Hi all; There's been a lot of speculation among the less realistic technology watchers that Intel may be playing coy with PC1066 because they're trying to pressure Samsung into pricing the parts more reasonably. The idea is that this way, Intel will force Samsung to sell PC1066 at a more reasonable bin split.
This is a "beyond silly idea". Intel would have to be run by insane maniacs to make a decision like that. But then again, they did sign the Rambus contract, BWAHAHAHAHAHAHA!!! LOL!!!
The real problem with RDRAM is that no one is designing with it. If Intel intends on bringing it mainstream, they should be making it as clear as possible that they intend on supporting it, and the PC1066 speed extension. Playing coy is exactly the opposite of this. In fact, if Intel doesn't support PC1066, several industry observers have noted that this would be a definite signal that Intel is out of the Rambus business. (See #reply-17125184 ) So by playing coy, Intel is sending a signal that RDRAM is dead, dead, dead. Does that make sense? Of course not. If Intel wants RDRAM to be mainstream and affordable, they need to get as many memory makers making it as possible. To get that, they need other companies to design with RDRAM, and to get that, they have to convince design engineers (like me) to use the stuff. But we're a tough sell, and we're certainly not going to use RDRAM if Intel is making noises like it's going to drop RDRAM. In fact, Intel already told us, in private, that RDRAM is toast. So we don't need further convincing. That was quite enough, even for those of us who didn't see it coming back in late 1999.
If your memory technology is obviously in good health, well then you can play games with the memory makers. But RDRAM is not of good health. Pretending to not support PC1066 would be like a very elderly person visiting a mortuary and taking a nap on a marble slab. They're not going to do it because they don't want any accidents to happen.
Intel has a problem with Rambus. They bought a 5 year license to use Rambus stuff. That means that they have to be completely out of Rambus by the end of those 5 years, or they have to renegotiate. It's fairly clear that Rambus screwed them good on the negotiations. Intel went from paying essentially nothing to paying $10 million per quarter. But Rambus has brought nothing but disaster to Intel. This is a well known and often remarked upon fact.
So with this history, is Intel going to want to renegotiate 5 years from now? BWAHAHAHAHAHA!!! LOL!!! BWAHAHAHAHAHA!!! Of course not. But products have fairly long life cycles in the semiconductor industry. So in order to avoid another pasting from Rambus, Intel has to get completely out of RDRAM by 5 years from this past November.
Intel can extricate themselves from Rambus, but it is going to take a good bit of those 5 years. To get out of the Rambus bind, they have to replace all their RDRAM product with something else as quickly as possible. Since they are bringing out dual channel DDR chipsets for workstations this year, there really isn't much reason to extend the RDRAM chipsets in that niche. In notebooks, desktop and servers they're already good to go.
So I would be a bit surprised if Intel did push memory out to PC1066. It's not automatic that Intel will avoid PC1066, they can still meet what I believe is their 5 year goal, but dropping PC1066 is what I'd expect.
-- Carl |
