Thursday, March 09, 2006

Post IDF news summary

Conroe/FX60 generated much of the fun for IDF.

Time for the post IDF realtime news:

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

37 Comments:

Anonymous Anonymous said...

I know you are biased on AMD, but you back up your views by facts and reasoning.

Keep up the good work. Your blog provides tons of good information to a lot of viewers!!

10:17 PM, March 09, 2006  
Blogger Sharikou, Ph. D said...

I found one thing odd: DELL didn't cheer for the NGMA during IDF. Maybe they were just afraid of Osborne.

10:24 PM, March 09, 2006  
Anonymous Anonymous said...

Deep down, both Intel and DELL know the trend is against them. They are losing market share in a big time. They are still doubtful whether NGMA can save them.

Another reason that DELL is silent, DELL is going to AMD ?

10:36 PM, March 09, 2006  
Anonymous Anonymous said...

This Intel plea for support in the future is paththetic and sad.. man what have these idiots been doing with all that money we have all given them over the years.?? not listening to the public thats what, we want 64bit with lots of memory they give us Itanic that wont float.
Even National semiconductors had 64 bit processors.!
All your links have real wins for real products and no BS. I think that Dell and anyone who who cant see past Intels weak performance and poor leadership dosn't deserve our respect, Thats where I think Intel is at
I dont care what BS Intel is promoting now or in the future they have blown it.
My advice to them is go behind closed doors, sack the idiots get the engineers back control, produce some clean competative chips, stop confusing the market with STUPID cpu names (best name was Pentium -even if it was crap), introduce these new chips when they work and in volume not some 6 months from now geezz
Disssatisfied Intel user.

11:30 PM, March 09, 2006  
Blogger Sharikou, Ph. D said...

Who is a good engineer replacement for Intel CEO? I can only think of one guy, Mooly Eden, the Israeli dude who led the Pentium M project and now the Core project. He is considered the INTEL saviour now and probably has some engineering background on relatively newer technology. Other INTEL VPs only had experience on 486/386.

11:45 PM, March 09, 2006  
Anonymous Anonymous said...

Notice how Nvidia with its adoption of Hyper transport can do designs for AMD..
Its about time Intel came into line to support the Graphics innovators instead of actively compeating with them.. maybe they will make friends instead of enemies. HT is a open design with lots of potential higher speeds wider busses.
With optical connections deep into CPU's and GPU's will be they way of the future abosolutly.. anyone with half a brain can can see that.

11:58 PM, March 09, 2006  
Anonymous Anonymous said...

There was some ex Intel guy, who did a web cast recently (last 6 months) about x86 design and Itanium, and all the little interactions various additions have.. but more importantly that design is about giving the public what they want not what engineers think they want.?
cant think of name?

great. webcast he was - is a genius ,I vote for him..?
great standup comic as well..sharikou

12:08 AM, March 10, 2006  
Anonymous Anonymous said...

I followed your link on VP's at Intel, and looked on Intels site. Now I know why Intel is in the state that its in, those type of people need a total lack of integrity to sleep at night, no wonder Intel was taken to court, Those anti-competative practices are the stock-n-trade of those idiots, thank god I have no Intel stock... I wouldn't be surprised if they use ENRON accounting to coverup all their deals.

12:22 AM, March 10, 2006  
Anonymous Anonymous said...

I could run Intel.! First thing would be go make friends with AMD pay them at shite load of cash adopt intergated memory controller and HT bring it into the best Northwood P4 design and make it compatible with the AM2 socket, hell I would add memory controller and HT to Itanium and have a dual core - one Itanium and one Northwood P4 talkin thru HT and shared cache ala AMD and still have it sit in a AM2 socket. all the while I'd sell and make Opteron under license..

12:41 AM, March 10, 2006  
Anonymous Anonymous said...

www.theinquirer.net writes:
ANANDTECH has a re-run of the comparison between an overclocked FX60 and the next generation Conroe. This time around, they had their own benchmarks running, they used the right BIOS and drivers, they exactly the same. Even then, well the FX60 takes a battering. And the 2.66GHz Conroe being previewed was a high end mainstream part, the 3GHz Conroe will make the FX60 look like a Celeron D.

