POST OF THE DAY
Advanced Micro Devices
Re: Intel 10x Faster in 3 Years

Format for Printing

Format for printing

Request Reprints

Reuse/Reprint

By Roleplayer
December 9, 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!

Grunchy wrote: And, Sander's worst nightmare just came true: Intel has a multicore hammer clone running for both desktops and laptops, on the new 90nm process, with 65nm coming on hard. I think Banc of America has this one nailed. I'm looking seriously at shorting AMD.

Here is why Intel is blowing smoke:

1) AMD is ahead of Intel in demonstrating and shipping dual cores. Dual cores were planned for Hammers all along. I think it highly unlikely that Intel will pass AMD on the basis of dual core technology in less than about an 18-month time frame. They may catch AMD late next year, but catching isn't passing.

2) AMD seems to have no heat and power issues at 90 nm. Intel has not yet demonstrated good heat and power at 90 nm. Dual cores are going to require more power and better cooling than single cores. I can buy that Otellini is talking specifically about Itanium or such, but not mentioning that; Intel is famous for FUD. But imagine 10 Prescotts in a single box (okay, more realistically, imagine 8 Prescotts with the extra 2X in performance coming from other efficiencies). That's something that's going into a computer room, not a desktop, not a laptop, and probably not a workstation. I can imagine a dual core being a desktop CPU by 2008. More than two cores? Not without huge breakthroughs in materials or software requirements.

3) In order for Intel to pass AMD with multicores, they will have to learn how to make multicore processors where each core outperforms its AMD counterpart. Intel is behind AMD here as well.

4) Intel is therefore stating their intent to equal in all and surpass AMD in some of the three critical areas of multicore design, power characteristics, and overall performance, and relatively soon. I smell the smoke, and it's not from the power supply trying to keep up with four Prescotts. IMO B of A is scent impaired.

5) Let us suppose that the 90 nm footprint advantage goes to dual cores. That leaves AMD's dual core volume at about where they are now. They won't be making fewer CPUs. Intel is talking about 70% dual cores by the end of 2006. Well, by the end of 2006 Fab36 should be producing product for revenue, possibly a lot of product. Any Intel hopes to outfab AMD with multicore products won't come to fruition until 2007 at the soonest, and IMO later.

Now I have all my "sell" antennae quivering, in large part because I didn't sell last time. But it is just as capital an error to sell at $24 instead of $48 as it is to not sell at $24 when AMD is dropping to $14.75, B of A's AMD target, assuming that it was reasonable to forecast $48. My short-term sales are targeting $30, and short term is six months, and I expect it to be about 1/4 of my controlled shares. But lets paraphrase Otellini with things that we know today, and see how bad it really sounds.

"We are behind AMD in almost every meaningful characteristic of CPUs right now. The only advantage we have is many times the manufacturing capacity. In order to leverage our advantage we are going to try and force CPUs to become as large as possible. The approach we have chosen is to put as many cores as possible into one CPU. We believe that it will take us three (or is it four?) years to push this technology to the point where AMD will have trouble making enough such CPUs to match us. This should enable us to solidly retain our position and profit margins at the highest end of server CPUs. If it translates to other sectors, so much the better."

So what are my warning signs about this strategy actually showing the possibility of success? First of all I'm looking closely for indications that a supercomputer built with an 8 core Itanium will significantly outperform four dual core Opterons with comparable power usage. Needless to say, I have no such indications today (but if you know of some, I'd be glad to follow any links). I'm looking for four core Itaniums outperforming two dual core Opterons as well.

Of much greater concern to me is the mainstream battlefield. Has Intel solved its power problems? Not that I see. Has Intel retaken the performance crown? Not that I see. Has Intel improved its bin splits? Not that I see.

So instead of looking at this long range forecast as spelling doom for AMD, I have to ask: is this Intel's way of saying, "We're not catching AMD for at least three years, and this is our way of spinning things so that we can keep some shreds of pride"?


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.