Hi Daniel Schuh; Did you notice Tenchusatsu's statement on DDR being likely for desktop next year? This is quite the switch for the guy.
Before reading all this, do take note of his statement after getting back from the Microprocessor Forum:
Oct 14, 2000
The demos were enough to convince me that DDR will indeed make it onto desktop systems, probably in full force by the middle of 2001. #reply-14580362
Compare this to what he's been saying for the last year:
Feb 2000
So to generalize the way Intel sees it, DDR is great for systems with ultra-high capacities of memory, while RDRAM is great for smaller-scale desktop systems and workstations. #reply-12892422
As usual, Dan, you're tossing up assumptions in the air and passing them off as fact. I can guarantee you that the 760's north bridge alone will cost at least $60 more to manufacture than the Tehama north bridge. #reply-12906448
<<DDR only in servers; no DDR in desktops>>
I sure hope this is true and not just anti-DDR FUD. I'm sure DDR will make it into some desktops thanks to AMD and their goal to be the "un-Intel."
On the other hand, the transition of Timna from SDRAM to RDRAM will mark a MAJOR milestone for mainstream acceptance of RDRAM. Although the transition isn't a 100% guarantee (what is a 100% guarantee in life, anyway?), it will definitely be a goal that Intel and Rambus will shoot for. #reply-12970507
Mar 2000
But for the most part, I think Intel will continue to back RDRAM in the future, except in servers where the higher RAS and capacity of DDR is critical. #reply-13045992
As for X-Box, Microsoft said that it will feature 6.4 GB/sec of memory bandwidth, but only 64 MB of memory capacity. I state the obvious: Rambus is the perfect solution for these memory specifications. Why? Because you can implement 64 MB and 6.4 GB/sec using only four RDRAM chips, each chip having its own channel to the system. You can't do that with DDR. #reply-13233312
[In reply to "When Intel finally gets their ship of fools repointed away from the waterfall, they will begin announcing DDR chipsets." ]
You see DDR and PC100/133 SDRAM support as the start of a retreat; I see it as stop-gap measures until RDRAM technology and production matures. #reply-13267551
If DDR makes some serious penetration into desktops and workstations, then Rambus will have a LOT to worry about. According to Intel's outlook, however, this may not be the case. #reply-13267876
Apr 2000
I still think X-Box will support Rambus. It only makes sense. The only argument against Rambus in X-Box is the huge price premiums and limited supply we're experiencing right now. But surely that hasn't slowed down Sony's Playstation 2. I'm sure later on, Microsoft will change their minds and go with Rambus over DDR, just as they changed their minds once before and went with Intel processors over AMD. #reply-13341977
Intel is going with RDRAM for Willamette, and DDR SDRAM for Foster. If you can't tell the difference between a desktop and a server platform, yet continue to post like you do, well then you deserve to be chewed out by the RMBS bulls on this thread. #reply-13343952
No, actually the real test will be Coppermine with Micron's Samurai DDR chipset. This will be the definitive apples-to-apples comparison between RDRAM and DDR SDRAM. #reply-13393601
There are advantages to supporting DDR in servers over RDRAM, too many
for even Intel to ignore. But on the desktop, beyond PC133,
RDRAM is looking to be the better technology over DDR. We shall see. #reply-13399696
Yet I always suspected that implementing DDR traces on the motherboard is a lot harder than the anti-Rambus coalition is making it sound. Maybe that's why DDR is seen as such a great solution for servers, because when servers already require 12-layer motherboards, DDR signal integrity becomes a non-issue. #reply-13400110
As for AMD's DDR chipset, I have no idea when that will be out. Probably Q3 or Q4.
On the Intel side of the fence, Willamette will be here either in Q3 or Q4. My guess is late-Q3/early-Q4. Also, Via's DDR chipset for Pentium III won't be here until Q4, maybe even Q1 of next year. #reply-13405221
The Rambus longs will be glad to know that nothing has changed with Intel's steadfast commitment to RDRAM in their roadmap for the next five years. #reply-13451572
I think it's remarkable that Intel is sticking to its guns on Rambus, at least for the desktop. (Of course, the Intel-bashers on the AMD thread will look at these statements and laugh.) #reply-13458056
However, DDR has several features that make it more desirable for servers, while RDRAM's low pin count makes it more desirable for desktops and workstations. There will be some crossover, for sure, but it's apparent to me that Intel wants RDRAM to succeed on the desktop, and they will do everything it takes to continue past the huge hurdles and headaches of initial RDRAM implementation. #reply-13463171
But DDR on the desktop may be tougher than it seems. I can't be sure about that, though, or else I'd be going gung-ho in RMBS stock. #reply-13485376
Barry, Pat Gelsinger has always been a staunch Rambus supporter. While I believe every one of his comments, I'm also afraid that the promoters of DDR will find some way to make it work on the desktop without resorting to 12-layer boards. #relpy-13490717
I'm sure DDR can be implemented on 6-layer motherboards, which is what Athlon requires anyway. It'll just take a lot of engineering effort. On the other hand, I really don't think dual-channel DDR can easily be implemented with just a 6-layer motherboard. #reply-13492758
And second, there is a chipset being developed that will support two Athlons and two DDR channels, all on one north bridge. Sounds like such a chipset will be part of a low-end server platform. (Incidentally, that north bridge will have more than 1,000 pins. That's not a trivial amount.) #reply-13495318
Gene, one of the reasons why DDR may require additional layers compared to RDRAM is that the electricals of DDR aren't as well-behaved as RDRAM. #reply-13495412
May 2000
Remember that the roadmap he's talking about is desktop chipsets, not server chipsets. DDR is still going to be supported for servers. But at least the good news is that Intel is still gung-ho on RDRAM for the desktop. That ought to pour some cold water on the rumors that Willamette will be paired up with SDRAM through the MRH-S translator component. (Perhaps such rumors were started by those in the anti-Rambus coalition?) #reply-13652209
-- Carl
P.S. There are a few correct statements of fact in the above, and even some opinions that are reasonable. That is why I don't classify Tenchusatsu as a thread moron. He is, however, a bit tied to the official Intel viewpoint, and this colors his opinions on the Rambus issues. |
