Greg A. Woods wrote:
> You should, indeed "MUST", ignore CPU frequencies when comparing processors.
yes - that's clear - and that is the reason why we tested this things :-)
> You should also look at system implementations, not just CPUs. The
> system (and memory) bus, DRAM architecture, etc. all play important
> factors with CPU-intensive jobs.
yes - 100% agreement on this, although most of our tests show that in
this low-cost segment (V120,E280 are entrylevel-servers) Intel-based
machines are faster. (and our Xeon machines do have something like PCI-X
and DDR-RAM, while those older low-level Suns often only provide
32Bit/33Mhz PCI). What I really wanted to say is that high clock
frequences can often compensate for other (maybe architectural) deficits.
>
> Look at the SPECint_rate2000 numbers for the whole system to get an idea
> of how different systems compare:
>
> http://www.spec.org/cpu2000/results/
this shows me that our test are about right, a 2Ghz XEON is definitly
faster than our 500/700Mhz Sparc's :-)
> You'll also want to verify that the virus scanner was compiled by the
> vendor with the best compiler available for the target system
> (e.g. Intel's compiler for iapX86 systems, Sun's SunPro compiler for
> sparc64 systems, Digital's compiler for Alpha, SGI's compiler for MIPS
> based systems, etc.). GCC still sucks badly when it comes to getting
> every ounce of performance out of a CPU.
"sucks badly" is a nice understatement imho, although I must say we have
NOT verified what compiler our vendor's use. (i.e we are just
using/testing the official binaries).
What is a fact is that we got much better performance with our Xeons for
this particular workload, we have other workloads where the Suns are
much better, but this was not was the original poster asked for.
Stefan
