On 10/08/07, karumba <[EMAIL PROTECTED]> wrote: > mike, thanks for clearing that up! > > @tadao amras, our C2D runs perfectly smooth with ~40 players at low CPU > usage. > > Mike Dussault schrieb: > > The packet queue threading was originally added because of a lame piece > > of behavior in Windows where if you send > 6 UDP packets without giving > > up a timeslice, Windows says "no" and forgets about your packets. So > > Windows needed the thread. But Linux doesn't have that problem so it > > didn't need the thread in the first place, plus the thread caused all > > sorts of problems under Linux because we couldn't get Linux to give the > > thread high enough priority to service the packets waiting in the queue > > fast enough. > > > > > > -----Original Message----- > > From: [EMAIL PROTECTED] > > [mailto:[EMAIL PROTECTED] On Behalf Of karumba > > Sent: Thursday, August 09, 2007 3:17 PM > > To: [email protected] > > Subject: Re: [hlds_linux] srcds performance on core2duo > > > > mike, thanks for the info. > > is there any "threading" now for linux? or is there no benefit when > > using dual/quad cores? is the "threaded-code" only temporarily disabled? > > > > Mike Dussault schrieb: > >> net_queue isn't used on Linux anymore. > >> > >> -----Original Message----- > >> From: [EMAIL PROTECTED] > >> [mailto:[EMAIL PROTECTED] On Behalf Of karumba > >> Sent: Thursday, August 09, 2007 2:43 PM > >> To: [email protected] > >> Subject: Re: [hlds_linux] srcds performance on core2duo > >> > >> uh...i just ordered a new core2 duo root :| does someone knows which > >> net_queue* parameters make sense for this CPU? > >> > >> tadao amras schrieb: > >>> Hello, > >>> > >>> I noticed some strange cpu usage while running srcds on intel core2. > >>> We're running some cs:s servers on various cpu platforms. All servers > > > >>> are running the same linux distribution, same packages and same > >>> kernel > >>> with realtime-preempt patch. The gameserver is always cs:s, 20 > >>> players > >>> and tickrate 100. > >>> > >>> Server#1 is an older p4 with 2.8ghz. 20 players need about 80-90% of > >>> the cpu but the tickrate is stable. 10 players need about 30% cpu > >>> usage. > >>> > >>> Server#2 is a quad core xeon X3220, 2.4ghz. 20 players need about 10% > > > >>> (rare peeks go up to 20%) of one core. 10 players are sometimes not > >>> noticeable at all as srcds is going beyond 0.3% cpu usage (per core). > >>> > >>> Server#3 is a (dual core) core2 duo E6600, 2.4ghz. 7 players already > >>> need about 50%. 10 need about 70% and more than 15 need >90% cpu. > >>> > >>> Does anyone know a reason for this behavior? - Or is srcds just not > >>> running reliable on core2 architecture? > >>> > >>> I'll also try to get some cpu/player graphs ready. > >>> > >>> Thanks for reading, > >>> tadaoamras
Hello, problem solved. The mainboard was completely broken. I ran some benchmark tests and got results like i had a pentium 2 (or even older). p4: 431.6 xeon: 612.1 core2: 151.3 Thanks for helping, tadaoamras _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds_linux