6:18 AM, March 10, 2006  
Anonymous Anonymous said...

Later INQ Reported: Intel supports guerilla benchmarketing

"Intel must be pretty desperate about AMD gaining market shares while Intel made the analysts news with earnings warnings two quarters in a row. So desperate, actually, that the worlds largest chipmaker broke its own rules. Intel people we talked to about the fact were surprised and all could not remember that chipzilla ever did this. ...

While one may criticize the fact that Intel did that in the first place, the setups really looked comparable judging after what Hexus and Anand reported. Mind you that Intel themselves gives out excellent guides to reviewers on how to configure a system and in its own benchmarks coming with review kits always provides exact data on how a competing PC was set up. You may not even get heavy on the hardware sites for taking the chance – it´s their business to do so, and how much more certain about a mature platform can you be, if the chipmaker itself set it up? And for AMD, a setup it is."

6:31 AM, March 10, 2006  
Anonymous Anonymous said...

TechReport described the benchmark condition here:

"We used test systems pre-configured by Intel before the show, and we had very limited time to conduct testing or inspect the systems. We were not allowed to look inside of the case of either PC, and the scope of the benchmarks we were allowed to run was defined by Intel. We weren't given the leeway to record our own custom timedemos for the games, and we didn't have enough time to run each test three times or even reboot between the tests....

our role really was confined largely to clicking a few icons and menu items to kick off a test and then writing down the results."

6:32 AM, March 10, 2006  
Anonymous Anonymous said...

IDF: Tanya Harding crown herself the champ?

6:34 AM, March 10, 2006  
Anonymous Anonymous said...

I think Bob Colwell is a better candidate for INTEL CEO position, see

http://www.theinquirer.net/?article=14310

7:42 AM, March 10, 2006  
Anonymous Anonymous said...

the 3GHz Conroe will make the FX60 look like a Celeron D :-))))))))

7:44 AM, March 10, 2006  
Anonymous Anonymous said...

AMD will release a 4.8GHZ Socket AM2 on 65nm process by the time Conroes hit the shelves 6 months later. So hold on.

7:47 AM, March 10, 2006  
Anonymous Anonymous said...

It's sad, INTEL reached 3GHZ in 2002, it's still at 3GHZ today, zero progress in 4 years.

7:50 AM, March 10, 2006  
Anonymous Anonymous said...

except that conroe will have like twice the IPC as netburst, so 3 gigahertz then is not the same as 3 gigahertz now

9:55 AM, March 10, 2006  
Anonymous Anonymous said...

So, Netburst is hot and slow? When can I get a Conroe?

10:11 AM, March 10, 2006  
Anonymous Anonymous said...

Bob Colwell is the man, He was probably forced out by the marketers at Intel because he spoke about what the public want, rather than screw the public for as much as we can.
Anyway the conroe demo will only make IT managers wonder if Intel has forgoten where Intels real focus should be, the server market first!.

1:22 PM, March 10, 2006  
Blogger Sharikou, Ph. D said...

Nothing of the IDF benchmarketing has changed my assessment that INTEL is five generations behind AMD64. We can now safely say those gaps won't be closed in 2 years, as INTEL return to the 1995 Pentium 3 start line. Bob Colwell indeed had the vision, but now it is probably too late. AMD will launch its super weapons one by one and ahead of INTEL. The AMD64 architecture is so much more elegant and scalable. Pentium 3 is last century's technology, adding a bunch of cache and opmitizing a few instructions are hardly innovation, they are enhancement at best. Colwell said that he left a todo list on the Pentium 3 and Israli amateurs picked them up, and did it a fairly good job.
The direction today is not to pursue highest single thread performance but focus on overall system throughput, AMD64 shines with its modern Direct Connect Architecture. INTEL will be stuck with ancient FSB architecture for years to come. Unless INTEL brings out some real innovation real fast, it won't survive another year of massive market loss.

1:43 PM, March 10, 2006  
Anonymous Anonymous said...

here's a link to insightfull articles from Bob Colwell our man to head Intel.
No he's to good for them, AMD should get him as their Intel-speak man.

1:45 PM, March 10, 2006  
Blogger Mark Wendman said...

