Throw Away Your RAM
TechReport reports that the 940-pin “Athlon64” will require registered RAM to operate. That means your current unbuffered RAM won’t work with it.
Searching in Pricewatch got one result, a Corsair 512Mb module for $137, but that was a Pricewatch misprint. Zero results for Pricescan.
So, right now, there’s zero registered DDR400 memory.
Eventually, some is bound to show up, but expect to pay a $10-$15 premium over unbuffered sticks. Also expect slightly lower memory scores from it.
Not too big a deal if you’re building a system from scratch, but an additional unwelcome cost for the average person reading this who is upgrading from a socket A board.
AMDZone reports that Athlon64s will be priced at a “small premium” over XPs.
That sounds good until you look at the current XP prices at those speeds, and just what you may be getting from your money.
Remember, there’s “little” and “big” Athlon64s. The 2800+ mentioned in particular has so far only been seen in roadmaps in the “little” or socket 754 version. Wouldn’t be surprised at all if that came in at $125-150.
Most of the rest will probably be $250 or more.
Perhaps more importantly, since AMD is apparently scared excrement-less at the thought that potential Opteron buyers might buy “big” Athlon64s instead, don’t expect the price of “big” Athlon64s to differ very much from the price of 1XX series Opterons.
Sockets, sockets, sockets
On one spot of a single webpage, we have this statement:
“The 940-pin Athlon 64 will be the variety at launch. The extra one pin is required to work on the 6-layer Opteron motherboards.”
On the same page, posted later, we have this:
“I’ve seen no reason to believe that 939 Athlons won’t work on 940 pins.”
I have been told by those over at that website that the first statement isn’t really correct, that this would only apply to dual-processor boards.
Let us hope the first statement is wrong, but then, that’s the most “official” statement we have on the subject.
And that’s the real problem, not what a website said or didn’t say. Why does AMD leave its customers guessing about basic specifications?
Why Do We Have To Guess?
We know more about the motherboards Intel will put out a year from now than we do about the initial Athlon64 boards, at least “news-you-can-use” material.
Intel will go to socket 775 next year. It will use new technologies like DDR2 and PCI Express. Say what you like about Intel’s socket changes, but at least they are announced well in advance so one can make informed buying decisions.
Intel sets forth clear, simple rules. The Xeons are off in their own little world. You buy a mainstream Intel motherboard, it will run any Celeron or PIVs with the same number of pins for the socket and the same (or lower) FSB. These processors will run in either single or dual-channel operation, so you can upgrade step-by-step if you like.
Compare this to AMD. We’re going to have THREE different sockets for one product, one of them just invented. What will work with what? Who the hell knows?
We’ll have single-processor Opterons. We’ll have “big” Athlon64s. What’s the difference between the two? Who the hell knows?
Will October’s Athlon64 work in next year’s Athlon64 mobo with DDR-II? Who the hell knows? Most likely, it will work, badly (that onboard memory controller, you know).
Will next year’s Athlon64 work in tomorrow’s Athlon64 mobo? Who the hell knows, but the answer is probably the same as the one just above.
In contrast, ask the same questions about October’s Prescott, and we already know the answers. They both happen to be “No,” but “No” is a lot better answer than “Who knows?”
Why We Have To Guess
AMD is a lot more secretive than Intel. After long observation, I think I know why.
How can AMD tell you what they’re going to do three or six or nine months from now when they don’t even know for sure.
This socket 939 is a great example of seat-of-the-pants management. Until about two weeks ago, AMD took the Marie Antionette approach to anybody wanting a dual-channel DDR system was “let them buy Opterons.”
We thought this a bad idea over a year ago. As we said then:
“Let us hope AMD hasn’t fallen into the same sort of trap by cutting Hammer Athlon down in the (probably vain) hope of maximizing Opteron sales.”
Well, I guess some people are slower, but it must have finally dawned on AMD quite recently that forfeiting the advantages of dual DDR to Intel in mainstream products wasn’t too bright, so at the last minute, voila, dual-channel Athlon64s.
Oops, can’t fit dual channel in socket 754. What to do? Take an Opteron and call it “The Opteron That Isn’t An Opteron,” which of course will only work in Opteron motherboards.
Oh, oh, The Opteron That Isn’t An Opteron is going to be stiff competition against The Opteron That Is An Opteron. What do we do?
We must take away one pin and create The Opteron That Really Isn’t An Opteron!
Oh, oh, The Opteron That Really Isn’t An Opteron doesn’t have any Opteron Motherboards That Really Aren’t Opteron Motherboards. I guess we’ll have to sell The Opteron That Isn’t An Opteron with Opteron motherboards until we have The Opteron That Really Isn’t An Opteron platforms ready.”
And just to make everything perfectly clear, we’ll call both of them, along with “The Opteron With Everything Below The Waist Cut Off That Is Really, Really, REALLY Not An Opteron” the same name.
Who needs Monty Python when you can watch this? It will get better, though.
If you’re unfamiliar with Monty Python and Search for the Holy Grail, you might want to skip the next paragraph
In September/October, AMD will trot out the Knights Who Say No. “We are the Knights Who Say No. No, this Athlon64 is not an Opteron. No, they are not identical. No, we’re not going to tell you how they differ, but they’re really, really different.”
Or you could just give the job to the former Iraqi Information minister
With this kind of madness about, how hard is it to believe “Just to show the world we really mean it when we say this is The Opteron That Really Isn’t An Opteron, The Opteron That Isn’t An Opteron will only work with Opteron motherboards but the Opteron That Really Isn’t An Opteron will not?” It may not end up being true, but when you consider what they are doing . . . .
If you were a company that did things like that, you’d want to keep everything secret, too. 🙂
If There’s Anyone Sane In The House
All these cock-a-hoops are apparently an attempt to preserve 1XX Opteron sales (and get more money for them than for functionally equivalent Athlon64 mobos).
Since mobo makers will have no problem offering Opteron That Really Isn’t An Opteron mobos just as loaded as Opteron That Is An Opteron mobo (all they have to do is change the socket), this is a dubious strategy at best.
It would have been better if AMD had just standardized on a single socket like we suggested last April, and use the mobo to enable/disable functions. Maybe they’ll think of that next year.
It would make more sense for AMD to give up on the notion of 1XX Opterons and call any SMP Hammer an Opteron. Better yet, paper launch the damn things in September and go with socket 939 from the start.
Otherwise, they’re in for a PR disaster. They’re just going to look like fools doing and especially trying to explain this. They’re going to have enough problems selling the first generation of these chips as is without that additional anchor.