Monday, April 30, 2007

The deleted Intel emails must be 100x more incriminating

WSJ reports that newly released Intel emails and internal memos demonstrate how Intel Corp. used anticompetitive practices against AMD in the Japanese microprocessor market.

Intel must have deleted a ton of 100x more incriminating email.

Terminating sanction is the only remedy for Intel's discovery abuses.

Friday, April 27, 2007

Hector gets the money to finish the job

$2.2 billion ought be enough to ramp K10 and R600 and finish Intel once and for all.
If Hectors needs more, he will get it.

The notes can be converted to AMD stock at $42, or about 50 million shares.

The folks who lent this money saw the K10 in action and decided they can make a huge fortune from the $2.2 billion invested.

Meanwhile, people found that the $65 x2 3600+ smashes Core 2 Duo in 64 bit computing. Dell customers also found the old K8 dog mauls quad core Intel.

Thursday, April 26, 2007

Some note that Intel's HPC sales to USA dropped to 0

The trend is set. Scroll down the page and see the light:

"AS to the Intel rep, ask him for his druggist's name. Then ask him how many DARPA HPC awards for computer technology Intel received at SuperComputing '06 last November in Tampa despite having 30 core 2 Duo and 1 Quad core entries. The answer is 0. Then ask him how many AMD got and for which chip. The answer is 4 for the 939/940. IBM got the other 8. Ask him where Intel finished in DARPA's Petascale contract competition. Answer is last IBM split honors with AMD/Cray and IBM has agreed to make the P7 interchangable with the AM3+. Sun's Rock was 3rd. Intel's wonderful technology has shut them out of the HPC market through the end of 2010."

Intel systems can't scale due to the FSB bottleneck.

Opteron rules.

Wednesday, April 25, 2007

More proof that Intel Core CPUs caused notebook explosions

A reader pointed out that Acer is doing battery recalls on AMD powered notebooks. Clearly, faulty batteries did not cause AMD notebooks to explode.

As I analysed long time ago, the Intel CPUs were the cause of the explosions so vividly shown on various web sites. In some cases, even houses and trucks were burnt down by Intel notebooks.

Tuesday, April 24, 2007

AMD = performance

AMD's R600 leads Nvidia by about 20% even with less and slower memory. The R600 is 29.6% faster doing Call of Duty 2, 21% faster on Oblivion, 16.3% on 3DMark06, 29% faster on Cadalyst C2006, 29% faster on Autodesk 3ds Max v8 OpenGL, 41.1% faster on Autodesk 3ds Max v8 D3D, 26.4% faster on Catia 02, 58% faster on Maya 02. It was a total slaughter.

Those tests with small differences are clearly sign of CPU bottleneck. Wait till you see it on K10.

Just to remind you, the 2.5GHZ K10 is 33% to 66% faster than Intel's 3GHZ quad core.

K10 is the cold blooded killer

2.5GHZ K10 is 33% to 66% faster than 3GHZ QX6800 quad.

Intel needs a 4.8GHZ quad core to compete against the 2.5GHZ K10.

Patty better exercise all his options and save some money.

Because by 2Q08, he won't have a job.

Monday, April 23, 2007

Upgrade time: X2 PC + 19 inch LCD for $399

DELL deal that frags Pentium XE 965 in all benchmarks.

Intel's graphics are not suitable for Vista Aero interface.

Official: AMD leads by 50%

AMD officially confirmed that K10 will be 50% faster than Intel's fastest quad core.

That's a staggering and crushing performance advantage.

Both AMD and Intel CPUs are roughly at the same clockspeed at about 3GHZ, as both are limited by the same physics and manufacturing equipment they purchased from tool companies, such as Applied Materials.

Below the top speed, there are about 10 speed grades. The fastest is about 50% faster than the slowest. Thus, each grade is about 5% faster than the one below.

On desktop, Intel now has about 10-12% IPC advantage with Core 2. This puts AMD's X2 6000+ at roughly the same speed as Intel's E6700. This situation has caused tremendous problems for AMD.

Now, with 50% performance advantage, the slowest K10 CPU will be faster than the fastest Intel CPU.

It's reported that Intel execs have exercised their options. Congrats to Patty. At least, he can make some money before becoming jobless next year.

C2D's pony trick rediscovered

Cache is pretty much the only trick --- I told you that long long time ago, and told you again recently. Also, Core 2 temperatures are danerously high for normal operation. Scientia explained why Intel hasn't released anything faster.

