To my knowledge all the new Dual Xeons have HyperThreading. I have multiple Dual Xeon servers to prove it.
----- Original Message ----- From: "K. Mike Bradley" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Saturday, July 03, 2004 10:41 AM Subject: RE: [hlds] Windows 2000 or Windows 2003? > Dave I know peeps here will hate me for continuing on with this but this is > a forum for issues like this and I must make a point. > A Dual Xeon (which is a Pentium 3) does not have HT. That's a Pentium 4 > thingy. > > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of David Fencik > Sent: Friday, July 02, 2004 2:33 PM > To: [EMAIL PROTECTED] > Subject: RE: [hlds] Windows 2000 or Windows 2003? > > I'd like to add that I have said that hlds is not multi-threaded, which is, > in fact, false. It is not coded in a way to take advantage of > hyperthreading, however. > > So, perhaps I should have said that hlds is not "hyperthreaded"? > > I haven't made any benchmarks, but run a fairly large gameserver hosting > company. > > All of my systems are dual processor. I have noticed on dual-xeon systems > that hyperthreading will impair the performance of large servers. > >>From the task manager, a dual proc system with HT will show 4 cpus. > Watching the cpu usage of each process, a large server will "bottom out" > at 50% of a cpu (25% at the task manager, during 32 player avalanche, for > example). When this happens, the server lags out. > > The fix is to disable hyperthreading, which will allow hlds to use one full > processor if needed. > > I can imagine that perhaps a smaller server would run better on a single > proc system with hyperthreading enabled, but don't have the means or desire > to test. > > Dave > > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of Steven Hartland > Sent: Friday, July 02, 2004 11:41 AM > To: [EMAIL PROTECTED] > Subject: Re: [hlds] Windows 2000 or Windows 2003? > > You clearly dont understand how HT works. Here's a brief overview: > HT makes a single CPU core look like two CPU's it does this so that the OS > can schedule additional tasks on the second "virtual" CPU and hence make use > of potentially idle execution units in the "physical" CPU. > > The problem comes from at least two potential issues. > 1. The "physical" CPU may not have any idle execution units due to the > design of the code being run and hence a conflict now exists. > > 2. The data and or code needed to satisfy the second "virtual" CPU's process > requirements invalidates in some way the data / code for the primary CPU's > process. This causes additional pipeline stalls reducing NOT increasing the > efficiency of the CPU. > > So yes HT can help but it does not always help due to the potential > conflicts for resources that exist which don't exist in a true SMP system. > > Tomshardware has some nice info on this: > http://www.tomshardware.com/cpu/20040528/index.html > > Steve / K > > ----- Original Message ----- > From: "K. Mike Bradley" > >> I am going to try this one more time. >> >> Again, the Operating system HAS THREADS TOO !!!!!!!!!!!!! >> >> I AM PRETTY SURE THE OS HAS AT LEAST ONE THREAD !!!!!!!! >> Lets pick the csrss.exe (Client server run time sub system) process > (which >> btw services win32 calls ... Something HLDS.exe needs). >> >> HL one main thread >> PLUS ++++++++ >> OS at least one thread (but probably several dozen more) THAT ADDS UP >> TO at the very least ... TWO. >> >> A MP (Multi processor) system would therefore have better performance. >> >> Because two threads run simultaneously. >> This is the point I was making and I did say I don't know about HT but > with >> MP HLDS.exe is better. >> >> If you got bad results with your benchmark testing HLDS.exe on MP, I > would >> look at it again. > > ================================================ _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds
