On Thursday 06 May 2004 16:44, Malcolm Baldridge wrote: > > I know it's probably heresy for a Linux user to use Intel, but in a value > for money comparison THESE days, a P4 is money well spent. They run > cooler, support SSE/SSE2 (the latter I've used to brilliant effect in an > experimental anti-spam product which does wire-speed connection monitoring > + diversion), and has higher memory + subsystem bandwidth with the FSB800 > dual-channel chipsets (865PE/875). > > When you couple the Intel CPU with the (FREE!) Intel Compiler, you really > have a winner hands-down. Rebuilding the OpenSSL library with ICC gave me > a 35% performance advantage over the best equivalent settings in gcc-3.3 in > the Pure C crypto cores, to name one small example. > > With the very small number of exceptions if you hand-pick the low-end > Thoroughbred cores, run so bloody hot - they are little nuclear power > plants. Yes, the new top-end Prescott P4s top them in heat dissipation, > but those aren't necessary for your application. > Does this imply that AMD no longer has a big advantage in floating point performance over intel? When I bought my last motherboard a few years ago, an Athlon 800Mhz, floating point was supposedly significantly faster on the AMD chips, and since I do a lot of realtime synthesis, going with AMD was a natural choice. Have things changed since then? Larry Troxler
