Date: Wed, 22 Dec 1999 13:36:22 -0500 From: Tom Embt <tom@embt.com> To: Thierry Herbelot <herbelot@cybercable.fr>, djb@Wit389306.student.utwente.nl Cc: freebsd-current@FreeBSD.ORG Subject: Re: BP6 (Was Re: Success with ATA drivers and UDMA66) Message-ID: <3.0.3.32.19991222133622.014684a0@mail.embt.com> In-Reply-To: <386085B1.4365799F@cybercable.fr> References: <385B2944.6ABE249E@cybercable.fr> <19991218021726.A931@relativity.student.utwente.nl> <385B2944.6ABE249E@cybercable.fr> <3.0.3.32.19991221092604.0145e0c8@mail.embt.com> <385FFD94.9BB9C5E2@cybercable.fr> <19991222003931.C455@relativity.student.utwente.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
>[SNIP] > >> I would like to know how HOT other people's processors get. In the >> stationary situation I have a system core (=3D processor average)= temperature >> of 46 and a case temperature of 50 degrees Celcius/Centigrade. > >What do you use for temp. watching ? (I fetched a little hack which is >called wmtempmon). My temps are somewhat lower : around 35/36 =B0C, as >I've installed "Alpha" coolers, bought from www.3dfx.com. One colleague >at work uses the same sink/fan combo, but with peltier and a monstrous >PSU to get to 572MHz. I've also loaded the latest BIOS from Abit. Perhaps you mean www.3dfxcool.com? Here's my story: I have two PHO fans from that place on my SL36C 366's (540's now). I had to do alot of tinkering to get them stable at that speed, in the end it all comes down to temperature and luck. One of my CPUs will do 550MHz (my goal) with a little TLC, but the other has to be prodded harshly to get it to 540Mhz. In the end I have done the following: - Realize that my heatsinks wouldn't fit over the angled levers on the ZIF sockets. Disassemble ZIF sockets and straighten metal levers in small vise; reassemble ZIF sockets. - Lapp the poorer CPU with successively finer grits of wet/dry sandpaper (it's mirror-like copper color now :) - Lapp the heatsink for this CPU - Turn around the fans on the heatsinks (quieter and works with well with next modification), using 3/16" standoffs made from vinyl hose and covering the side gap with electrical tape to create a ducted "sucking" effect. - Create a complicated cardboard air ducting system inside my case (I kid you not!!) which uses a case80HO (80mm case fan) to suck air in the side vents, blow it in between the CPUs and the AGP card, where the CPU fans suck it back up through the heatsinks, and it gets ducted back out the rear of the case. This dropped CPU temperature by something like 4-5 degrees Celsius IIRC, and is the key to this whole thing working reliably. I think the reason it helps so much is that it bathes the heatsinks in room-temperature air instead of internal case-temperature air, and does not allow "dead air" to recirculate around the heatsinks. On the downside it can reduce circulation for the rest of the components in the case, and should be done with caution. - Determine the "right" amount of heatsink goop to use, get it right and don't ever touch them again. Tip: apply this stuff with a clean single edged razor blade. - Play with CPU speeds and voltages doing ALOT of "make -j 12 buildworld" on a nice warm summer day to seperate the settings that _mostly_ worked from the ones that _really_ worked. Some would run for >30 minutes without a hitch only to sig11 at the last minute. Turns out 2.2V is the sweet spot for both CPUs. In the end I have a very stable dual 540MHz system who's CPUs NEVER exceed 45C and case temp NEVER exceeds 40C (BTW, no A/C here, ambient 65-80F). Was it worth it? It depends what you're time is worth and if you enjoy this sort of thing. It's not for most people, but I certainly had fun. Next project will be a water cooler (faster & quieter, yeah!), and after that I may put a pair of peltiers between the water plates and the CPUs. One must remember that a pair of fast CPUs is not always super useful. My compiling is still pretty much disk/memory bound, and what else uses the real horsepower of two fast Celerons? Games? hah - I've only an original TNT in this box, which is certainly the limiting factor in 3D stuff. One place I've found this system to really shine is running distributed.net clients. In RC5-64 I can get just over 3.0 Mkeys/sec, however right now I'm trying to win the CSC contest (I've personally done about 0.006% of the keyspace so far, out of ~57% total by all distributed.net members). Am I sick and twisted for being so concerned about busting keys? The BP6 is certainly a fun board to play with in FreeBSD, particularly now that the HPT366 is supported (Thanks Soren!), SMP is getting better every day, and there's always new sources to compile. PS - should this be moved to -hardware??? Man, I'm going to regret writing all that, now I've got replies to deal with ;) Tom Embt tom@embt.com To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3.0.3.32.19991222133622.014684a0>