Sharikou,

I think you are overly dismissive of the recent Anand / IDF benchmark. It was not rigged, it might have been slightly imperfect, but more important the advantage is clearly intels. NGA Intel devices - Meron Conroe and Woodcrest are very good designs, and they have clearly proven that neither Hypertransport nor Integrated memory controller are all that they are cracked up to be especially with respect to small core count CPUS and small CPU count Servers ( ie less than or equal to 4 ).

IMC is not in the least critical to throughput when the emory bus transactions no longer suffer from P4's notorious REPLAY upon 31 stage pipeline stalls / reloads. Cowell himself said that the Haifa design team is merely following up on and improving upon the old list of suggestions the P3 team had as things to do next to carry forward improving throughput on short pipeline X86. You can listen to his online talk at Stanford where he says this as part of the history of the early Itanium design troubles.

While you obviously have an axe to grind ( over and over ) right now your words are so hollow and so empty in meaning with respect to the intel NGAs as to belie a mild obsession on your part to badmouth what are obvious excellent revisions which have resulted in the best X86s being now from Intel.

Obssession and ranting denial in the face of noticeable improvment is sick. Conroe is a good, even great design and Intel does have the manufacturing expertise, 65nm capacity for volume dual core and this changes reality with respect to your sick over the top badmouthing.

While Intel is a bit late, they bit the bullet and undertook fundamental excellent improvements to the basics of the core, the cache, the SSE3 FPU and and the 4 issue double wide execution logic. Your wave of your bad mouth cannot counter the facts nor that Anand / Intel at IDF undertook a fair but quick test, which Anand issued a correction to a minor error of his openly - given the test was done quickly it was pretty representative of the performance differnces between the 2.66Ghz Conroe and a slightly overclocked 2.8Ghz X2 .

You can go on and on and rant and rave like a lunatic but it won't change the performance advantage Intel has produced in Conroe even if obviously this is almost a nightmare for your sick mind.

Take care, look youself in the mirror and realize the Intel Haifa design team has done well, very well indeed, and give the guys credit for their excellent engineering accomplishments. To say otherwise is flat out outrageous lie and indicates a twisted sick mind you may possess.

Best Regards,

Mark Wendman

9:28 PM, March 10, 2006  
Blogger Sharikou, Ph. D said...

Mark,
How can you prove to us that the benchmark was not rigged? I concluded that the Conroe box must have been doped, because Woodcrest will be 10-20 slower than Opteron 280.

I simply repeat my challenge to INTEL: you want our money? prove it!

If Intel wants to show its future CPU is faster, just give two boxes to two 3rd parties and do a fair test.

You are aware of the BAPCo history, the crippled compilers and the recent Skype deal, right?

You asserted that Conroe is good, did you take one home for a test drive? or did you merely rely on Anand?

Intel Israel did OK work, Bob Colwell praised them, but they are basically amateurs who lack 64 bit computing experience. AMD's folks grow up with 64 bit. So far, we have zero proof that Intel Israel managed to get AMD64 compatibility worked out. There was no mentioning of Conroe's basic 64 bit capabilities during IDF. Anand didn't even run CPU-Z to see if x86_64 was there.

The whole Conroe/FX60 thing was an insult to our intelligence.

10:19 PM, March 10, 2006  
Anonymous Anonymous said...

Sharikou barks, but the caravan rolls on. This time THG writes:
"This huge "leap" already sparked discussions what these new products mean for AMD. Clearly, Intel is out to trump AMD. It is our observation that the Core micro architecture scales better with every added MHz than Intel's Athlons and Opterons. The higher the clock speed of Conroe, the more significant the distance to AMD's current Socket 939 Athlons. The mainstream 65W Conroe is expected to top-out at 2.66 GHz at introduction; an Extreme Edition is expected with 3.33 GHz, which theoretically means that AMD will have to offer the equivalent of an 4+ GHz Socket 939 Athlon to match the fastest Conroe."
http://www.tgdaily.com/2006/03/11/idf_spring_2006_wrap_up/

3:28 AM, March 11, 2006  
Anonymous Anonymous said...

Another cheerleader? This one didn't even get Intel's permission to pushed the buttons of the Intel "setup".

