Replies to post #4701 on Advanced Micro Devices Inc (AMD)

[Dan3]

Re: They would need aHT, too.

That's the key. Opteron's fabulousness isn't primarily due to having 64-bit registers. It's due to the integration of the chipset's crossbar switches and memory controllers with the CPU on one piece of silicon.

Just as the on-die cache has proven to be a huge benefit in improving performance and lowering costs, so does on-die chipset boost performance and lower costs.

And that whole concept is something that AMD has apparently flown into the market under Intel's radar, leaving Intel years behind.

[blauboad]

Intel has to distribute Beta versions, get iterative feedback, make corrections, rework masks, etc. in order to get a reliable product. It would take at least a year in the field before it becomes a product. Once you start distributing something of that magnitude then the rumors will be everywhere.

So, realistically, AMD has at least a year lead on 64bit x86. Good to know. I'd expect a FUD campaign, though, if Hammer takes off. Probably some confusing marketing--"Pentium now with 64+ QuantaBits"--to stall until Yamhill is out. ;)

We have the benefit of the 3DNow! and SSE extensions as guidance. Intel's strategy was to watch AMD with 3DNow! and then come up with something that duplicated (most) of the functionality but was totally incompatible with 3DNow!. Effect was to blow AMD out of the waters because everyone coded for the huge Intel market.

I think there are some important differences here. I doubt very many people bought processors based on whether or not they supported SSE/3DNow! or not. And even among the few who might have--how many of them bothered to check if their software took advantage of the extensions? And would they notice the difference?

64bit is a little different, in that it will probably be much more at the forefront of the marketing campaign, not just an extra little bell/whistle. Alot of Hammers are going to sell just because they are x86-64bit, and I doubt this was ever true of the 3d extensions. Software support will be much more of a concern to the end user, particularly OS support, and that takes a long time. Unless Microsoft is secretly coding for Yamhill, and Yamhill is implemented in Prescott, then I think Yamhill would just be too late.

And if Yamhill won't fly, then Intel will have to implement AMD64. I just can't believe that Intel is unaware of how badly Itanium will fail as a general-purpose CPU. They must have contingency plans.