, it's a different story. This was an option exercise and auto sell transaction. Mr. Richard first bought 18760 shares of AMD at various exercise prices, total cost $260,000. Then he sold 15,010 shares for a net gain of $200,000 in cash and he had 18760-15010 = 3750 more AMD shares under his belt. But keep in mind he has to pay AMT taxes on the 3750 shares he kept, so the $200,000 cash must first be used to pay taxes on the 3750 shares, then pay taxes on the gain from the 15010 shares sold. In the end, the gain is much smaller.

I was wondering why Henri Richard, an insider, could make such transactions on January 3, 2006, right after the end of 4Q05 and before the earnings report. But the fine print in footnote on the Form S4 indicated that this was an automated transaction following "to a Rule 10b5-1 trading plan adopted by the reporting person on August 19, 2005".

Mr. Richard was in fact losing a lot of profit on these transactions, these options has big time value which was lost in the exercise and sale. I guess Henri Richard didn't expect AMD to gain 35% Q/Q growth back in Aug 2005.

Another AMD VP, Thomas McCoy also adopted a similar trading plan.

Friday, January 13, 2006

German Crooks Are Smarter

In this article, the Fool exposed the Wall Street crooks. Basically, when the crooks rate a stock at buy, it's because they own the stock, if the crooks rate a stock strong buy, it's becase they want to dump the stock.

But that's too obvious, isn't it?

Today, Deutsche Bank downgraded AMD to sell, its analyst Ben Lynch cut AMD's target price to $26, because he believes INTEL is rolling out new chips for servers, desktops and mobile and 2005 Q4 will be AMD's last swan song. Incidentally, Ben Lynch has a zero (0) star rating on EPS estimate accuracy for most of the stocks he covers, which is kinda lame for a person that is supposed to earn his food by writing analyst reports -- a monkey could do a better job by throwing darts and hitting target 50% of the time.

Of course, we know from the Fool.com article, that the analysts have their motives, and being accurate is not their goal. Ben Lynch's analysis looks like both an AMD downgrade and an INTEL upgrade -- only two companies compete in 99% of the x86 market. Let's check Deuthsche Bank's INTC holdings on Nasdaq.com, we found:


see? Deutsche Bank held 88 million shares of INTC previously, but had already dumped 23 million shares, as of September 2005, it still had 65 million INTC shares to dump. INTEL's insitutional ownership dropped to 56.6%, there are more institutional sellers than buyers, but Deutsche Bank stood out as the largest seller of INTC shares. Another German bank, the DRESDNER BANK AG, sold 17.9 million shares of INTC, or 97.9% of its INTC holdings. Remember, AMD's FAB30 and FAB36 are in Dresden, Germany, and the folks there know what AMD has on its shipping docks and where they are heading. Perhaps, not by coincidence, the DRESDNER BANK AG started heavy buying of AMD stocks.

Clearly, if Ben Lynch tried to do a straight upgrade of INTC to dump it at higher price to other investors, he would be caught red handed. Being a smarter person, he embedded the INTC upgrade message in downgrade of AMD to sell. Of course, INTEL investors saw the AMD downgrade and were all cheered up, now they can comfortably buy the shares Deutsche Bank was dumping.

But look at this, folks, after fragging INTEL Presler with Athlon 64 FX60, and fragging INTEL Paxville with old Opteron 880s, AMD is ready to mass frag INTEL with 12 socket AM2 processors. AMD will soon release the socket AM2 Athlon 64 X2 5000+ and the FX62 (2.8GHZ), plus a slew of more powerful ones.

INTEL's Merom is not expected until Q4 2006, and the speed will be at 2.33GHZ due to heat problems. That means AMD will enjoy 100% of 64 bit mobile market for another 270 days. Windows Vista x64 is coming.

