Numbers lie, the p2 cache (even p3 katmai, not coppermine) is 512kb, but it's sloooooow, so the new cache is extremly, by far, more superior.
Testing hlds on such a sucky pc, migt be a really good idea if they really are able to optimize it to run smooth, BUT, it will simply never happen. First, it should be tested in a LAN-environment, to rule out network-connections as the source of lag, then with max settings, sv_updaterate 100 (default is 60), sv_maxrate 20000 (default is 9999 afaik), on the clients: cl_updaterate 100 (default is 20), cl_cmdrate 100 (default is 30). And if not with a 1000hz kernel, at least with pingboosting, so the hlds-instance never drops below 100fps. And since 32-players is possible, it should imho, be tested with that number. Coming and pretending that a p2-333 can handle a 10 player server with 100fps on a real map (like aztec etc), is just plain far-fetched dreaming at the moment. I do like the fact that someone is actually looking into it though, but anything running below 100 fps (serverside) should not even count as a benchmark, comparing cpu % without fps is pointless. - Sindre >===== Original Message From "Matthew Donnon" <[EMAIL PROTECTED]> ===== >uh Stefan, >current p4 cpus (533 and 800 fsb) have exactly the same amount of L2 cache >as the p2 (512k). >Older (400fsb) units had 1/2 the amount (256k) > >Matt > >----- Original Message ----- >From: "Stefan Huszics" <[EMAIL PROTECTED]> >To: <[EMAIL PROTECTED]> >Sent: Saturday, January 24, 2004 3:15 AM >Subject: Re: [hlds_linux] CPU usage > > >> Marcelo Bezerra wrote: >> >> >If he is profiling to decrease cpu load, a slow machine will more >> >clearly show problems. >> > >> > >> Yes, but on the other hand a P2 is different enough from a P4 that a >> possible error in the code might not show. >> Also, even though the P2 is slow in MHz it has a (comparatively) huge >> cashe. It's been a well know fact for a few years now that HLDS demands >> lot's of cashe to perform good. >> It might even be the constant moving of data from main memory to cashe >> that is the real culpit on the high MHz servers running multiple HLDS >> instances. >> >> -- >> /Stefan >> >> Software never has bugs. It just develops random features. =) >> >> >> >> _______________________________________________ >> To unsubscribe, edit your list preferences, or view the list archives, >please visit: >> http://list.valvesoftware.com/mailman/listinfo/hlds_linux > > >P2 = 512kb L2 cache running at 1/2 core speed & mounted externally to the >CPU core >P3 (katmai) = 512kb L2 cache running at 1/2 core speed & mounted externally >to the CPU core >P3 (coppermine) = 256kb L2 @ full core speed >P3 (tualatin) = 512kb @ full core speed >P4 (willamette sub 2ghz/400fsb) = 256kb L2 @ full core speed >P4 (northwood 533fsb and above) = 512kb L2 @ full core speed >P4 (extreme edition) = 512kb L2 plus 2048 L3 both at full core speed >Xeon (p4 architecture) = 512 L2, options for 1024 or 2048 L3 @ full core >speed >AMD Athlon (classic/thunderbird/palomino/throughbred) 128L1 +256L2 @ full >core speed >AMD Athlon (barton) 128L1 & 512kb L2 @ full core speed > > >_______________________________________________ >To unsubscribe, edit your list preferences, or view the list archives, please visit: >http://list.valvesoftware.com/mailman/listinfo/hlds_linux _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
