AMD Takes The High (Price) Road

Add Your Comments

Taking The High Road

We have a roadmap for AMD CPUs for 2004. This includes all the models (including Newcastles and 90nm chips) that are supposed to come out this year.

What is most noteworthy for the average person reading this is that the slowest Newcastle will run at 2.2GHz and have a 3400+ rating.

Why is that important? It is important because it almost guarantees that Newcastles are going to be expensive for a very long time to come.

Why is that?

  • AMD has pegged its Athlon 64 official prices to Intel’s official prices, and has kept to it.
  • After February, Intel doesn’t plan to cut prices on its mainstream processors until they come out with the 3.8GHz Prescott sometime in the third quarter (best guess around Labor Day). (When the 3.4GHz non-EE chips come out, they will fill the second rung of the Intel pricing totem pole, at $417. The top $637 will be filled by Intel’s 3.6GHz processor when it comes out, so the release of the 3.6GHz will not trigger a round of price cuts.

    Since we’ve known for quite some time what Intel’s pricing roadmap is planned to be, and since Intel’s pricing follow a pretty set pattern, we can easily project AMD’s socket 939 prices for most of the year.

    CPU Speed Cache 130/90nm
    April 1 May 15 Sept. 1
    Nov. 1
    Athlon 4000+ 2.6GHz

    512K

    90nm

    —-

    —-

    —-

    $637

    Athlon 3700+ 2.4GHz

    512K

    90nm

    —-

    —-

    —-

    $417/$278

    Athlon 3700+ 2.4GHz

    512K

    130nm

    —-

    $637

    $637/$417

    $417/$278

    Athlon 3400+ 2.2GHz

    512K

    130nm

    $417

    $417

    $278

    $218

    All dates are approximations. May 15 means the date the Prescott 3.6 comes out; September 1 really means the date the Prescott 3.8 come out, and November
    1 signifies the date an Intel 4.0GHz processor comes out.

    There’s some doubt about the price of the 3700+ later on because it doesn’t exactly match Intel’s offerings, so I’ve provided the two Intel price points.

    Want to see what socket 754 looks like?

    CPU Speed Cache 130/90nm
    April 1 May 15 Sept. 1
    Nov. 1
    Athlon 3700+ 2.4GHz 1Mb 130nm
    —- $637 $637/$417
    $417/$278
    Athlon 3400+ 2.4GHz 512Kb 130nm
    $417 $417 $278
    $218
    Athlon 3400+ 2.2GHz 1Mb 130nm
    $417 $417 $278
    $218
    Athlon 3200+ 2GHz 1Mb 130nm
    $278 $278 $218
    $178
    Athlon 3000+ 2GHz 512Kb 130nm
    $218 $218 $178
    $163

    Now let’s see what Intel has in stock for the year.

    Prescott Prices…

    Ed

    Prescott Pricing

    In contrast, let’s look at the pricing of the three lower-end Prescotts (the 2.8 and 3.0 are likely to become the most popular O/Cing processors).

    CPU Speed Cache 130/90nm
    April 1 May 15 Sept. 1
    Nov. 1
    Prescott 3.2 3.2GHz

    1Mb

    90nm

    $278

    $278

    $218

    $178

    Prescott 3.0 3.0GHz

    1Mb

    90nm

    $218

    $218

    $178

    $163

    Prescott 2.8 2.8GHz

    1Mb

    90nm

    $178

    $178

    $163

    <$163

    Let’s compare the CPU prices of a 3GHz Prescott (which I suspect will end up being the O/Cer chip of choice), a socket 939 Athlon 64 3400+, and a socket 754 Athlon 3000+ (which will be the cheapest CPUs in their prospective lines):

    CPU Speed Cache 130/90nm
    April 1 May 15 Sept. 1
    Nov. 1
    Prescott 3.0 3.0GHz

    1Mb

    90nm

    $218

    $218

    $178

    $163

    Athlon 3400+ 2.2GHz

    512K

    130nm

    $417

    $417

    $278

    $218

    Athlon 3000+ 2.0GHz

    512K

    130nm

    $218

    $218

    $178

    $163

    Granted, a socket T Prescott may entail some extra costs a Hammer won’t. The most advanced models will require DDR-II and a PCI Express video card, but the lesser models apparently can be configured using DDR and AGP. It should also be noted that should the 250MHz FSB come into play, most people will have to run their current memory at speeds slower than the FSB, or buy new RAM.