I also looked at AMD's institutional ownership data as 09/30/2005, there were two houses had the largest increase in AMD holdings. One is FRIESS ASSOCIATES LL... went from 0 to 7.3 million AMD shares, and the other is TURNER INVESTMENT PA... from 0 to 4.68 million AMD shares. So I look at their INTC holdings, surprise, both sold out their INTC shares.

Thursday, January 12, 2006

INTEL Merom processor having heat problem

The 64 bit successor to the 32 bit Core Duo seems to have thermal problems. According to tgdaily.com, INTEL will launch the Merom processor in September 2006, if it can solve the heat problem by then. The initial clockspeed will be 2.33GHZ.

AMD will enjoy 100% of the 64 bit market for another 9 months.

Wednesday, January 11, 2006

Setev Jobs plays the benchmark game again

We heard from Steve Jobs before that Apple PowerPC Macs were 2X faster than 3GHZ Pentiums before. Now, it's the other way around, INTEL Pentium is 2 to 5 times faster than PowerPC. Infoworld caught Jobs red handed this time.

Whatever, in any case, INTEL chips are only half the speed of AMD's. Jobs can't cheat on that.

Monday, January 09, 2006

INTEL Pentium Extreme Edition 955 (Presler) fragged one more time

AMD released the Athlon 64 FX60 CPU, which was hailed by folks at NVIDIA as the FASTEST PC processor in the world. AnandTech did some extensive benchmark, the 65nm Presler dual die chip is totally fragged.

Not surprisingly, reviews of the FX60 chi pooped up everywhere on the net, CNET did a review on Athlon 64 FX60 and no surprises -- INTEL's 65 nm Presler got fragged again.

DELL XPS is the ultra low end

From this benchmark by PC Magzine, the DELL XPS 600 is the lowest of all in every single test. I suggest the dude try out some real 64 bit chips before embarassing himself in the public. DELL is still a screw driver company after so many years, so pitiful.

HardOCP bought the DELL XPS and tried to get some tech support after some serious problems, they give DELL Support a score of 4 of 10. I think they were being generous.

Rahul Sood, CTO of VoodooPC, said this about DELL XPS with Quad SLI: "throwing Quad SLI at an Intel based gaming system is much like putting lipstick on a pig."

Sunday, January 08, 2006

A Condensed History of INTEL

1) Chip Design Age -- ruled by Andy Grove (1979 -- 1998)

Bob Colwell and his colleagues designed the Pentium Pro in 1995, which has become the core of Pentium III, Pentum M and Core Duo.

2) Manufacturing Age -- ruled by Craig Barret (1998 - 2005)

No advances in CPU design. Hyperpipeling was used to jack up clock speed which was servred as as a marketing tool, made INTEL a funamentally a manufacturing company. Bob Colwell quit. Pentium 4 hit thermal wall. 90nm shrink made things worse than 130nm.

3) Marketing Age -- ruled by Paul Otellini and Eric B. Kim ( 2005--)

Going back to Bob Colwell's Pentum Pro with some modifications. VIIV, Core Duo, Leap Ahead were invented as new brands and new labels for the old 32 bit technology. Single cores glued together with FSB is being used to compete against AMD's multi-core with Direct Connect Architecture. 65 nm showed little improvement on performance and performance per watt over 90nm.

INTEL Pentium D 950 (Presler) may overheat

There are two reasons that a CPU reduces its speed when running, one is there is little work to do, the other is the chip is overheating. INTEL Pentium 4 CPUs are basically overclocked chips, they can't run continuosly at their specified speed for extended periof of time, due to the extreme power consumption. INTEL employs a technique called thermal throttling: the CPU lower its speed to prevent melt down of the silicon core. Another technique INTEL uses to reduce heat is to add instructions that basically does nothing, this way, the chips looks running at full speed, but saves energy by reducing the work load.

According to the Tech Report, the 65nm Pentium D 950 (Presler) has a real heat output of 140-145 watts, partly due to a defect in the CPU design. It found that , with the stock INTEL CPU cooler, the Pentium D 950 has to throttle itsef to a lower speed. However, if using an expensive cooler like a turbo fan, the throttling problem is resolved.

