Re: Coppermine doesn't use compiler tweaks for SpecINT95...
There's a fresh version of the compiler for coppermine. Not that there's anything wrong with that :-). And if the guys who updated that compiler are any good, it's been optimized for coppermine - again, not that there's anything wrong with that.
intel.com
C Compiler Intel C/C++* Compiler Plug-in V4.5
I agree that .18 coppermine, with full speed L2 cache, running on a dual 133 memory bus workstation board, using simd instructions, runs a little faster than a .25 Athlon with 1/2 speed L2 cache, on a single PC100 memory bus motherboard, when it isn't using simd instructions.
But remember, AMDs business plan is to achieve a 30% market share - where Intel, with close to a dozen FABs to support, needs to completely dominate the market.
What's the cost of the two platforms? What's the current availability?
In 60 days, when there are standard (single processor, single memory bus at PC133 or Rambus) motherboards for coppermine available from Intel, there will also be .18 Athlons using VC133 with DDR 200 and 266 coming a few weeks later.
Athlon vs. Coppermine is looking great for AMD.
IMHO AMD's problem, and Intel's triumph, right now is in the K6-X vs Celeron contest - where 5 times as many processors are sold and Intel has a significant performance lead.
Dan |