    We’ll definitely need to do the math come April and May, and find out just how capable a socket T Prescott is.

    However, for an overclocker, it sure looks like Intel will have a $200 price advantage on CPUs over any dual channel Hammer system until the fall, and a socket T Prescott ought to be at least price-competitive with soon-to-be-obsoleted socket 754 performance systems.

    How Likely Is This To Happen?

    There are two ways these prices will drop more than indicated.

    Intel Launches A Price War It’s possible Intel might decide to accelerate their pricing cycles to help socket T Prescott sales if they sag, but all that would probably mean is that the prices you see for September and November would come a couple months earlier. Besides, Intel doesn’t seem exactly loaded for bear at the moment. Unless they can ramp up Prescott by considerably more than expected, they’ll probably leave well enough alone.

    AMD’s Parity Pricing Collapses These prices tend to support Smith Barney’s contention that AMD isn’t going to be making a whole lot of these things any time soon. The prices are just too high to expect big sales from them.

    There’s two elements to AMD’s high prices. There is price parity with Intel, and there is high speed grades.

    The prices on the socket 939 chips didn’t surprise me at all (given the price parity policy). What did surprise me is that AMD did not come out with a socket 939 Newcastle running at 2 GHz. Then the socket T/socket 939 matchup would have looked like this:

    CPU Speed Cache 130/90nm
    April 1 May 15 Sept. 1
    Nov. 1
    Prescott 3.0 3.0GHz

    1Mb

    90nm

    $218

    $218

    $178

    $163

    Athlon 3200+ 2GHz 512Kb 130nm
    $278 $278 $218
    $178

    rather more competitive than this . . . .

    CPU Speed Cache 130/90nm
    April 1 May 15 Sept. 1
    Nov. 1
    Prescott 3.0 3.0GHz

    1Mb

    90nm

    $218

    $218

    $178

    $163

    Athlon 3400+ 2.2GHz

    512K

    130nm

    $417

    $417

    $278

    $218

    Wouldn’t you agree?

    So long as AMD keeps price parity with Intel and keeps the speed grades up, you can’t expect prices to become “reasonable” any time soon, especially for socket 939. The question then becomes, “How long can they keep it up?”

    The AMD answer, of course, will be “From now on,” but realistically, they can’t keep up such a top heavy pricing and speed grade structure indefinitely. Not even Intel is as top-heavy as this.

    So the question is not “Will it collapse” but “When will it collapse?” through speed grade reductions, price grade reductions or both.

    The answer is “Sometime after AMD supply exceeds customer demand and unsold processors start piling up” but of course that isn’t really an answer to the question.

    My best guess at the moment is later rather than sooner, and that it is more likely we’ll see lower speed-rated processors before we’ll see lower prices. “Later” for lower prices probably wouldn’t come until late summer at earliest, and early 2005 at latest.

    Killing The Messenger…

    Ed

    Killing The Messenger

    This is the roadmap I’ve been afraid I was going to see since last summer.

    Hate to say we told you so, but we told you so.

    I already know what the likely fanboy, “Write Ed an email or (much safer) call him an idiot or on Intel’s payroll in some forums” response is going to be. Denial.

    It will take time for the truth to slowly sink in for many, and it won’t be pretty when it does.

    I really don’t take any joy in saying these things, but sometimes the truth hurts. I’ve thought and think AMD is doing both itself and yourselves a lot of harm by following this policy, simply because AMD risks alienating their current user base without any guarantee that they’ll find others who will pay these prices. They have the right to do it, but I don’t think it’s going to work.

    I can see why AMD is trying to do what it is doing. It’s responding to a lot of the financial traumas it suffered over the past few years, and it thinks that x86-64 is the wonder drug that will cure their financial disease and solve all their problems.

    The problem is all this is a short-term strategy, an attempt to grab a few quick bucks fast, and forget the aftermath and long-term consequences.

    There’s a number of reasons why AMD is doing this, and some of them would be understandable. If you can’t make a lot of chips for whatever reason, price the few you can high and hope you grab some quick cheap bucks to keep you afloat and look good for the financial markets for at least a few quarters.

