By all accounts, AMD has done everything right against a seemingly insurmountable competitor. It's a true David and Goliath story of how the little processor company that could (little, relatively speaking) is taming a lethal and giant predator. Starting with the introduction of its hybrid 32/64 bit AMD64 architecture -- a technology that I wrongly predicted would lose in the market to Intel's pure 64-bit Itanium -- AMD has been on a roll.
In addition to making the big bet on AMD64, the company's early bet on its HyperTransport memory technology appears to have been a good wager as well. Now called Direct Connect, AMD's approach to memory is to build the controller that gives its CPUs access to memory right into the CPU. This differs from Intel's approach, which has been to separate the two. AMD's design choice is at least partially responsible for the way AMD's new dual core processors are kicking butt [another part of AMD's roll] against those of Intel, even with slower performing memory. Intel is on record with ZDNet as saying that AMD's memory design choice will run out of gas at some point.
But, in Dan Farber's interview of AMD's Fred Weber, the chip company's CTO spins the rumor mill saying that Intel will have a Direct Connect-like technology in its chips by either 2006 or 2007. If that's the case, then it will be yet another Goliath-taming tip of the hat to AMD by its arch nemesis. In an attempt to confirm or deny the rumor, I checked with Intel spokesperson Bill Kircos, who did neither when he said "I'm not saying Intel will or won't go this route someday. History shows we've gone both ways. We integrated the memory controller in our CPUs as far back as 1990 (i386SL w/ Page Mode DRAM + SRAM + FLASH MC) and 1992 (i486SL w/ Fast Page mode 3.3v DRAM, x4, x8, x16). We do it now with our X-scale based products. And, we had it as part of a desktop PC chip project called Timna that we cancelled in the late 90s." In his e-mail response that, in the name of media transparency, can be found here, Kircos also reiterated some of the points he made to me in a previous correspondence about the risks of integrating memory controllers. Those points were incorporated into one of my previous blog posts entitled Intel on AMD's early dual core wins: Not so fast.
Argue as it did (that hybrids were a bad idea) and as it may (that integrating chips with memory controllers is a dead end), Intel may have a tough time arguing with the resulting market share numbers. From AMD's public relations counsel came this barometer of the headway the company is making:
For Q1 2005, Gartner shows AMD continuing to make strong market share gains in several key categories, including:
- AMD's share of the U.S. x86 4-way server market grew to 22.5%, a 35% increase quarter over quarter, and an impressive 325% increase year over year.
- AMD's share of the U.S. x86 market grew to 9.0%, a 17% q-o-q and 34% y-o-y gain
- AMD's share of the worldwide x86 market grew to 5.7%, an increase of 18%
- q-o-1 and 50% y-o-y
- AMD's share of the worldwide total of servers shipped grew to 5.2%, a 6% q-o-q gain, and an impressive 49% y-o-y gain.
For Q1 2005, IDC also showed similar strong progress across all key categories, including:But, while the numbers certainly vindicate AMD's long-held convictions, they still reflect one of AMD's trouble spots: corporate desktops and notebooks. Lenovo, which has taken over IBM's PC business, has announced that its next round of ThinkCentres will be based on Intel's dual core Pentium D chip. But in China, the company has announced that its dual core offerings will be based on AMD's dual-core Athlon 64 chips (the desktop chip) rather than the Pentium Ds it is using domestically in the U.S. I pinged Lenovo for some details on the rationale behind these decisions, but the company so far hasn't gotten back to me.
- AMD's share of the 4-way server market grew to 18.5%, up 35% q-o-q and a staggering 1441% y-o-y
- AMD's share of the 2-way server market grew to 6% , up 18% q-o-q and 76% y-o-y
- AMD's share of the overall server market grew to 5.5%, which represents a 12% q-o-q, and a 72% leap y-o-y
HP has slightly upped its commitment to AMD over the last year. This was brought to my attention by AMD when the company's director of worldwide enterprise and commercial marketing Bruce Shaw responded to my inquiry regarding AMD's lack of traction on the desktop front. Said Shaw:
We expect our success on the server front --more than 50 percent of the Global 100 --will enable us to make inroads into the client side. It's also important to note that we have already had success with the Fortune 500 from a desktop perspective through our partnership with HP. Customers such as H&R Block and Northeast Utilities have standardized on an AMD platform because it offered them the things they demanded in addition to performance -- lower TCO and image stability.
Although the company wasn't able to get back to me in time with the specifics on which machines were in use by the two companies cited by Shaw, a brief perusal of its Web site shows the Athlon 64 turning up in two (5000 series systems) of the company's 15 total business desktop offerings. Two of HP's lowest-end business systems also have 32-bit (non-AMD64) Athlons in them, while none of HP's higher-end more corporately focused and manageable 7000 series systems use the AMD technology. Then there's Dell, which has no AMD offerings. Also, one note. Intel has obviously recognized image stability as an issue because it has an entire program dedicated to it called the Intel Stable Image Platform Program. I discovered it while searching Gateway's site for AMD-based systems and I noticed that at least two of that company's business desktop offerings were a part of the program.
AMD is obviously aware of the challenges it faces. Based on how it has done on the server side (against many odds), you simply can't count the company out. Though completely coincidental, Shaw's 50 percent number also jives with another of AMD's 50 percent citations. Recently, the company launched a newsletter called 50x15 Connections to go with the company's 50x15 program. The program, according to the inaugural issue, is about AMD's effort to provide "50 percent of the world with affordable computing capability and Internet access by the year 2015." Though the program's goal characterizes its target as a person that might not otherwise be able to afford a computer or Internet access (in other words, not really a business), we can assume that if the company is at all successful in attaining 50x15, that that there will be at least some drag effect that results in better business penetration.
All this said, the numbers still speak for themselves. There are few offerings and even fewer businesses demanding AMD64 on the desktop. Perhaps, as Shaw implies, once those companies are more exposed to AMD64 on the server-side, they'll be able to specify AMD64-based desktops with more confidence. We'll see.