The Israeli dudes aren't so smart after all, certainly no match to the grand masters at AMD, whose K10 design will frag Intel to death.

Saturday, April 21, 2007

3GHZ Opteron SpecInt_rate score published, Woodcrest fragged

2P dual core Opteron 2222SE SpecInt_rate2006 score: 56.6

2P dual core Xeon 5160 SpecInt_rate2006 score: 50.7

4P dual core Core 3GHZ Xeon 7120N, Specint_rate2006: 50.7 (8 cores)

2P dual core Opteron 2222SE, SpecFp_rate2006: 52.1

2p dual core Xeon 5160 (3GHZ) SpecFp_rate2006: 45.1

2P quad core Xeon 5345 SpecFp_rate2006: 51.7

My comment on the AMD inventory situation

Hector made one mistake: he didn't drop X2 prices sooner.

But that mistake is now forcing him to take the corrective action which is not too late. By selling X2s at around $125 or even lower, the $900 million inventory can be cleared by the time of K10 launch. Since AMD basically has one FAB (FAB38 will ramp down soon), its switch to K10 will be rather quick.

The above moves will deliver multiple blows to Intel.

First, the fire sale of X2s will make all Core 2s tough sale. AMD solutions are far more attrative when priced competitively---AMD solutions always come with good graphics, sufficient for Vista. Expect Intel to be seriously wounded in this first round.

Then, once the X2s are cleared, come the K10s with crushing performance advantage. This second blow may prove fatal to Intel. Don't be fooled by Intel's profit numbers, its cash flow that matters. To stay competitive, Intel has to spend a lot of cash on new FAB technology. AMD on the other hand, has no problems using Chartered and TSMC...

As of today, you can get a DELL Desktop PC with dual core X2 3600+ and Nvidia graphics, loaded with Vista, all for $339.

Friday, April 20, 2007

AM2+ quad crusing at 3.04GHZ

65nm K10 desktop quad at 3.04GHZ. Intel's 45nm Penryn got pre-fragged by about 50%.

Intel may refuse to replace CPU if you overclocked -- Intel always knows how to extract money.

Thursday, April 19, 2007

K10 starts shipping in Q2

AMD strikes back.

I predict that 3D gaming will be 90% faster on AMD platform. The R600 is only optmized for AMD and that gives a 50% lead.

AMD's 1Q07 was largely due to its failure to adjust processor prices. I have said long time ago that AMD must create a dual core Sempron to prevent Intel from dumping the Pentium Ds.

AMD's inventory reached a staggering $900 million at the end of 1Q07, while this indicates a potential problem for some, it offered AMD the golden oppurtunity to flood the hell of the market with massive number of the Athlon X2s at ultra low prices. The prices keep falling, you can get an X2 for $60. I expect the X2 6000+ to be sold at around $105 once K10 is out.

Thus, starting from Q3, we will see the collapse of the prices of the K8 and Core 2 architecture. All K8 and Core 2 will be sold at below $120. The result will be a huge Intel inventory build up, leading to another price crash, Core 2 CPUs will be sold at $65, and K8s will be sold at $55. K10 will make all the money.

With the K10 onslaught, the overcapacity of the x86 market will lead to the BK of Intel by 2Q08.

My RECOMMENDATION to AMD on strategy

To quicken the collapse of Intel, AMD should PERMANENTLY lower the price of desktop CPUs.

Except the FX line, all AMD desktop CPUs should be sold at price between a narrow Window of $50 to $150. Since AMD's APM process is so good, the chips they produce are nearly identical, this allows AMD to mark their CPUs at higher rating.

The pricing table should look like the following
1) Athlon 64: $50 to $60
2) Athlon 64 X2 4000+: $65 but with 10% volume-- this makes sure all Pentium Ds will be sold at lower than $60.
3) Athlon 64 X2 5600+ and up : $85 to $95
4) K10 dual core : $120 to $150

With such a pricing scheme, Intel's ASP will collapse. In less than three quarters, Intel will BK.

AMD's 45nm Shanghai quad core running

The 45nm Shanghai is at least 40% faster than Penryn. The low power 65nm K10 can do 1.9GHZ at 45 watts, about 10 watts per core.

AMD's game plan is simple. Barcelona will take the server market, render all Intel Xeons total junk. The 4x4 platform will get Barcelonas, so high end gamers will be happy. On the lower end, high clock K8 dual cores will be competing with Intel's Core 2 in the $100 market.

