AMD Items

Add Your Comments

Another Contradictory Roadmap

There’s an article with an updated AMD roadmap here.

The information presented there rather differs from what Digitimes recently reported.

Basically, there’s two new and two missing items:

  • A 2.6GHz FX-55 is supposed to come out in a few months. These will almost certain be 130nm chips, and frankly, they’re like the Northwood 3400+ in that both chips are being pushed to the end of the process limits.
  • AMD will adjust the PR on socket 939 A64s upward 100 points from the PR rating of socket 754 chips running at the same speed. This is a conservative adjustment which shouldn’t give anybody heartburn (I had figured on 200).
  • What is missing from this roadmap are two processors: the 1.8 and 2GHz models mentioned in the Digitimes article. Since these are the processors you’d want to buy, this is a rather important omission.

    One problem with roadmaps is that different versions from different times can surface at the same time. Another problem is that journalists and/or the sources don’t get the details exactly right. This becomes a lot more likely when the product mix is as confusing as AMD’s.

    For instance, as of now, an Athlon 64 is socket 754, except when it’s socket 939. The same is true for a Newcastle. If the latest revealed roadmap is correct, a socket 939 2.4GHz processor with 512K cache will have a higher PR rating than a 2.4GHz socket 754 processor with double the cache.

    So it is quite conceivable that somebody speaking to Digitimes might have said that Newcastles would be available at speeds from 1.8GHz-2.4GHz, but somewhere lost in the confusion was the detail that half of them are socket 754s and half are socket 939s.

    Or it could be something else.

    There’s a very simple approach to take to all this. Come May, AMD will either have a mainstream socket 939 or not. If they do, fine. If they don’t, don’t buy a Hammer until they do.

    People Are Talking…

    People Are Talking

    There’s two AMD related webcasts out there. One is another talk for a Stanford EE class, this time by AMD’s chief Hammer architect, Kevin McGrath.

    It’s basically an overview of the Hammer architecture, with a few comments on improvements to be made within the next year or so. That was previously reported here.

    One brief comment he makes during the speech is that the single/dual-channel memory architecture split is ohly there for branding purposes, not because dual-channel is so much more involved than single channel.

    Another AMD related speech was the one given by Chief Financial Officer Robert Rivet for the Morgan Stanley Semiconductor Conference 2004
    . This one is rather more interesting because it gives some additional insight into what is going on in the heads of the AMD execs.

    This presentation is rather odd in that the presentation is missing. Only the Q&A is available.

    The comments were rather odd, too. When asked about the level of success of desktop Hammers up to now, Rivet announced that “We’ve teased the market.” He pointed to just cinematic and gaming developers/, and said the rest of the market was “curious” about it.

    He then went into a curious mix of BS creative reasons and real reasons why AMD wasn’t ramping up too quickly.

    The creative reason was that AMD had no need to ramp up at a “crazy man’s pace,” since AMD was not in a position like it was with the K6 ready to “run out of gas.” Funny, but I seem to remember the Athlon rampup being fairly leisurely, and the K6 actually getting a die shrink to .18 micron to make it last a little longer. 🙂

    That made about as much sense as saying that it was more important to sell first to the “right places” than to sell a lot of processors.

    The real reasons intermingled with the others was 1) “We need an operating system.” (he said it, I didn’t, penguinphiles) and 2) needing 90nm production to get “the right size and die.”

    Once that happens, then they’ll get crazy and ramp up at a raving lunatic’s pace during the fourth quarter, with Hammers going to over 50% of total CPU unit production at that time.

    To be fair, he also said a number of things that would seem to indicate that if anything, AMD might be able to do a bit better than September/October to release 90nm processors, and that MS rather than Dresden might end up being the bottleneck.

    The subject then swung to ASPs, and that was a very curious discussion, too. Mr. Rivet said that AMD had a “philosophy change”, and no longer wanted to find itself making 22% of the world’s x86 processors, but only getting 10% of the revenue from the sale of x86 processors.

    Apparently, the new philosophy says that once you sell companies expensive server CPUs, they’ll also be willing to buy expensive desktop CPUs from you, too.

    Mr. Rivet thought to be “relevant,” AMD needed something more like 25% of x86 revenues. This is a rather stunning figure, since based on some other comments he made, it would seem to demand that AMD would have to increase its market share about 75% while increasing their average price about 50% above today’s levels.

    Nice work if you can get it.

    However, when one questioner tried to establish a rather modest milestone for this audacious ambition (an average ASP of $100), Mr. Rivet said, “It will take a while,” with “a while” apparently being the time it would take for Fortune 500 companies to begin seriously jumping whole hog on the AMD bandwagon.

    Shaking My Head

    The good news is that there is one force in the world capable of making all these wonderful things happen to AMD. The bad news is that force is Intel. Not just Intel, but an Intel screwing up.

    Can AMD break down the Fortune 500 walls with attitude and an Opteron? That’s not too likely to any big degree without Intel screwing up.

    Well, yeah, Intel is screwing up at the moment, and probably will continue to do so for some time to come.

    Then again, they screwed up over the PIII, too. For a couple years, Intel didn’t seem capable of doing anything right. What did it get AMD? About 5% more market share, for a while.

    OK, this screwup will probably be worse; Intel’s actually trying to sell these furnaces, which could lead to a few million bad customer experiences.

    Will it be any better this time waving an Opteron around? It can’t hurt, probably will help a bit. But it is a shaky marketing plan indeed that only realistically can work in the short- and medium-term when your competitors is screwing up simply because there’s a distinct possibility they’ll stop.

    For us, well, we don’t seem to register in this brave new world. From this speech, it becomes quite clear that when AMD says “customer-centric,” they mean “corporate customer-centric.” I don’t quite see how 4-way Opteron performance is going to make me or you want to pay rather more for a Hammer.

    The sum of the message seems to be, “We’re looking for a better class of customer.” Well, I’m sure that would be nice to have, but don’t you think you ought to get them first before tossing your current customers aside? We may be a bunch of scuttlebums by corporate standards, but shouldn’t you try to keep at least those of us with a little more in our wallets?

    And what happens to AMD (and for that matter Intel) when everybody (including corporate America) turns into a bunch of scuttlebums and decides that a $40 processor is good enough?

    It seems like AMD wants to move upscale just when the rest of the world is going in the opposite direction.

  • Leave a Reply

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