<DDR certainly has a lot of pins but isn't SLDram a narrow bus?>
I think SLDRAM is somewhat like RDRAM in that it has a narrow bus and depends on sending data via packet protocol. However this "Anything but RDRAM" consortium has given up on SLDRAM and is now looking into DDR SDRAM. In my opinion, they would have had a better chance working on SLDRAM than DDR SDRAM.
And, of course, this consortium also promises not only DDR SDRAM, but a DDR-2 later on down the road, promising double the peak bandwidth, even though they fail to address how inefficient that bandwidth will be utilized. Next, they'll dump that and go for the "Son of DDR," or "DDR The Sequel: This Time, It's Personal."
What many people are forgetting is that Rambus isn't sitting still. They're not going to rest on Direct RDRAM and say that this is the last DRAM technology you'll ever need. Even as we speak, Rambus is busy coming up with their own follow-on to Direct RDRAM. Hopefully this can both increase the peak bandwidth of RDRAM and apply some of the lessons learned from the implementation of the current RDRAM technology.
And one more thing. I don't know where the president of the "Anything But Rambus" consortium gets the idea that the industry doesn't want a packet-based memory interface. I can tell you for sure that Intel themselves is moving towards narrower, high-speed interfaces in future system designs. (An example would be NGIO.) And these interfaces will be putting requests and data into packets, much like the RDRAM interface.
But then again, I guess the guy considers Intel to be only a "minor" player in this industry ...
<If AMD (cough, cough) gains a foothold in market share, then this would be a concern.>
Execution is key. AMD has a history of poor execution. Now it's apparent that the now-defunct SLDRAM consortium also failed to execute on implementing their DRAM design specs. It's likely that this new DDR SDRAM consortium could trip over its own shoelaces once again.
Tenchusatsu |
