Date: Mon, 28 Aug 2000 22:00:47 -0700 From: Kent Stewart <kstewart@urx.com> To: gjohnson@gs.verio.net Cc: Mike Meyer <mwm@mired.org>, Marc van Woerkom <marc.vanwoerkom@science-factory.com>, questions@FreeBSD.ORG Subject: Re: FreeBSD is being extremely slow.. Message-ID: <39AB437F.E1B686EB@urx.com> References: <14762.54705.346152.495600@guru.mired.org> <39AAE9EC.DFD5E4E@urx.com> <14763.2428.985901.162062@guru.mired.org> <39AB2575.8E16B4C1@gs.verio.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Tony Johnson wrote: > > 1.5x gain. 10% gain. Post your computers speed using one of the unix > benchmark programs, overclock your box , then do the same benchmark > program. I bet you will find that you did not get a 1.5X gain! Look at > all the factors of computer speed that you have no control over. > > 1. Cabling standards > 2. IDE and Scsi standards > 3. Disk rotational speed > 4. System/Video Ram speed > 5. pci bus saturation. Yes STB has driver patches for thier 128 pci > video cards because they hog up your pci bus. Overclocking your cpu > would only make this problem worse because the pci bus bottle neck is > now worse. If we were worrying about all of that, we would still be running at an 66MHz FSB. That could be the wrong end and a 133Mhz FSB would impact things less. My video cards with one exception are all agp. I don't have any STB products in any of my systems. The only PCI cards that mattered were the NIC's. I have a couple of Adaptec 2940's but setiathome doesn't test disk through put. I had only one benchmark I considered important and that was running setiathome. A Celeron 300a @450 ran seti wu's in an average of ~33,600 seconds/wu for 312 wu's. A Celeron 433 in the same system is running an average of ~45,900 seconds/wu for 228 wu's. I can supply you with all 628 times if you want. Actually, I have the results for all 628 wu's. Running Celeron 433a and a 66Mhz FSB using PC-66 memory averaged 54311 seconds/wu. Replacing the PC-66 memory with PC-100 dropped the average to what I gave. These times seem to be comparable to those produced by running a couple of Monte Carlo programs that I am interested in using. I didn't have a setup to endlessly get new data and capture the results like I could with setiathome. The times for two P-II 400's running Win2000 were 30989 seconds/wu and 29620 seconds/wu for ~150-200 wu's. The speeds changed when version 2 was introduced and this is the average for version 2. In addition, I have only tabulated the first 530 results, which included version 1 of setiathome. All of the data was captured, I just didn't take the time to add the cup_times to my spreadsheet. I'm running FreeBSD on the 400's but I haven't tabulated the data. The FreeBSD times are all visibly longer than the W2K times. Enough so that I have dropped 1000 places in the order of results produced. A P-III 450 running W2K Server is averaging 28,450 seconds/wu. The total wu's processed by all 5 cpu's is just over 4016. Data for the last ~2000 wu's were captured. The worst is an old P200 overdrive that is averaging around 180-190K seconds/wu. > > Not only do you make your computer more unstable by overclocking it, but > if you used one of the unix benchmark programs , I'd bet that you did > not see 10% speed increase in your computer. You are also damaging your > equipment. If you wish to throw away $$$ , could I give you my mailing > address :-) I agree on one point and that is when you use a speed that produces a wierd PCI speed. For the rest, I would think that you've been reading too many books. You may be right. That wouldn't surprise me too. I have always been told that Intel only manufactures one speed. They just test them at different speeds. If you run faster than the design speed, you can be in trouble. In my real world, Cray basically overclocked his computer's for years and then cooled them so they were reliable. Our small Cray required 230KVA of cooling. Kent > > > > Mike Meyer wrote: > > > > Kent Stewart writes: > > > Mike Meyer wrote: > > > > Marc van Woerkom writes: > > > > > > I refuse to support overclocking, please fix your system and > > > > > > then repost if you continue to have problems. > > > > > A wise decision. > > > > > I was once tempted to overclock a P166 to 180 or somethig MHz. > > > > > There were several weird errors due to overclocking that did never > > > > > show up under W95 but only under FreeBSD at that time. > > > > What's really wierd is that overclockers seldom go to even as much as > > > > 10% more CPU. For anything but very long-running cpu-bound tasks > > > > that's not enough to be noticeable! > > > That isn't true. You go from a FSB of 66 to 100 and clock for clock > > > that is a 1.5x gain. > > > > I've never heard of anyone doing that one before(*). The ones I see > > are more like the one here (166 -> 180), which is less than 9%. > > > > However, what I normally see are CPU speeds, which might be a > > different ball of wax. If you go from 66 to 100 FSB with 4x cpu > > multiplier and a 366MHz CPU, then the *CPU* clocked at 400MHz, which > > is right at 10%. > > > > I'm not into this stuff. The damn things are flaky enough without > > going out of my way to make them worse. > > > > <mike > > > > *) The exceptions are the guys doing liquid-cooled systems, and > > getting 2 or 3x. On the other hand, they admit they're doing it for > > hack value, and are spending more on the system than it would have > > cost to buy a system running at the resulting speed. > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org > > with "unsubscribe freebsd-questions" in the body of the message -- Kent Stewart Richland, WA mailto:kbstew99@hotmail.com http://kstewart.urx.com/kstewart/index.html FreeBSD News http://daily.daemonnews.org/ To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?39AB437F.E1B686EB>