DSCUBED wrote:I have noticed that when running either the GUI, console, or screensaver client, they are all relatively slow. I am running them (not at the same time) on my Dual 1.8GHz PowerMac G5. I also have a Dell laptop running Windows XP with much less CPU, RAM, and Hard drive speed. The funny thing is, that the Dell can run circles around the PowerMac. With other Distributed computing projects like SETi and anything else under BOINC, the PowerMac can toast the Dell.
So what I am wondering is: why is the PPC Mac client so inefficient? DId I set it that way? Can it be fixed.
Does your PPC Mac report "Extra AltiVec boost OK" If so, you've done all that you can do. If not, restart the client and/or add -forceasm to the client so that FAH's optimizations are being used.
When you compare CPU speed (and, to some extent, BOINC speed) you're asking just how fast is the processor clock-rate which is a measure of how fast it is when doing integer arithmetic. FAH (and some BOINC projects) uses almost no integer arithmetic but relies mainly on floating point arithmetic. Both the G5 and your Dell laptop (but not a G4 or the oldest Pentium PCs) have additional circuitry that optimizes simple floating point arithmetic by performing multiple calculations in parallel. The G5 calls it AltiVec and the not-so-old Pentium hardware (P3/P4 and later) calls it SSE. The FAH client is already optimized to use either AltiVec or SSE.
Neither AltiVec nor SSE are new. Since they were developed, a number of additional hardware optimizations have come along. The Intel Core and Core2 technology have further improved SSE speed by reducing the number of clock-steps required to perform individual SSE instructions. No such improvements have been made to the AltiVec hardware.