8:01 AM, March 11, 2006  
Blogger Mark Wendman said...

Sharikou,

1st - the P4 when running optimized code was not nearly as bad as observed on P3 code base. P4 optimizations did something simple called loop unrolling - obvious as to implications, and when done, was often considerably faster with the same source code. So despite the obvious mistake of relying on an optimizer that was hardly ever adopted for enduser products, it was still an engineering advance, and while P4 was miserable at unoptimized code it met certain key design targets. These targets proved wrong and thats life. Everyone messes up one time or another, and this messup had significant enduser and corporate consequences.

As to whether the test was rigged, I don't know, and neither do you. But my intuition is that it was not, and my intuition is pretty good, partly based on experience of having worked at Intel at the start of the 386. While you are so emphatic about dissing Intel, your perspective is clearly over the top.

What are you going to do, if the test was not rigged and you are proven to be wrong? I won't say you are wrong, as my conjecture as yours as to the actual validity of the tests - neither are proven one way or the other YET. I am honest about that. Are you?

As to feeling misled about prior data of older CPUs - I will not dwell on it because the main factor is what Cowell says regarding the near zero acceptance of P4 compiled optimizations in REAL end user application codes. If you cannot understand that, well maybe you are not as insightful, nor perceptive as you claim yourself to be.

You certainly seem rather intense in a way that can trivially blind one to facts of substance, especially when the details are subtle.

But having worked at Intel - despite my departure, I have real respect for the firm as a whole and also admit that sometimes like any large corporation, bad politics can hold sway over critical decisions that have considerable and durable impact. Cowell's talk is very honest about how the early days of the Itanium were dominated by Albert Yu and his not so positive management skills in an obviously huge undertaking.

I also agree that Bob Cowell would make a fine X86 design VP, as he is obviously very skilled. As to dissing or denigrating Intel Haifa's design team, I do take issue with that - they are a crack disciplined and very hardworking team, far better than you indicate - they work like crazy, and will make the NGA rollout very successful, is my guess, even if you are obviously dismissive, for what ever reasons.

As to the 64 bit extensions...

You make the claim as if 64 bit computing is some kind of magical design threshold different than the prior word size extensions that Intel migrated through time and time again through all its history? What kind of conclusion is that? Rather silly and simply wrong.

While their implmentation EMT64 is deployed on fewer Intel cpu designs as a percentage of product so far, I suspect they have shipped at least as many 64 bit CPUs in quantity than AMD has,if not they will very soon do so and quickly overtake AMD in actual qty.

As to the comparison between Woodcrest and Opteron 280, was this chip to chip, or system to system where HT in AMD has an obvious advantage in CPU core count scaling, that no-one denies? That, if the case, is not the core per se, but the HT links being better than intel's shared bus in larger core configuraitons ( ie beyond 2-3 cores ).

You have not responded to my points re HT and IMC on the desktop? They do help lower system costs and simplify inventory but in small core count if the bus traffic is not clogged by P4's notorious REPLAY, it is no big deal for performance is my take. Intel's solution in theory might be more expensive to manufacture except that with 65nm in HUGE volume before AMD, it does not matter, plus significant is that when dual core proliferates as STANDARD ( forced by Intel's transition to likely majority dual core in mainstream product in the coming quarters ) AMD's capacity diminishes significantly for dual core CPU output in relation to INTEL's massive already rolling multiple 300mm 65nm fab sites.

I think time will prove that I am correct re beleiving the recent Anandtech IDF results as substantially correct ( with the minor correction by Anand in the 2nd revision - his error on HIS personal benchmark of ?FEAR - incorrect graphics settings fixed in the 2nd try).

As to specifically 64 bits importance, it won't matter in the least until Vista, and even then it is largely irrelveant for the majority of endusers. Vista is weird as to the degree that eye candy and high end graphics are used in the OS, while nice to look at it forces you to buy a huge step up in CPU and Graphics performance for just that mostly, which on the face of it is not my cup of tea.

If I get Vista ( I will likely delay and delay, till an SP1 probably ), I don't think it will be all that useful in terms of applications I mostly use, and for what I use computers for. The only advantage of Vista seems that from the get go MS is trying to be a bit more responsible about relase quality since everyone has hammered on them for buggy and attack prine codes in the OS and included browser.

