Opteron News

Add Your Comments

There’s a rather interesting, mole-fueled analysis of Hammer in the the Inquirer today.

Some key points:

  • AMD can’t get yield on high-speed parts, and doesn’t expect to be able to any time soon.

  • On the other hand, they run rather coolly, though AMD won’t tell you how coolly unless you sign an NDA.
  • No DDR-II controller in the initial Opteron/Athlon64 chips, probably will see a DDR/DDR2 memory controller at .09, though not necessarily the first ones.
  • Athlon64 boards should be cheap.
  • The $46 million AMD paid IBM was not for current Hammer development, but next-generation Hammer development.
  • AMD isn’t just cooperating with IBM, they moved their CPU design team into an IBM facility.

    Will IBM Buy AMD?

    I greatly doubt this. IBM wants to become primarily a computer services company. They don’t need a distraction like competing against Intel.

    However, it is in IBM’s interest to make sure that somebody is competing against Intel. The question in IBM’s mind has to be, “Can AMD continue to do so?”

    And guess who’s put themselves in prime position to find out first hand?

    Due to their technology agreements and new living arrangements, IBM will know more about the true state of affairs at AMD than anybody else outside of AMD. After all, the AMD guys work in their building with them.

    If the horse is no good, they’ll leave it alone just like they left Cyrix alone (and probably pick up a bunch of good design engineers in the process, hell, they won’t even have to clear their desks, much less relocate).

    If the horse is good, though, they’ll back it up.

    Not by buying the whole company, though. That would make Intel too mad. They’ll just back them up one way or another. It might be buying a minority stake in the company, it could be something else, but it will be more than just agreeing to buy AMD processors.

    Whatever it is, rest assured that AMD will broadcast any serious IBM seal of approval.

    So now we have a spy in AMD called IBM. 🙂

    Some Other Opteron Observations

    Given what was said in the Inquirer article, the Opterons are probably not going to be able to overclock a lot, at least not under default conditions. If they could, AMD would be selling them at higher speeds.

    Overclocking would in any case be difficult since the initial processors are at the top of the Opteron voltage range. Voltage mods (if possible) would be necessary (though we don’t have a clue how sensitive and/or responsive SOI circuitry will be to overvolting.

    Recently, rumors have been floating around about a 3400+ Athlon. AMD can probably do it, but if they do this, say in August, it’s an admission by AMD that they can’t solve the yield problems for Athlon64, Take One. You’re supposed to compete against your competitor, not yourself.

    Could this be another Thoroughbred situation for AMD? Maybe, but it doesn’t look like they have a fix yet (at least not for anything significantly above 2GHz).

    If they don’t figure out SOI real soon, forget about Athlon64 Generation One selling big time or even being very available. I think they’ll have to go forward with the release of the product or lose all credibility, but the first generation of “big” Athlon64s will be priced high and end up going to niche groups.

    Athlon64 Jrs. may fare a bit better; if priced right, you could see a mini-boom of dual processor systems.

    AMD can survive .13 Hammers doing little in the sales department, but then everything rests on a quick, painless transition over to .09 micron (without these yield problems) and the availability of Windows x86-64 fairly early in 2004.

    Will they? Let’s hope so, but if towards the end of the year or early next year you suddenly start hearing talk about 90nm Athlons, that will be a sure sign the transition is going badly.

    Ed

  • Leave a Reply

    Your email address will not be published. Required fields are marked *