Does Core Duo qualify for VIIV sticker?

VIIV was said to mean 64. But since INTEL was slow in copying AMD64 instruction set, the Core Duo and Core Solo are pitiful 32 bit chips. Does two 32 bit cores glued together qualify for VIIV?

Saturday, January 07, 2006

32 bit Core Duo Notebooks will fully depreciate in 6 months

AnandTech did some performance benchmarks on INTEL's spanking new Core Duo 32 bit x86 mobile processor. The results are unimpressive, for instance, compared to Pentium M 1.86 GHZ, the Core Duo 1.86GHZ with two cores showed only 11% performance increase in Business Winstone 2004 and about 22% in Office Productivity SYSMark 2004. It seems that the antiquated shared FSB architecture in all INTEL designs really hurts performance when you adding cores. Also battery life is no beter for the 65nm Core Duo, indicating that going 65nm provided INTEL almost nil performance/watt benefit. The Core Duo is only 2.16GHZ max, while the AMD Turion 64 ML-44 is a 2.4GHZ 64 bit processor.

Since Windows Vista requires 64 bit for full enjoyment, buying an obsolete 32 bit Core Duo for marginal performance improvement will be a fast route to depreciation -- if IRS approves.

Thursday, January 05, 2006

SUN should adopt customer-centric innovation

SUN is being run as an innovation-driven or innovation-centric company, spending huge pile of money on R&D, most of which generated little revenue and is being open sourced, probably in a hope that someone else can profit from it.

Sharing is good for mankind, but you have to have resources if you want to keep sharing. You have to make some money or you die out. One way to innovate and make money the same time is through customer-centric innovation, the prime directive of AMD. The more I think about this term, it makes more sense. You innovate to solve real problems of your customers, you innovate and get paid and you have more and happier customers. You make more money, and you hire better talents, you share more, and the cycle continues...

INTEl Core Duo(Yonah) delivers little

PC World did some benchmarks on Core Duo (Yonah) based notebooks from HP and DELL, and found almost no improvement on business application performance and battery life. The only area Core Duo shined was in multi-tasking. So if you run two heavy duty applications the same time on your notebook, there will be a benefit from using Core Duo, but expect battery life to be even shorter in that scenario.

DELL XPS in flames?

Look at this picture. Did DELL put a tempertature reading panel there?

DELL XPS PC will be fragged

Showing off gaming PCes at CES powered by the INTEL Pentium 4? That is not a bright move. Dude, the age for a high school kid building cheap PCes and making billions is over. Gamers today are smart people, and they know any Athlon 64 FX CPU will frag an INTEL P4 player in no time. What's so interesting in the DELL XPS PCes? Michael Dell was touting about its 30 inch LCD monitor. But for gamers, performance is the difference between life and death. An XPS will be slaghtered by Athlon 64 due to the sub standard performance of the Pentium 4, assuming the same skill level and asuming the XPS does not burn its motherboard because of the extreme heat.

Let's see how much longer DELL's $5 billion asset can sustain its $80 billion market cap. This company is so amateurish, and it's sad that corporate America is still sticking to this screw driver company for their desktops.

AMD should release quad-core in 2006

AMD's CTO Phil Hester, in his interview by CNET, derided INTEL for sending out development versions of future chips, such as Paxville and Dempsey for review. He said what AMD has in development is two or more steps ahead of current dual core Opteron. AMD is set to release quad-core Opterons by 2007.

The situation is indeed troubling for INTEL. As DELL CEO Kevin Rollins acknowledged, next generation of INTEL server chips will be behind AMD's current generation. I think Rollins was talking about the Dempsey CPU and the Bensley platform. At least today, Rollins seems to be more clear headed than Michael Dell who claimed that INTEL would overtake AMD in dual core performance. The Paxville proved itself to be so bad, it damaged INTEL's image irreversibly. Looking ahead, even the obsolete single core Opteron 246 was able to beat Bensley/Dempsey/Blackford on quite a few tests.

