Wednesday, May 03, 2006

Steam Engines versus Diesel and wishful thinking

There is too much confidence about Intel capabilities. It seems that a lot of analysts, investors, and technology enthusiasts think that Intel can pay its way out of any problem, thus, if AMD is Intel's problem, Intel will use its resources to kill the problem, and AMD will be destroyed. I disagree with that point of view, my reasons follow.

Intel is a greedy monopoly. The business model depends on both characteristics to be able to exploit customers and to have happy employees stuffed with stock options. That is one reason to think that Intel won't incurr in any price wars (1). Management, who relies on stock options as their main source of income (2), wouldn't want to do that; not even for long term good prospects, their exploitative and greedy attitude demand instant satisfaction. "In the long run, we are all dead" (Keynes) would seem one of their mottoes.

But they can try a similar approach: If the crux of their problems is that they have been outcompeted, and the market finally became aware of that, then the company has to offer better products. But since they are so hopelessly behind in the technology (While AMD is already doing Silicon On Insulator, Integrated Memory Controller, the upcoming single die quad cores, the coprocessor architecture that HTT 3.0 allows, APM; Intel "counters" with things such as "CopyExact!), they can't come to the market with truly better products.

The next option is to try to outcompete AMD following the brute force approach of putting more of everything known to work and selling the package for about the same price the competition sells. That is a sort of price-war: Not to manufacture products of equivalent costs to that of the competition's while selling them at steep discounts, but to manufacture extremely costly products to sell them at about the same prices as the competition. That is to sell a 4Mb L2 cache at the same price as 2Mb, 65nm at 90nm price.

That approach has been tried many times before in many markets and technologies. The latest steam train engines were by far the most complex ever designed and built, they had the largest boilers, etc.; because only through sheer sizes they were able to be up to the capabilities of the uprising Diesel/Electric engines. And the efforts to keep up were futile in the end:

The design ideas and techniques of Steam Engines were exhausted, played out, but Diesel's had the whole world ahead.

Intel's new lineup of products are steam engines. There aren't any new ideas to further the basic Pentium Pro architecture; I, for one, am skeptical from the professional point of view (summary here), about some of the design compromises Intel chose; and in particular, there are some assertions that my considerable understanding of this subject are not capable of explaining, such as the claims of higher power efficiency, or how is Intel going to successfully deal with the problem of trading off combinatorial complexities for marginal benefits so pervasive in all engineering disciplines; especially with Intel's track record.

I recently heard Henri Richard from AMD repeating the common saying that not even with 9 women it is possible to have a baby in one month; that is, it doesn't matter how much money Intel may have, the design gap is not possible to be closed in less than two years.

The problem is that AMD is in high gear, not in a reckless drive, but *cruising*, ramping two times more production capacity, with the highest market share it has ever had, highest revenue share, highest mind share, highest credibility.

Thus at Intel they have to follow for years the route of profitless existence that AMD followed until its recent emancipation. I wonder what makes the market think that AMD, which has the same management and engineers who reversed the roles with Intel at the brink of bankruptcy, all of a sudden are going to screw it all and at the same time, Intel's management, who allowed this reversal of roles, is suddenly aware of all of their mistakes and going to pull off a momentous come back after those two years.

That is wishful thinking. The market overvalued Intel and undervalued AMD. AMD demonstrated its worth, flew to the point of reaching $42 per share, too high so as to not force the comparison with Intel, thus forced the market to decide whether Intel was overvalued or not. But still the market's money was on Intel, every positive news about Intel was received with joy, every positive news about AMD with concern; that leads to a very easy to understand psychological phenomenon: Wishful thinking.

There are those who say that Intel has managed in the past to quash AMD's rebellions, that's true, but AMD never managed to get this strength before, Intel hadn't screwed it up so badly; in the past Intel could dump products in the only segment in which AMD was competitive, and succeeded because AMD never was able to get out of the "value desktop player" typecasting; today, it is Intel who is outcompeted all across the board but in mobiles in which they only have a bit of "residual" competitiveness.

The wishful thinking that supplies the credibility for more promises and more dissapointments will make the slide toward insignificance of Intel and Dell just slow and uneventful, but equally inexorable.

(1) There is a distiction between "channel clearing", what Intel is doing and proper price war.
(2) Unconfirmed


Anonymous said...

A good call, in any technical exploits changing lots of thinks a once is bound for failure, slow and steady is a sure bet

Anonymous said...

i have a different opinion. take a look at Intel's dual core transition from Pentium D to Core Duo. from benchmarks done by various 3rd party site, Core Duo is actually competitive against Athlon X2, and in some area, surpass X2. then the question comes. if Yonah can do it, why can't Merom/Conroe/Woodcrest do better?

