If by "benches",some of you consider running superPi,wprime,pifast,everest built-in "tests","sisoftsandra" and similar ,then yeah these are about as useful as a chocolate teapot.
You can marvel at the results that mean jack sh*t since they represent no real world application workload,just a hypothetical/theoretical and uarchitecture tuned programs with
zero practical value.
Give me real world a/v compression/decompression,give me archiving of files,give me Maya/Lightwave/3D Studio results,give me gaming
at resolutions end user actually use(
to eliminate GPU bottleneck use SLI/CF if needed!),give me compile times with proper optimizations for each uarchitecture,give me data on how well I/O is done on the platforms tested(CPU usage of SB devices,sata/IDE and USB performance),give me power draw at the wall socket in idle and full load,give me temperatures in idle and load,give me ratio of perf/$ of the platform tested at the end of the review,give me brake down of numbers which would say which system excels in which type of real world workload.This is what matters,not some pathetic list of "benches"(what a stupid word btw..) that never had nor will ever represent a real world performance of one CPU.
Bookmarks