    Eventually, though, they’re going to have to ramp up Hammer production (perhaps waiting for 90nm), but by that time, Intel may have well said, “me too, and more” when it comes to x86-64, which would take away AMD’s claim to fame.

    In any event, entry-level prices will have to drop, even presuming AMD can keep up a Intel price-parity policy, and given the events of the last few years, that’s an awfully BIG assumption to make.

    Intel: The Bargain Solution?

    Intel is likely to find itself this spring with at least the opportunity to become the “bargain” company for overclockers. Next May, AMD overclockers who won’t buy Intel are more likely than not to face the unpalatable choice of paying more than for Intel to get about the same performance, or paying the same to get somewhat less.

    That’s what the situation will be if socket T Prescotts can do 4-4.2GHz overclocked without too much fuss. If this can be done with lower-end Grantsdale boards which continue to use DDR and perhaps AGP, that could become the O/Cer’s best (or more accurately, least bad) choice three-four months from now.

    Of course, even the price of the lower-end solution will hardly be a “bargain.” It won’t be that Intel becomes a bargain; it will be that AMD will stop being one. It will become the lesser-of-the-two-evils. (Well, if Intel is evil for high prices, doesn’t AMD become even more evil for the same and higher prices?)

    How Far Goes Your Fandom?

    The AMD strategy is clear. To them, enthusiasts have a choice. If they’re really enthusiastic, they can pay about ten times what they were paying on average for XPs to get an FX. If they’re not quite that enthusiastic, they can pay about six times what they were paying for an XP to get a socket 939 Newcastle. Finally, for the cheap low-lifes who’ve been paying $50-60-$90 for a CPU, AMD will say thank you for that loyalty by coralling you cheap sheep into the “bargain” sector and only charging them two to three times the price for a third-rate solution.

    OK, sheep, are you going to follow this shepherd, or are you going to wait until he sobers up?

    If you don’t like being considered sheep, well, I’m not the shepherd; I’m just noticing that the shepherd is getting out the sheering knives.

    Does your socket A look a lot better now than it did before you started reading the article?

    My Personal Reaction

    Before this info came to light, I had figured socket 939 would be the next big thing, and I figured that would be my next system that I would pay for. After this information, I’m more inclined to buy a socket T, and much less inclined to get a socket 939 with my own money. The socket T looks like the (more than slightly, less than somewhat) better choice; AMD can wait until they come to their senses.

    Mind you, I was leaning towards the socket 939 even assuming a $278 price tag for the CPU; I wasn’t exactly holding out for $50. $278 was a stretch, but $400 snaps the rubber band. I suspect the rubber band for most readers doesn’t even stretch as far as $278.

    Socket 754? I’m actually writing this using a socket 754 system. I’ll talk more about this in the days ahead, but for right now, it’s certainly OK, but it’s hardly awesome.

    Your Personal Reaction?

    You may well not be as inclined as I to go Intel (and I’m not suggesting this is a must-have at all), but I think most of you will feel as disinclined as I to pay $400 for an AMD CPU. Some may feel inclined to look at socket 754, but that feeling is likely to pass if it does turn out that an AMD system costs almost the same and does less.

    Whether you believe me or not right now, so long as your wallet decides when to buy, you’ll probably be on the sidelines for quite a while, maybe until 2005.

    How do you feel about that?

    I’m not taking a poll, but I’d like to know how you feel about the whole situation and how it will affect what you’ll do the next six months.

    This is pretty high-grade speculation, I’m just taking what AMD’s policy has been up to now with Athlon 64s and just projecting it out. Until they change the policy, these projected prices are what the policy says they’re going to be.

    So when you write me, don’t tell me, “No, it’s not going to happen” or “I don’t think so” unless you have better reason than “I don’t want to believe that.” That’s about as convincing under the circumstances as convincing me Santa Claus exists by telling me you believe in him.

    Assume that it is going to happen, at least for the next six months, maybe a year. Just to cover the “AMD changes its mind” option, please tell me at what price point you would spring for a new Athlon 64 system, and whether you’d want socket 939 or settle for socket 754.

    You know where to find me.

    Ed

  • Leave a Reply

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