To: Ian Anderson who wrote (20277 ) 5/13/1999 4:26:00 PM From: Tenchusatsu Respond to of 93625
Why Merced doesn't need RDRAM yet: I already explained on the INTC thread why the 460GX, the chipset supporting Merced, will not need RDRAM. The 460GX is a server chipset, and servers require huge amounts of memory, on the order of gigabytes. You need to have multiple memory channels in order to support this capacity. The additional channels also increase memory bandwidth to the point where even the processors can't consume any more bandwidth. Supporting huge amounts of memory using RDRAM would have been overkill in bandwidth and expensive. Plus, the Merced platform was originally slated for a late 1999 release, before the latest delay. The RDRAM platform would have still been in its infancy in that timeframe. RDRAM will be supported in future server chipsets. <Also I think Merced will be like the P6, in that it will run 32 bit (i.e. current) applications, slower than existing processors, and only show a performance improvement when the apps are recompiled (i.e. never in it's lifetime) Intel might get the software infrastucture ready by the time we get to the Merced III.> You're arguing that Merced is a dead product because of lack of IA-64 software development. Well, how do you get the ball rolling in the first place? A software developer interested in developing high-performance IA-64 applications isn't going to wait for the so-called "Merced III" before starting development. They'll want to start development as soon as possible, which means the release of Merced is critical to their plans. As for Merced running 32-bit programs, well obviously it's not supposed to be a screamer in performance, especially in light of Intel's upcoming Willamette and Foster 32-bit processors. 32-bit compatibility is there to grease the skids between IA-32 and IA-64. Tenchusatsu