So AMD is far far ahead, no less than 5 years. INTEL's Bensley/Dempsey/Blackford is no more than an imitation of 2001 Athlon MP. However, with this dual FSB architecture, INTEL would buy some time in 2P 4 core area and attain a performance level closer to Opteron 280. Though AMD will still be unmatched in 4P 8 core and above with Opteron 880, 2P server market is the volume market. Apparently, DELL's CEO was counting on Dempsey to bring some parity to the 2P server market and to slow AMD's penetration, after the failure of Paxville.

For it to kill off the Xeons once and for all, AMD must speed up the development of quad-core Opteron. Since Opterons have no problems scaling up to 16 cores and above, and INTEL can't scale above 4 cores even with dual FSB, one single quad-core Opteron will outperform any 4 core Xeon system, and two quad-core Opterons will be like today's 4P Opteron 880, and will be unmatched by any other.

Tuesday, January 03, 2006

Workstation benchmark for INTEL Paxville performance

Hexus conducted a test of the Paxville, the tests were Alias Maya and Pro/Engineer Wildfire the results showd that INTEL's dual core is at about 50% the speed of AMD Opteron 880.

Monday, January 02, 2006

DELL CEO basically says INTEL leaps behind

Very interesting interview with the dude's CEO.

Q: Is there any concern that by not adopting AMD technology, that you're leaving a significant amount of business on the table?

Rollins: We certainly worry about that, and we track that very, very carefully. But if you look at where the inroads have been made, they've been made in those fairly small unit volumes, they've been in the four-way servers. That's where AMD has made its success. To some extent two-ways, but it's the four-ways where they've been taking the bulk of the market share. Again, are they doing well there? Yes. Are the four-ways as a percent of the overall x86 market very big? No. So you're losing some units by share in a fairly important category, but is it strategically a killer? No. Financially it's not. But it's something we worry about, so we're tracking that very, very closely as to what we would do.
Intel's got a better roadmap coming up, in terms of performance and in terms of thermals in the coming year, so that's been very encouraging for us.

Q: A better roadmap than they've had, or a better roadmap than AMD?

Rollins: A better roadmap than what they've had, but again, it gets us much closer to parity with what AMD has been doing.

So according to Rollins, INTEL WILL have something closer to what AMD has now. Is AMD sleeping?

64 bit entertainment is here

MSI has created a media PC powered by AMD64 technology. It incoprates an Athlon 64 rocessor, 802.11g for transmiting media over wireless and NVIDA graphics for video and game play. It can also act as a TiVo box to record video. The most interesting thing is that it runs the 64 bit version of Windows.

64 bit entertainment is good for the masses.

INTEL's vision is software

After reading this article by Business Week, I came to the conclusion that INTEL's goal is to become an omnipotent software company, all these stuff, such as entertainment, healtcare, financial services, are software. Of course, platform can be either software or hardware or both. In the case of centrino, it was a platform of hardware with 3 components, Pentium M , INTEL wi-Fi and INTEL chipset. However, to have a truly functional platform, you need to have the full software stack. I expect INTEL to invest heavily on the MacOS platform for its VIIV initiative. Since VIIV will be mostly a software specification with some hardware requirements, Windows VIIV will benefit AMD equally. To exclude AMD, INTEL may sign exclusive deals with Apple, or even acquire Apple for this purpose. Sounds good on paper.

But Bill Gates won't be amused, and expect him to retaliate.

