> I'm a physicist, and I can tell you that I couldn't care
>less about integer performance, when you get stellar FP
>performance.
You're right, but I bet most single image MP systems are not being sold to crunch numbers, but to run business apps like SAP or Oracle. I have no hard numbers though, so I could be wrong, but my guess is most scientific number crunching is done on (linux) clusters because they so much cheaper than single image MP systems.
>What the hell does that have to do with anything? A CPU
>isn't chosen because its performance/mm² is good!...
Of course not; I just tried to point out that Itaniums lead in (spec)FP is more due to its large caches than to some inherent superiority of IA64 over x86. That doesnt matter when you are in the market for a platform though, I agree.
>Just one more thing: If Itanium is neck and neck with
>Opteron in SpecINT (which it is) and MUCH FASTER than
>Opteron in SpecFP, then it is logical to assume that, in
>any linear combination of FP and INT ops, Itanium is faster
>than Opteron, right?
Probably. But not very often will a such a machine be bought for both integer and FP performance, so I am not sure how relevant it is to declare one overall fastest platform. Also, SPEC is one thing, but I hear lots of echo's of people actually coding on Itanium systems, that see no where near the theoretical performance unless they spend huge ammounts of time optimizing and using PGO. While this may not be a big issue in many situations, there are tons of others (like yours ?) where you just can't afford to let PGO run 24 hours after each minor recompile to get decent performing code. Anyway, this is just hear say, I have no experience on IA64 whatsoever, so YMMV.
>Plus, SPEC also shows Itanium to scale that much better
>than Opteron...
Careful.. you are comparing a 24 GB 8-way Itanium with 6x8=48 MB L3 cache to a 4 GB opteron with 4x1=4 MB L2 cache. I guess its not all that surprising the Itanium beats the opteron.
= The views stated herein are my personal views, and not necessarily the views of my wife. =