Advanced Micro Devices
Re: AMDS Advantage is Gone

Format for Printing

Format for printing

Request Reprints


By eachus
February 19, 2004

Posts selected for this feature rarely stand alone. They are usually a part of an ongoing thread, and are out of context when presented here. The material should be read in that light. How are these posts selected? Click here to find out and nominate a post yourself!

Whilst Eachus has argued that in the short term it is good news for AMD due to 64 bits now becoming a must have feature and other reasons, he is probably correct but now that Intel are going 64 bit one of the major sources of product differentiation with Intel is now lost. It doesn't matter which way you look at it from any long-term perspective this can only be considered very bad news. Now before some fool tells me why I am wrong consider this. If you are selling something that customers want and your competitors aren't obviously you are in a better position than if your competitor is selling it as well.

If you think what follows sounds like an AMD fanboy posting, you could be right. If so skip on down. But that doesn't make the facts wrong...

Let's look at the situation a year ago, and the situation as it looks to be a year from now. Back in early 2003, AMD was only starting to ship Bartons to finally catch up with Northwood, and Intel was moving the goalposts with higher FSB speeds, and Hyperthreading. A year from now AMD will be all 90 nm Hammers, and in fact some customers will be looking forward to AMD's second generation of 90 nm Hammer offerings, including dual core versions. Intel will still be seriously troubled by power consumption, and anyone with a high-end PGA 775 system will consider a water cooling system to get the noise down--if one doesn't come stock. Intel will be hoping to fix the problem with Tejas, and at the same time push clock speeds even higher.

On the server side, AMD was almost nowhere a year ago. A year from now they should have a large fraction, probably over 50% of Sun's new system sales. (Even if the UltraSPARC systems account for a majority of Sun's system sales on a dollar basis.) Opteron systems will also be sold by HP, IBM, Fujitsu-Siemens, and Cray, in addition to the current cast of second tier players. As far as the average customer is concerned, Intel Prescott and Nocona systems may be fine if you want a low-end box, but the high-end AMD64 (not x86) systems will all be Hammer (and Opteron) based.

As for laptops, last week--and last year--Pentium-M (Banias) was king of the hill. Was the Dothan delay to retrofit AMD64 technology? I doubt it. So do you want a laptop that can last 5 years? 4 years? If so shouldn't you be looking at an Athlon64-M laptop?

Does that sound like something that AMD stockholders should be wringing their hands about?

(Leaving AMD fanboy mode.)

As I keep trying to point out, part of marketing is figuring out how to sell your company's strengths. Intel's current strengths, as a company, are high-volume production at a low cost, and the ability to provide both the CPUs and the chipsets. With Hammer, AMD has not only internalized the Northbridge part of the chipset, but AMD is selling the complete chipset for server chips. This reduces Intel's advantage, but where AMD has chosen to fight is with a very clever scalable design that allows them to sell 64-bit capabilities at a 32-bit price.

That price incidentally is not just dollars; it is power, cooling, board real estate, and processors per rack. To understand this compare a "true" 1U dual-CPU system, and one which fits in a 1U slot but needs some cooling air access towards the back. (I am not saying that all dual Nocona systems will be like this, just giving an example from recent memory.) Both systems are 1U, but in practice, you can only put 43 of the second design into a standard 6-foot rack. You may be able to mix-in other units not as deep--and not as power hungry--without going to a second rack. But it is an issue, and you would really prefer to buy the system that could be cleanly stacked, even if you are only buying a dozen 1U boxes. You might want to put some 6U RAID boxes, and even some networking gear in the remaining slots, and it is much easier to layout if you can stack the CPU boxes.

The same arguments apply to blade servers. If you can put a dozen (or 14, or 16) single CPU blades into a 6U box who cares? Users start caring a lot if they can get dual CPU blades. That may be possible with Pentium-M/Banias, and you run into power limitations with dual Xeon blades. (You may be limited to 8 or so per blade rack.)

So going forward, I see the situation in the server market as: if you need a single CPU server, why are you even looking at AMD? Power and cooling concerns for the closet you want to put it in? Okay, I'll buy that. For dual CPU systems, I still expect AMD to own the high-end, and Intel to own the low-end, with a lot of skirmishing in the middle.

When you start stacking dualies, or switch to four, eight or more CPU systems Opteron will be the clear winner. Will Intel get back into the four-way game with Potomac? Possibly, especially if it supports four threads per CPU to deal with the longer pipelines. (It is strange to talk about Intel getting back into a market they are still in the process of losing. But I think that whatever the reason for AMD's price cuts on 84x Opterons, that will be one effect.) Intel can cut their 4-way Xeon prices, but it won't matter as much. The four-way Xeons have, for the most part, been competing with dual-CPU Opterons, and I don't see that changing anytime soon. The big effect of the 84x price cuts will be to make four-way Opterons more competitive with (more) dual Opteron boxes.

On the desktop? Hmmm. Anyone want a Prescott that can't handle AMD64 instructions? It only consumes more power than a Northwood, and doesn't perform as well on benchmarks. Thought so. When--and I think it is when not if--Intel starts shipping 64-bit enabled Prescotts, then there will be reasons to consider buying a Prescott. Before then, the recent announcement by Intel should have customers preferring to pay a bit more for a Hammer CPU.

Again, the big effect there will not be on Intel's sales, but on AMD's. AMD has been predicting a very rapid ramp of the Hammer family this year. I am starting to suspect that their expectations are too low. When (see above) Intel starts shipping 64-bit Prescotts towards Christmas, the effect will probably be to dry up any remaining demand for Athlon XPs, and to finally give consumers a reason to buy Prescotts instead of Northwoods--and Pentium-Ms.

All in all, as I have been saying, a very good year for AMD, and a good year for Intel as well.

Become a Complete Fool
Join the best community on the web! Becoming a full member of the Fool Community is easy, takes just a minute, and is very inexpensive.