I hope this discussion keeps you trying to be honest, rather than the ranting you seem to be on a roll on recently.

Yours Truly,

Mark Wendman
http://mark-nano.blogspot.com/

8:10 PM, March 11, 2006  
Anonymous Anonymous said...

Sharikou,

Intel has claimed many things in the past but I do not recall that they ever claimed their perfomance edged over the opteron in the past. They talked about their process technology and they did talk about them introducing dual core to the market first. It would be sucidal and stupid for a company with Intel size and history if they claim at their IDF that they have a product that fairs better perfomance than a smaller company product if they really do not.
Their credibility can not survive somebody else (3rd party benchmarkers) proving them wrong. Intel knows they would have no control over bench marking once their product is out in the market. So, I would trust they understand that if they rigg any demos, it may kill them if somebody would prove them wrong. You also have to give credit to Anands. His website is well respected by many techies. You did not have problems with reviews and tests he ran proving the Opteron is better than the Xeon, did you?
Intel is a paranoid company and they screwed up badly recently. But their paranoia is so fundamental that it showed at IDF with them guarding their demo systems closely. They have done this before. As a matter of fact, it is a practiced culture.
At IDF they talked about the future because they do not have an answer now for current AMD products. Assuming AMD will not combat them with a new product that beat their claimed performance in time for Conroe release, they can claim they would have better perfomance per watt.

One more thing, you keep rambling about the core archeticture being 5 generations behind. If AMD's "elegant" design you keep talking about get a beating in performace by a 5 generations old technology, what would that say about the elegancy of AMD's design.

9:26 PM, March 11, 2006  
Blogger Sharikou, Ph. D said...

Mark,

Let's look at it this way. The theoretical max speed of the CPU is determined by the clockspeed. In the ideal situation (such as running a simple loop), the 3.6GHZ P4 can be 2x the speed of a 1.8 GHZ Athlon 64. So the problem becomes IPC for real application loads, because no one is using their computer for the fun of a simple loop. That IPC number definietely varies depending on the state of the CPU and memory. IMC increases IPC by reducing memory latency -- which can cause 300 idle cycles in (unidirectional) FSB based architectures. HT can also help, with HT, I/O is not using the memory bandwidth. IMC+HT is an elegant and scalable design. It's much harder to jack up FSB frequency as every device listening on the FSB introduces noise. Intel Israel's solution was to throw cache on the problem, as long as the code and data are in the cache, you are happy, little latency there. Now, you can imagine there are code that can fit into the 4MB cache of the Conroe and can't be fit in the 1MB cache of the FX60 (1MB because FX60's caches are not shared). In that case, Conroe will definitely have an edge. No IMC can match the speed of cache. But, do you think that represent an average usage case? My reqiest to Intel is simple: let's do a set of typical tests, Intel's favourite BAPCo SysMark if you like, but limiting testers to a set of pre-arranged buttons is doping by implication. There are also other unknown parameters both hardware and software, without subjecting the box to anti-doping tests, the results are useless. Intel definitely used Anand as pawn to infuse credibility into something of zero credibility otherwise.

Tests show AMD64 CPUs run 10-40% faster in 64 bit mode, there are many reasons, for example, in 64 bit, AMD64 has double the registers, which can be viewed as 0 latency cache. Somehow, EM64T runs slower in 64 bit mode than 32 bit mode. According to Intel, it had 4 teams study doing 64 bit on x86 and gave up after all concluded it was not doable. So, it was AMD invented x86_64, and Intel is apparently still wrestling with that problem.

As for Intel's manufacturing, it's really bad. I estimated that Intel's yield is only 30%. Intel has at least five 300mm FABs and seven 200mm FABs, they are pitiful. AMD has one 200mm FAB and is cranking out so many kinds of Opterons, Athlon64s, Smeprons, Turions, Geodes, Chipsets, and is taking 21.4% of the market. Once FAB36 is converted to 65nm, AMD will be able to take over 50%. AMD's manufacutring is years ahead of INTEL, my friend. APM3.0 is The most advanced in the world, no one else is even close. The only advantage Intel has it has the money to buy newer equipment to try out new things, like it's trying 450mm wafers now. But, yield is the key.


