Advanced Micro Devices Inc (AMD)

[wbmw]

Joe, Re: On Intel side, adding second FP unit to P4 would blow Itanium out of the water, eliminating its most notable advantage over competing chips

The difference between today's DP Xeon and IPF SPECfp scores is about 80%. I'm also using the 3M cache version, so that the idea of fitting SPEC within cache is less of an issue.

HP Integrity rx1620-2 (1.6GHz/3MB Itanium 2) - 2692
HP ProLiant ML370 G4 (3.6GHz, Intel Xeon) - 1504

For "blowing out of the water" level of performance from Xeon, you're asking for quite a lot. AMD uses three floating point units, by the way, and it doesn't offer much of a performance advantage for them, at least not in a comparable OEM system.

ProLiant DL145 (AMD Opteron (TM) 250) - 1534

Re: The tradeoffs were not as favorable, but may be in the near future.

So x86 is going to get all these amazing features in the future, but IPF will stand still?

P.S.

Here's a diagram of Opteron's floating point architecture, courtesy of www.chip-architect.com. You can clearly see three floating point units: FPMUL, FPALU, and FPSTORE. Granted, these have specific purposes, but they are designed for the common case, which needs to load and store data as math is performed. Add and multiply operations are also commonly seen together. I can't imagine getting much more performance by adding a fourth unit.