l2aider1 Posted May 3, 2008 Share Posted May 3, 2008 Monk, Thanks for taking the time to respond. I appreciate the insight into your reasoning. And you may be right. Was just curious to know if there were some compelling evidence to prove one having a significant advantage over the other, or if it were more of a matter of personal preference. I'm sure theres probably not more then 5-10% difference between comparable configurations. Off topic have you guys messed around with Dual Harpertowns at all, as we are potentially looking at a Dual Harpertown configuration running on the 5400 Chipset. Would be curious to see how they fare in regards to game server hosting. Link to comment Share on other sites More sharing options...
Monk Posted May 4, 2008 Share Posted May 4, 2008 First thing, Game servers are sensitive to clocks inside your motherboard..This also has an effect on other syscall performance as well, since the lower the faster (less expensive) * TSC ACPI-fast HPET i8254 * Xeon Woodcrest: 580 nsec 1429 nsec 1290 nsec 4180 nsec * Opteron 2200: 201 nsec 1110 nsec 1760 nsec 4420 nsec * The code I used is useless for everyone else unless you have a stratum 0 clock locally attached via PPS iface (serial port GPS), so you have a baseline device to use to see how shitty your various clocks are inside. HPET is only used on Windows 2008, the TSC skews on multiprocessor servers, i8254 is very slow to read and ACPI uses multiple IO port reads since it lives off somewhere behind a bridge.. I'll post some more instruction benchmarks as soon as I finish writing the code and running it on various intel and AMD boards.. Link to comment Share on other sites More sharing options...
adamnp Posted May 7, 2008 Share Posted May 7, 2008 nothing:: 22 cycles locked add: 22 cycles cpuid: 262 cycles Figured I'd add this for the hell of it, New Q9300 machine Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.