Hi tinkershaw; Re: "Yes, Intel has one, ONE DDR based desktop chip on its road map; Intel has stated repeatedly that it is only there if the price parity between DDR and RDRAM doesn't appear in a timely enough fashion. So this DDR chip is not even a sure thing to be released. But if it is, then this will be DDR's big chance, maybe its only chance with Intel on the desktop. It had better succeed big time or DDR and Intel on the desktop are dead."
1st, Intel only needs one DDR desktop chipset. The current desktop chipset they have for SDRAM is the i815. All Intel needs is one.
But, Intel has arranged that if their DDR desktop chip doesn't make it, there will be some alternatives:
ALi M1741
ALi M1681
ALi M1671
SiS 645
SiS 650
VIA P4X-266
VIA P4M-266
Note that each of the above is a distinct northbridge. In order to keep DDR out of the P-4, in addition to Intel screwing up their 845B, all seven of the above chips would also have to fail. BWAHAHAHAHAHA!!! Yeah right.
Let's review a little history.
Intel didn't think PC133 would be a good idea for the P-III. They thought that PC100 was good enough, and then PC800 (Rambus). But just one chipset, from VIA, provided support for PC133 to the P-III, and that chipset took so much market share from Intel that Intel was forced to follow their tiny competitor into the PC133 market. Eventually Intel came up with i815.
Losing that market share to VIA was stupid, and Intel doesn't intend on doing it again. Now, there are 7 candidates up to take that chipset market share away from Intel. I would bet that Intel is pretty intent on getting the 845B to work.
And if Intel fails to get the 845B out on time? All that will happen is that one or more of the above 7 chips will ruin Intel's market share for P-4 chipsets. And then Intel will eventually show up late to the market with the 845B, just like they did with the i815. Intel being late won't stop the DDR roll out for Intel's P-4. It's all but inevitable now. Why do you think that every single memory maker has unbuffered DDR now? Remember that the only DDR that's good for servers is the registered stuff, the unbuffered is designed for desktops. (See #reply-16008910 for all the links.)
In order for RDRAM to remain on the desktop, Rambus longs have to pray that all eight of these chipsets don't work. And that Nvidia doesn't come up with a P-4 license. And that ServerWorks' DDR chipsets can't be converted to desktop use. And that Intel's Plumas DDR server chipset can't be easily modified for a desktop version. And that neither can IBM's Summit chipset.
Remember, you've got a lot of engineers working at a lot of companies to get DDR on the desktop. The savings of replacing RDRAM memory with DDR is on all those machines is tremendous, and that kind of attraction will make companies put out extreme effort.
Hey, I have no doubt that of these 8 announced (God knows how many are still secret) P-4 DDR chipsets some of them are going to be cancelled before they are done. It's a very competitive industry, and this is what always happens. And when one or two of them are canned, the longs will have a little party. But their only hope is to have all of them go away. Not much of a chance of that.
And if all 8 did fail, then what? No one has started any RDRAM designs for 18 months. That's why there are no RDRAM chipsets announced from anyone other than Intel. That's why Nintendo, PixelFusion and Sun removed RDRAM from their plans. That's why Microsoft went with DDR.
For 18 months, memory designers have known for a certainty that (1) DDR would be available, and (2) that DDR would be cheaper than RDRAM. This is not to say that memory designers knew that RDRAM would stay expensive, just that they knew that DDR and SDRAM would remain considerably (i.e. 25 or 33%) cheaper. Given these facts, every designer went with DDR in order to avoid ending up with a COGS that was saddled with RDRAM.
If designers had known for a fact a year ago, that RDRAM would be as cheap as it is now, then, in the absence of the knowledge that SDRAM would be even cheaper, they likely would have started RDRAM designs. But the fact is that no one knew this. All we knew was that DDR and SDRAM would be cheaper than RDRAM, and that is exactly what has happened. But that was enough to swing the designs over to DDR and SDRAM. (Now do you know why I was willing to bet people that DDR was going to be cheaper than RDRAM?)
Engineers get points for being safe, not for betting the future of a product on a price crash. The memory makers (even Samsung, Toshiba & NEC)) repeatedly told us that they were going to keep DDR and SDRAM cheaper than RDRAM, and that is exactly what they did. You could have read all about it in the trade press, but oh yeah, you losers believe that everything in the trade press is intended to make Rambus look bad, LOL!!!
Answer the question: "Why didn't anyone besides Intel decide to use RDRAM?" In other words, why didn't VIA, AMD, SiS, Serverworks, ALi, IBM, Nvidia, ATI, PMCS, Apple, etc., use RDRAM? Why did Intel decide to use DDR?
-- Carl |
