To: Charles Gryba who wrote (149827 ) 11/27/2001 11:12:25 PM From: wanna_bmw Respond to of 186894 Constantine, Re: "There's really no real current need for 64-bit on the desktop. The reason AMD is doing it is because they hope that will help them get some tier-1 enterprise solutions and because they can't/won't license IA-64 and because it's a good marketing idea. Intel on the other hand will introduce something like x86-64 on the desktop for the sole purpose of antagonizing AMD's solution, making it obsolete and forcing AMD to lose $$$$ and time to emulate Intel's extensions." I agree that there is no current need for 64-bit on the desktop, but I really disagree with the rest of your prediction. For one thing, AMD's chances of getting enterprise design wins are even less than on the desktop, because it requires a much higher level of commitment. AMD can't convince anyone that their first foray into the enterprise market will allow them to establish a new instruction set, even if it does borrow from a currently popular instruction set. AMD's only selling point for x86-64 is that it will offer similar performance to their highly efficient 32-bit core, offer potentially less bloat in the code, and offer 64-bit addressing. Even AMD admits that there will be no performance gain from the instruction set itself. At Microprocessor Forum, Fred Webber only spoke about the potential to shrink the code by as much as 10%. However, in order to even realize those kinds of gains, AMD will have to depend on someone else writing a very good compiler. Right now, Microsoft isn't interested. They haven't announced an x86-64 compatible operating system, they haven't announced x86-64 compatible development kits, nor have they even acknowledged AMD's 64-bit solution (with the exception of typecasts in the header file, but that hardly disproves my point). I only see AMD possibly having some luck with the open source community, which only account for low end front-ends and workstations. The market for AMD's 64-bit instruction set barely exists. Intel, on the other hand, would be idiotic to create another 64-bit x86-like instruction set. For one there is no need technologically, and for another, Intel is in no danger of AMD overturning the development community over night. In AMD's best case scenario it would take years to build up a decent software base, and at any time, Intel could change directions in the future. There is only one potential problem that I foresee Intel facing, and that is the 64GB memory limit with the 36 address lines found in current Intel CPUs. Therefore, their best bet is to increase this number when the time comes. It will of course make more heavy use of the segment registers, but IA-32 processors will be very capable of handling the minimal overhead involved. As for needing to surpass the current 64GB limit, I don't see this happening in the next 5 years on the desktop. The server market would obviously have Itanium as the 64-bit alternative. wbmw