Patty will be out on the streets by 2Q08.

Intel's collapse will be fast and violent.

Wednesday, April 18, 2007

Anand learnt his lesson the tough way

Instead of being the button-pushing pumper he was during the Conroe IDF, Anand did some more testing this time and exposed Intel's sick lies. Intel claimed 45% performance increase on gaming* with the 45nm Penryn.

However, Anand showed almost %0 performance boost in modern 3D games. The 3.2GHZ 45nm Penryn scored 145.3 FPS in HL2, and the 2.93GHZ Conroe did 131.3 FPS, the gain was 10%. However, that was with a 10% clockspeed increase. Clearly, Penryn shows no IPC advantage at all.

Fundamentally, Penryn is just a dumb shrink coupled with some more cache.

Intel's Q1 saw revenue below $9 billion. This was achieved when AMD is in a transition period, and Intel had all the cards.

Soon, Intel will be finished off by K10.

Meanwhile, people are flocking to the 3GHZ X2 6000+. And, look at this deal.

Some reader pointed out a cute observation. A 2x Barcelona ought to be enough to frag anything Intel will have in the future. Eight K10 cores has 8x the performance, nothing in Intel's future roadmap can scale that high.

* I hope Intel wasn't using Microsoft Minesweeper for the benchmark.

Sunday, April 15, 2007

Intelers are in denial

As I have shown in this comparison: the C2D and K8 architecture are neck to neck in terms of performance. Except the SSE advantage, the only trick C2D has is cache. Once the cache is reduced to the same level of K8, the C2D loses its advantage. Similarly, when C2D's cache is not enough, it's also slower than K8. As we noted earlier, 3GHZ Opteron conclusively frags 3GHZ Woodcrest in both integer and floating point computation.

Essentially, Core 2 and K8 cores have equal IPC in terms of integer, and Core 2 is at least 15% slower on FP.

K10 core will have 40% IPC lead over Core 2 on integer (per core).

K10 will have 200% IPC lead over Core 2 on floating point (per core).

Friday, April 13, 2007

Athlon 64 X2 3600+ at $57+free shipping

$57, and you get a Pentium XE 965 killer. Geroge Ou is quiet now. He is waiting for Intel to drop the Pentium D950 to $45.

Looks like AMD is determined to kill Intel at all costs. That's the right move. What da heck. Flood the market with 37 million dual cores per quarter, leaving Intel the rest of the market.

I heard that Hector Ruiz has ordered the K10 troops to commence their attack early. The sharp drop of prices of K8 is just the preparation for much faster chips to be launched soon. The K10 will open a performance gap that will never be filled -- as Intel BKs in 2Q08.

Thursday, April 12, 2007

Intel's 2Q08 BK is coming

AMD has everything in place. Capacity, OEMs, channel, design wins, K10 will simply wipe out Intel. Wonder why DELL has just released a 2P Opteron server called PE 2970?

Once K10 is out, all Core 2 will be sold at below $100.

Wednesday, April 11, 2007

More proof that Conroe is a one trick pony: cache

Now, we have the most direct comparison of the Core 2 and K8 microarchitecture.

DailyTech tested the Conroe-L, which is a single core Conroe with 512KB of cache. The Conroe-L 440 is rated at 2GHZ. Comparing to Athlon 64 3500+ (2.2GHZ, 512KB L2), the Conroe-L is definitely an undedog.

As you can see from the gaming benchmarks, the Conroe-L trails Athlon 64 in all tests. In one game, the Conroe-L is 13% slower. Clearly, the inherent average IPC advantage of Core 2 over K8 in desktop environment is less than 10%. I pointed out this long long time ago.

As we have seen previously, the 3GHZ Opteron frags the 3GHZ Woodcrest in both integer and floating point peformance.

K10 will simply kill Core 2.

PS: those who compare the prices of Conroe-L and Athlon 64 must note that Conroe-L is a future Intel chip no one has seen in the wild yet.

No pain, no gain, AMD rules

Hector's ATI move was the greatest strategy moves since Caeser's conquest of Gaul or the D-day of the allies.

Intel will be so obsolete by June 2007. AMD will simply frag Intel in multimedia, graphics and microprocessors.

The idea is simple. For every AMD GPU sold with an Intel processor, the price will be $35 higher. All of Intel's IGP business will be junked.

I saw discussion on AMD's past failure to capture more market with Opteron. The blunder was due AMD's timidness when dealing with Intel.