Eddie said...

For sure "Core" is going to be faster than current AMD products.

But that doesn't mean that if AMD wanted, it couldn't follow the same brute force approach to put 4Mb in a cache, march toward 65nm on the "double time", include a fourth execution path, etc.; follwing that approach AMD would produce chips much costlier than the ones it already has.

That is the basic argument, Intel products are costly brute force improvements, and they have peaked already.

AMD may come with real innovation and dust "Core"s for good. Perhaps not this year, though. That's why I write covered calls.

Anonymous said...

so even if intel is very behind on those "frontier" technologies, is it possible for intel to develop similar and competitive technologies (CIS for instance) in a short time?

George said...

Nails In The conroes Coffon.

#1 HyperTransport alows the prosessor to havec better acsess to the menory.

#2 Core IS just an improved pentium pro From 1995!

#3 Athlon 64 on the other hand is an archectecure that dates back to 2003.

#4 Intel is run by marketing not R&D!

#5 woodcrest wich is 2 conroe was beat by opteron 280 in alot of benchmarks

Anonymous said...

more Nails in conroes cofon

#6 in cinabench a athlon64 2800 socket came close to woodcrest with 3 less cores

Anonymous said...

is it possible for merom/conroe/woodcrest to perform better with single threaded applications, but will perform a lot less in multithreaded enviroment?

Eddie said...

so even if intel is very behind on those "frontier" technologies, is it possible for intel to develop similar and competitive technologies (CIS for instance) in a short time?

Henri Richard's words remembering the common saying about Intel and the performance gap: "Not even with nine women you can have a baby in one months"

The core interconnect is one major component of a microprocessor, not trivial at all to develop, test, and fine tune. Those tasks take many months to be done.

In general, being optimistic with Intel, at least 2 years to catch up

Eddie said...

is it possible for merom/conroe/woodcrest to perform better with single threaded applications, but will perform a lot less in multithreaded enviroment?

Yes, mostly because with a single demanding thread, the core that is executing it grabs the whole L2 cache for it.

I am changing my opinion about the shared caches, but initially I though that its sole purpose was precisely to improve performance in single threaded conditions.

But I have been thinking in real-world scenarios in which cache sharing is highly desirable.

Anonymous said...

so when do you predict that AMD will make another comeback at intel?

Anonymous said...

While is true that channel clearing and a price war are different things, their effect on the share price is the same.

that should have been taken into account in the analysis

Eddie said...

When is AMD going to "comeback"? Well, it is not behind, not even assuming Core2s were in the market; but I understand your question. AMD is limited by production and process right now, and that's were it is focusing: On ramping Fab36 and introducing 65nm products. There are some improvements coming for the next few months, but nothing spectacular for the remainder of the year. Nevertheless, take into account that AMD's current 90nm products are perfectly competitive with Intel's 65nm; thus a successful transition to 65nm will probably crush any Intel uprising, and then some more architectural improvements loom for 2007 and beyond.

Be patient, time is in our side because it allows the advantages in design to express in products.

I am contradicting about the price war: I say that Intel won't incurr in one, and at the same time it is clearing the channel, and I also say it will do a price war selling Woodcrest/Conroe/Merom, extremely expensive products, on the cheap.

Allow me to clarify: The difference between a price war and the channel clearing is simple: The Intel market is moving the same volume, but at smaller prices; that is, they are *not* taking back market share from AMD, a channel clearing, with normal production has only limited effects in non-Intel markets. A price war is different, it has the intention to influence the market for other companies.

Intel basically made a mess of the market in Europe with its channel dumping, and AMD actually did a price war with Opterons, it gave them so cheap that it managed to penetrate the server market so much that all in all jacked up earnings.

Anonymous said...

AMD delayed its 65nm line up chips to Q1 of 2007. is it because of AMD's low yield on new technology? What about Intel's yield on its Merom/Conroe/Woodcrest lineups?

Eddie said...

I don't know why AMD "delayed" 65nm.

I don't remember ever AMD planning to introduce too much of products at 65nm this year, though.

Anyway, let's say that they are behind schedule. So what? that may be simply because it gives more money to mass produce at 90nm and keep fine tuning 65nm than switching to 65nm.

I don't know about why, but I know that current AMD management is not friend of making public relationship stunts, they may feel that there is no need to interrupt any of those excellent 90nm production lines nor to start any of the 65nm yet; or it may be that there is a problem.

But I don't think so, why would they have problems?