Bill's Computer Unstable
Got down to San Diego. _I know, I know, guide to Beijing/Shanghai and San Diego are coming soon._
Anyway, checked out Bill's Athlon 64 3200+ Shuttle SN85G4 is unstable after about 5 minutes. Here's how to diagnose. Essentially, after about 5 minutes of running flat out, the machine hard shuts down. No windows error, the whole machine shuts down. That points to either a thermal problem (e.g., the machine is running too hot and with the Athlon 64 new motherboards have a shut down diode) or it is the power supply that has a circuit breaker shutdown since the supply in the SN85G4 is only 240 watts as I remember and Bill has this machine loaded up with 2GB memory, 250GB hard disk, ATI 9600 All-in-wonder and the AMD Athlon 6400 3200+. Here's how to diagnose it:
# "Radified":http://radified.com/Articles/stability_testing.htm. There are many ways to test the stability of your PC, because there are a variety of applications that will put a heavy load on your CPU, such as SETI@home and RC5. See Radified for more details on testing. I particularly like Prime95 because it verifies results.
# "Silent PC Review":http://www.silentpcreview.com/article172-page1.html. They have an article about AMD Quiet and Cool technology. Installed this so his computer would power down when it isn't working hard, so I hope that at least normally the machine doesn't have a problem. This is the original C0 revision of the Athlon 64, so it idles at 800MHz (instead of 2GHz) at 1.3V instead of the full power 1.5V, this reduces the power requirement to 35 watts. The "thread":http://forums.silentpcreview.com/viewtopic.php?p=64530 on their site is quite amazing in its detail on cooling.
# "Speedfan":http://www.almico.com/speedfan.php, "Motherboard Monitor": and "CPUCool": . Tried these programs to see if it is a thermal problem. They all seem to give different readings, but pretty consistently, the temperatures in a long compute-intensive "memtest86+":http://www.memtest.org/, "prime95":http://www.mersenne.org/freesoft.htm or MPEG2 encoding caused temperaturs up to 70 Celsius. According to "anandtech":http://www.mersenne.org/freesoft.htm, 70C is really the limit as also mentioned in "AMDBoard":http://forums.amd.com/index.php?s=e58f78bb5692297a84ec87ef098690e4&showtopic=20337&st=0entry167429, although "HardOCP":http://www.hardocp.com/article.html?art=NTI0 says it is 85C
# Other thing to try is to swap out the graphics card to reduce voltage requirement. One hint here is that with 1GB memory stick removed, the system seemed much more stable with the full 2GB installed, it would only take about 5 minutes of memtest for it to shutdown.
So I didn't get to fix it, but if you're having the same problem, it might be a good time to get the thermal paste out and relube that Athlon 64 processor. Some folks have reported up to a 8 Celsius temperature drop by doing that. In the mean time, I've asked Phil to swap out the video card to get the power requirements back to something reasonable and see how stable it is. If I were to guess, I'd say it is the power supply, but it could be thermal. I can't find the Shuttle temperature shutoff spec anywhere, but that would be the next step.