Joe, <The difference is that with P4 faster on handful of apps, but slower on the vast majority of apps.>
That handful can only grow, especially as Pentium 4 volumes grow and as the P4 platform drives down into the mainstream. I'd say that even having a handful is pretty good considering that Pentium 4 was only released three months ago.
<I think there is a difference between P4 choking on code (that runs fine on 99% of installed based of computers), and moving to the next generation of computing with 64 bit generation of chips. (which incidentally may be a little premature).>
Well, at least we can agree on two things:
1) 64-bit adoption has nothing to do with 32-bit legacy performance (despite the claims to the contrary made by AMD marketing)
2) The mainstream migration to 64-bit won't be necessary for a long, long time
<On the subject of optimization, don't you think it is a little presumptuous for Intel to go to the software community and ask them to recompile their apps just because Intel's P4 team f*cked up?>
The Willamette team did not screw up. They made the right choices given market conditions and the fact that the price/performance ratio could have been even worse on 0.18u. They also recognize that improving performance without specific support from software is a lost cause. The P6 core already performs very well on legacy apps. Why design another P6, except for mobile and value?
<My guess is that neither do you, and neither of us knows how scalable the Hammer family will be. What is your point? I think you should leave the FUD to Mr. Phud (Elmer) and his apprentice (Semiconeng).>
I'll bet that Elmer, Semiconeng, and I know more about what to expect from Intel than you do. But there's no way I can back up that statement, so let's leave it at that.
<We had no idea how scalable .18u P4 would be, and it looks like the clock speed scalability is slightly lower than expected, while the IPC is significantly lower than expected, yielding a processor that is basically a yawn, rather than something that would destroy AMD.>
The problem with Pentium 4 isn't the clock speed scalability, nor is it the IPC, nor is it the dependence on optimized software. I feel all three of these issues are easily solvable. But the bigger question is whether the mainstream really cares about that much performance anymore.
Tenchusatsu |