Sunday, January 01, 2006

"Leap Ahead" reflects the same old INTEL-centric mentality

The "Leap Ahead" program is in stark contrast to AMD's prime directive. In essence, it is INTEL-centric versus customer-centric innovation. With AMD moving at light speed from $100 notebooks to 2048 node mainframes armed with quad-core glueless 32P SMP,INTEL is farther and farther behind on the technology curve. AMD quad-core will shatter any hope INTEL had on catching up with AMD, as INTEL's shared FSB simply can't scale to 4 cores. INTEL followed AMD to the book in the past few years. But it failed to follow AMD in one of the most critical directions when it actually desperately needed such guidance from its mentor:customer-centric innovation would be key to INTEL's success for its move to consumer business. Instead,INTEL is projecting a self image of a bigot trying to lead the way.

I predict that "Leap Ahead" will have the same fate as Itanium. With Itanium, INTEL tried to lead the world into EPIC computing by abandoning billions of investment on x86 sotware and hardware. The same intel-centric thinking is driving the "Leap Ahead" program.

Dual core server duel: Xeon vs Opteron benchmark

AMD challenged INTEL to a dual core duel, INTEL is not responding. For that, AMD has given a list of 10 reasons that INTEL won't duel.

For those who like to see an interesting fight, be disappointed. If INTEL does get to the stage, it will get beat up so bloody by AMD, you will shed tears of sympathy for the INTEL folks.

GamePC had a benchmark on the rare to find Paxville DP at 2.8GHZ, and INTEL got punched right and left, left and right.... Paxville had only about half the performance of Opteron in Apache bench.

Now INTEL is readying the dual core Xeon MP 7040, which is dual core for 4 way and up servers. Luckily for us, HP has submitted SPECINT scores for this spanking new INTEL wonder chip.

Here is the summary:

INTEL: 3.0GHz, Intel dual-core Xeon MP 7040, 4 cores, 2 chips, 2 cores/chip , SpecInt rate 57.2

AMD: Opteron (TM) 880, 4 cores, 2 chips, 2 cores/chip , specint_rate 75.1

INTEL: 3.0GHz, Intel dual-core Xeon MP 7040, 8 cores, 4 chips, 2 cores/chip specint rate 105

AMD: Opteron (TM) 880, 8 cores, 4 chips, 2 cores/chip , specint_rate 144

Now the math: (75.1-57.2)/57.2 = 31.3%,
(144-105)/105 = 37.1%

Now, let's look at number crunching speed, the SpecFP_rate scores:

INTEL: ProLiant DL380 G4 (2.8GHz, Intel dual-core Xeon Paxville), 4 cores, 2 chips, 2 cores/chip 40.3

AMD: ProLiant DL385 (AMD Opteron (TM) 280), 4 cores, 2 chips, 2 cores/chip , 74.4

Woops, INTEL dual core is only about half the speed of AMD in number crunching. Looks like INTEL needs at least a 5.16GHZ dual core Xeon to have to same performance of Opteron 280.

Now, if we compare 8 core, specfp_rate scores, INTEL looks tiny.

ProLiant ML570 G3, Xeon MP 7040, 8 cores, 4 chips, 2 cores/chip, SpecFP_rate2000, 48.2

Sun Microsystems, AMD Opteron (TM) 880, 8 cores, 4 chips, 2 cores/chip, SpecFP_rate2000, 153.

4 Xeon MP 7040 dual core is only 30% of 4 Opteron 880 in number crunching speed.

Maybe HP is not good at making Xeon servers? Let's look at IBM's dual core Xeon machine:
IBM eServer xSeries 346 (2.8 GHz dual-core Xeon Paxville, 2MB L2 Cache), SpecInt_rate 59.9, not bad, IBM pushed the score up two points.

AMD home made using offshelf TYAN Thunder K8QSD Pro (S4882-D), AMD Opteron (TM) 280, SpecInt_rate 76.8 .

Actually, the fastest AMD dual core CPUs are the Opteron 285SE and 885SE used in Sun's Galaxy line. The SpecFP_rate is 82 for 4 Opteron 285SE cores. I will update the link when SUN's score is published on spec.org .

Here is a preview of INTEL Dempsey processor performance benchmark.

(I changed the article date to 01/01/2006, as servers with INTEL's new chips won't be available by then)