Had AMD sued Intel on Opteron launch, Intel would have been killed already. But, now it's not too late. K10 will officially finish Intel off. DELL will be 100% AMD on K10.

Tuesday, April 10, 2007

AMD should definitely seek terminating sanction

It seems that Intel won't be able to produce the deleted emails. Those emails worth at lest $8 billion dollars.

Monday, April 09, 2007

Athlon 64 X2 3GHZ at $239

Athlon 64 X2 2.2GHZ at $99, X2 3GHZ at $239 (out of stock already). I hope AMD has enough stock of these.

AMD's pricing is disruptive, considering that Intel's Core 2 Duo E6700 (2.66GHZ) is sold at $509. AMD enjoys a 2x price/performance advantage.

The K10 killer will be demoed later this month. AMD should put an Intel box next to the K10, and show that the K10 is 40% faster. It should be noted that on the day K10 is demoed, Intel will cut prices significantly.

AMD's game is simple. Use high clock K8 to battle Core 2 Duo, and take the high end market with K10. In floating point, K10 is 3.6x the speed of dual core K8.

Once K10 is out, all K8 and Core 2 CPUs will be sold at below $100. I expect the Core 2 Duo E6800 to be sold at $97, the E6700+ at $96, and the Athlon 64 X2 6000+ at $95, the E6430 at $94, the X2 3600+ will drop a buck to $64. The K10 CPUs will be priced from $150 to $999. Intel will suffer a violent death.

Intel can only hope K10 is not as good as AMD claims to be. Otherwise, there is no way Intel can survive.

I think even die hard Intelers must agree that AMD is better value:

With $509 you can buy a Core 2 Duo 6700, or you can spend $239 on an X2 6000+ which leaves you $270 to buy an AMD 690G MB+ 2GB ram + a 320 GB HDD. The whitebox vendors will have no problem making the right choice of CPUs.

The 2.4GHZ Athlon 64 X2 4600+ boxed CPU+fan is now at $116 with free shipping. The slower Core 2 Duo E6300 is at $181.

The Athlon 64 X2 5600+ (2.8GHZ) is at $179.

Sunday, April 08, 2007

Intel's Quad core only slightly faster than the $64 X2 3600

You pay $65 for the X2 3600+ and get 65.5 fps in Rainbow Six: Vegas.
Or, you pay $1000 for the Intel's fake quad core QX6800 and get 88 fps.
If you look at other benchmarks, the picture is worse for Intel.

The pathetic "quad core" performance of Intel CPUs shows how outdated Intel's architecture is.

Saturday, April 07, 2007

AMD's klling game

As I estimated here, AMD is capable of producing 37 million dual core CPUs per quarter. This might explain why AMD is selling dual core CPUs at $65.

The drastic price drop of Athlon X2 CPUs indicates that there will be soon a dramatic performance hike on AMD CPUs. Intel's BK may be sooner than I think.


Wednesday, April 04, 2007

3GHZ Opteron conclusively frags 3GHZ Woodcrest in both integer and floating point

The truth is out. 2P 3GHZ Opteron beats 2P 3GHZ Woodcrest in Specint_rate2006 by 2.5%.
It kills 2P Woodcrest in SpecFp_rate2006 with a lead of 15.5%. Intel needs a 3.5GHZ Woodcrest to compete.

Last but not the least, a 2P Opteron is 13% faster than a 2P Clovertown Xeon 5355.

Now, imagine what K10 will do.

Monday, April 02, 2007

Study confirms my theory that Intelers have lower IQ

Survey result is here. The study had a sample size of 5000 and found 69% of the people who use Intel powered computers had greater learning difficulties. George Ou immediately comes to mind as an example. The study found AMDers to be smarter.

My old analysis is here, it was based on my observation of the Intelers posting on this blog.

Based on the study, one can conclude that Intel engineers are also dumber than AMD engineers, because Intel engineers use Intel computers and AMDers use Opteron.

AMD 8-Core to slaughter Intel

AMD is using two K10 quad core and HT3 to produce a 8-core monster.
I suggested that AMD should bundle two Opteron dual core with HT long ago.
Some Opteron boards work just like that, but they are using two sockets.
Now AMD is finally doing the cheap trick -- connecting a K10 quad to another with HT3, the other quad has direct access to memory.

What can Intel do? Packing 4 dual cores on the FSB? Yeah, you get 4 Conroes hanging on the bus, each core getting about 100MHZ.