I pointed out that Intel is five generations behind, sadly, IDF didn't change any of that, not even 6 months later.

My opinion is: no matter what Intel does today, it is already too late. No tech company can survive such a huge technology gap. Dr. Craig Barret once said if you are one generation behind you are dead.

IDF will speed up Intel's fall by creating an massive Osborne effect. A lot of people are really unhappy to find that stuff like GHZ and hyperthreading they paid good money for are actually useless.

10:59 PM, March 11, 2006  
Blogger Sharikou, Ph. D said...

One thing I'd like to add. AMD has mastered the most advanced SOI technology. Using the a set of complex and exotic techniques, it can jack up frequency by 40%. Also, it is using Z-RAM ( a capacitor-less, single transitor memory design only available for SOI) that can increase cache density 5x. So using the same die area, AMD will have 5x the cache. I am quite positive that AMD can implement these two things by the end of 2006 on its 65nm process.

11:05 PM, March 11, 2006  
Blogger Mark Wendman said...

Sharikou,

hmm, If the microcode implementation of EMT64 is less efficient than AMD64 implmentation you may be correct re 64 bit code potential speedup on Intel vs. AMD. Right now who's running 64bit on the destop X86?? If it needs to get fixed, I am guessing they will fix it when the market needs it.

On 32 bit code, you are likely to be very wrong.

As to 65nm process progress at AMD, I am less optimistic for AMD than you are, as the 65nm migration is not so simple. For any optical lithography at 65nm and below, edge roughness and pattern transfer is getting harder and harder for successive shrinks and I suspect Intel is up the learning curve faster than you give them credit.

I doubt that Intel is at as low a yield level as you imply, as that is not how they do copy exactly for fab startup. They don't ramp nor transfer until it is both manufacturable at high yeild and very profitable. So your claims are pretty hollow and silly. They have not done anything but copy exactly transfers since the late 80's, and yields are world class and always getting better, in all of their processes. They know their stuff and you can deny this ad nauseum, but your claim is hollow.

APM is definitely helpful but it is not the be all nor end all as you claim. Especially if the process integration is not encountering fundamental issues - APM is window dressing on an already working process, or not particularly useful to be kind to debug very low yields that are materials or integration related.

APM requires knowledge of centering targets mostly for parametric optimization - and gives NO clue as to actual yield issues that can be integration related. ( meaning if the process is NOT working well enough, valid statistical inferred process targets are often impossible to derive - duh?) APM can only center by near realtime tweaking of tools and not fix dead processes. Sorry but there are no miracles under the guise of TLAs ( 3 letter acronyms ).

Intel in all likelihood has APM but without calling it as such. They certainly have far more resources to implement anything as complex as APM and you can bet if they think it necessary to do in the fab, they will have done it and just not presented it to the outside world esp calling it by APM. ( inline process feedback is all it is, and they fund research woldwide on metrology but you have to read the literature to go past TLAs, to understand what they do publicly but don't advertise )

While that is not enough to confirm that they have it, it is enough for me to say you are just ranting as you seem to do and do, like the energizer bunny ranter!

Intel is very strong in fab AND design, and when push comes to shove upon shipping NGAs, you'll be pissing in your pants as to how it will shift your reality.

So keep ranting, it is becoming less and less likely you have any clue as to what will actually transpire.

Best Regards,

Mark Wendman

10:39 AM, March 12, 2006  
Blogger Sharikou, Ph. D said...

Mark,

On performance, let's just wait for 6 months and see what NGMA can really do.

On yields, no matter how you look at it, Intel has about 20x the capacity of AMD, but is only producing 4x the chips. Yields or FAB utilization. Your take. INTEL is building more FABs, as it can't produce enough chipsets to match its CPUs. That is what we were told by Intel execs.

APM3.0 can even reroute a wafer on the fly to change recipes, this is simply unthinkable with copy-exact.

AMD already have 65nm processors at hand, their goal is to reach mature yield (near 100%) when it ramps.

Intel will see a 20% revenue drop for 1Q06. For 2Q06, I project another 20% drop to around $7 billion. I see 3Q06 operating losses for Intel and mass layoffs in 2007. Intel simply won't have enough cash to feed 100K workers by then.

