AMD’s has had a few bad weeks with DDR: Doubly Delayed Releases, but here comes Santa Clara just in time for Christmas!
(Santa Clara is Intel HQ)
Gamers Depot has an interview with an Intel PR person.
Now how could that possibly be the source of good news for AMD?
Reading PR statements (or interviews with those folks) is like trying to pan for gold. You have to go through a lot of dross. Usually, you come up with nothing. Sometimes, you find a couple nuggets.
There’s a few nuggets in this one:
“We’ve already talked about 2 GHz in Q3’01. . . . The micro-architecture [will go to] (0.13micron) manufacturing process. . . . We will begin this transition to 0.13 in Q4’01.”
What the gentleman doesn’t say or point out is that these timeframes are later than what’s been floating around for these things to occur, especially the latter.
That’s very good news for AMD. It really narrows the time between Intel’s implementation of .13 micron manufacturing on Willy and AMD’s implemenation of .13 (provided AMD’s schedules don’t slip, too). More good news in a moment.
You will see .13 micron processors from Intel before that.
This can only mean the PIII. That in and of itself is no surprise and no news. Combining the two nuggets is big news. It confirms that Intel plans to use the long-used and much-mutated descendant of the Pentium Pro core as its major competitor against AMD for all of 2001.
This is pretty odd. Instead of trying to blow away your opposition by improving your best product, you upgrade the old nag and send it out to battle one last time?
If you were AMD, what would you rather compete against nine months from now? A 2.5Ghz Willamette, or a 1.5Ghz PIII?
It’s now Clawhammer vs. Little Willy
Sports teams love to get mismatches against their opponents. If you’re in football, you want to get your 6’8″ wide receiver going against a 6’2″ defensive end. Same kind of thing with basketball. You can have that in the CPU industry too, except here, the littler you are, the better.
A .13 micron Willamette vs. a .18 micron Palomino would have been that kind of mismatch, and it would have probably driven AMD from the high-end. That’s not going to happen now.
Instead, by the time Little Willy comes around, AMD should be pretty close to releasing its .13 micron Clawhammer. That won’t be a mismatch.
Either Intel thinks a .13 micron PIII is more than enough competition against a Palomino, or it’s not too sure about Willamette.
If you’re looking at Intel, when in doubt, think cheap.
Intel spent a lot of money getting .18 micron PIII fabs in operation. I suspect a changeover to a .13 micron PIII is less of a change (and maybe lets you reuse more equipment) than a .13 micron Willy change.
Wouldn’t be surprised if the .13 micron PIIIs have SSE2 instructions. Maybe Intel thinks that will give them an edge for a while until AMD adopts SSE2 with Clawhammer.
Another reason (and it could well be all of them) comes from this nugget:
(when asked about DDR motherboards for the P4) “We are still evaluating it . . . and at the highest speeds there are still issues. . . . This is a fairly mainstream solution and doesn’t peg the highest system bus speeds that DDR will eventually be capable of.”
Now I could be off-base here, but we do know that a dual-channel Rambus memory system does offer more bandwidth (3.2Gb/sec) than any near-term DDR solution (PC2100 will offer 2.1Gb).
The benchmarks that we have for Big (.18 micron) Willy indicate that it fares best against its DDR Athlon competition in applications that can use as much bandwidth as it can get.
As I’ve pointed out before, a DDR Willy might be a significantly poorer performer than a dual-channel Rambus Willy in bandwidth-hungry operations. Maybe Intel is waiting for faster DDR before officially committing to it.
If that’s the case, sure it’ll let Via make the initial DDR boards. Let them take the rap if or when the combo stinks. Not like Intel exactly loves them for taking half their mobo business away from them.
However, there’s a far more fundamental problem here, it applies just as much to AMD as Intel, and you can read about it here.
Tags: Systems & Components