Joe, <386, 486 and even Pentium were successful even without a serious software support. (I don't count DOS extenders, and Win3.x). All this happened when there actually was a need for 32 bit software.>
There is a big difference. By the time Windows 95 came out, there were barely any 286-based PCs out there. Most computers were already capable of 32-bit processing.
AMD is hoping that software developers will simply move to x86-64 with time. But even AMD can't hope for more than 20% total market share, and that share is destined to decrease as Intel continues building more manufacturing capacity. Then if AMD were even able to transition 100% of their manufacturing to Sledgehammers, x86-64 will only penetrate 20% of new PCs. And the vast majority of existing PCs out there won't ever be capable of x86-64 execution.
The only hope AMD has is to make x86-64 so ubiquitous that Intel will have no choice but to copy them. Judging from how little success AMD had in making 3DNow ubiquitous, I really doubt AMD will have any success with x86-64. And the market x86-64 is initially targeting, high-end database server applications which need 64-bit addressing, will already be taken over by Itanium and IA-64. (Or if not Itanium, Sun UltraSparc III.)
<What do you think is more challenging, starting from zero % market share with an incompatible processor or starting from the top of the 90% plus mainstream market with a compatible processor?>
Think Sledgehammer will have better performance than Northwood or Gallatin (Pentium 4 and Pentium 4 Xeon, respectively)? Remember that Sledgehammer will need specific compiler optimizations for best performance, just like Pentium 4 and P4 Xeon.
Tenchusatsu |