11:11 AM, March 12, 2006  
Blogger Mark Wendman said...

Shariko,

Copy exact is a system of copying manufacturing procedures, it is not necessarily eliminating the path of adaptive processes.

Copy exact does the following - EXACT toolset at every common process factory (the foundation), extensive cross training of personnel across sites, and pretraining before ramping a new site by dispersing a new fabs team, before ramp to be trained before needed. And running the same process ( except for adpative procedures ) at every site. GOod manufacturing practice in my eyes and makes the most revenue in the business anywhere. Yea that is terrible eh? I think it is great business.

Nowhere does copy exact preclude the methods of APM to be used, in fact they are likely to be used in whole or in part without the name used by AMD, as computer controlled manufacturing procedures that are copy exact (mfg procedures) across multiple fabs.

Forget the dire predictions, they don't cut it. even the bad Q4 was only $500m off a $9B nominal target, and you'd be hard pressed to tell me that was anywhere near 20% decrease. There is alot of upside to Intel with the new product transitions, and AMD I'd be curious what is happening even on 90nm in Fab 36 volumes??? even if they are slow parts compared to the Intel NGAs.

3:02 PM, March 12, 2006  
Blogger Sharikou, Ph. D said...

Mark,

Intel has already started seeing the Osborne effect. Dell is selling Intel servers for less than $500 now. People are expecting huge discounts on INTEL's soon to be end of lifed products.

6:12 PM, March 12, 2006  
Blogger Mark Wendman said...

The sky is falling oy vey, but Intel won't fall as low as you fear or possibly hope.
What can I say, please refute my points, don't tell me about osbourne, ozzy or otherwise.

There is older inventory what can I care about? Socket A Semprons next?

You can buy for $299 after rebate Dells too and the CPU costs Dell very lttle, which is just fine even if it is possibly a problem for AMD. After all Intel's die costs are not very high, even on 65nm, nevermind 130 or 90nm.

Intel always make a ton of profits, like BILLIONS as they should.

At least part of the discussion was interesting to encounter someone calling you on your bluffs eh?

Hope you turn out to be a super Conroe customer, and enjoy the product..
It is well designed and going to play fast games for some, and super CAD, simulation work for others..

6:25 PM, March 12, 2006  
Anonymous Anonymous said...

"They (Intel) have not done anything but copy exactly transfers since the late 80's, and yields are world class and always getting better, in all of their processes. They know their stuff and you can deny this ad nauseum, but your claim is hollow."

Nonsense... we're not talking about the Intel of the 80's. Intel got zero advantage from it's 90nm process, and it appears 65nm is going the same route. To compare - AMD is still on 90nm, but even at 65nm, INTC offers no current product that is comparable in terms of thermals or performance.

Some of us need to get out of the 80's and update our thinking!

Plus

6:53 PM, March 12, 2006  
Blogger Sharikou, Ph. D said...

Mark,

Actually, I projected a 15-20% q/q revenue drop for Intel's 1Q06, which you would not believe back in Feburary. Such massive sequential Intel revenue drop must be unimaginable for most people, especially in a growing global PC market. But it did happen and I accurately forcast it. The "Osborne effect" was already affecting Intel's revenue, as Intel already introduced NGMA in the fall IDF of last year. AMD's market recognition, public awareness of Intel's trouble and the demise of the Netburst, AMD's increasing capacity, AMD's legal crusade...are all shrinking Intel's market share and reducing Intel's sales.

For those know about Intel's NGMA, they have the following options in the next two quarters:

1) Buy Intel (P4, core duo)
2) Wait for NGMA
3) Go AMD

Now, those who choose 1) are not the smartest, you agree? I don't think Intel can count on such idiots to be the majority.

Those chose 2) must not be in a hurry to get a new PC. Also, the situation at the time of NGMA's arrival is probably not in Intel's favour, as AMD will have launched preemptive strikes in May and June.

So, 3) is the right choice.

Operating loss in 3Q06 is thus very real for Intel.

9:34 PM, March 12, 2006  

Post a Comment

Links to this post:

Create a Link

<< Home