Neeed help with compute errors

I'm baaaack. I cleaned out the dust bunnies and -- for no particular reason -- I reinstalled some older 400MHz memory instead of my current 533 MHz. I'm on day 3 of additional prime95 stress testing and no errors yet.

I think Bernd mentioned in some ancient thread(s) that he has to be careful how he compiles the apps because some of the libraries and load modules are not always compatible with this or that. There are a wide variety of architectures in use. I don't know the particulars.

Is there a software knowledgeable person, someone who knows about compiling windows apps like Einstein, who can shed light on this thought: My computer is SSE capable and I run the SSE version of the app. Could the compilation for SSE use libraries and routines that for some reason simply run a little flakey on my fpu and lead to the message "WARNING: Fixing yLower (-389 -> 0) [HoughMap.c 771]"?

After running this last and final stress test I intend to restart EaH and see if I encounter any more compute errors. If so, I then might run the non-SSE version if anyone thinks that is worthy of pursuit.

Not to be facetious, but how dirty is too dirty? I have no reference standard to say whether it was too dirty or not. I don't personally think it was all that dirty, but I'm not a heatsink.

For some reason my Dell computer won't let me see any temps, nominal or otherwise. However, I've been running prime95 stress test with maximum emphasis on heat and power consumption. I have a power meter attached to my computer's power line. Normally 2 Boinc threads consumes 150 watts but 2 threads of prime95 are currently using 165 watts for 3 days elapsed. Ambient room temp right now is 75F. That ought to push the temp even higher than Boinc does, yet no error is showing up.

Ambient room temp varies from 67-77F and I have a vortex fan located next to the exhaust port of the computer case so as to move the heat away from the local area. I cannot monitor the temps inside the case because my Dell apparently doesn't allow for it. I've tried Speedfan software and it cannot locate any sensors to report from. I have read the SMART data from the disk drive and it claims to be running an average of 37C or 1 degree C lower than the average for disks of this type. Maybe this can be extrapolated to say that the computer case on average is not overly hot, though peak temps are possible.

What I find noteworthy is that I won't just encounter a random here-and-there compute error or access violation. Seems like when I encounter the first error then every task running concurrrently and subsequently will also have access violations. And that may include non-Einstein tasks as well. Reboots sometimes make it better for a while and then it starts again. Monthly windows updates also seem to trigger this problem until after a few more reboots. At least that's the way it seems to me: flakey.

I stated in my last post that I tried speedfan and, in fact, that is what I used to get the SMART data from the disk drive. However, the version I originally downloaded was 4.33 and it didn't seem to support the Dell Dimension 8400 motherboard that I have. I don't think Dell wanted anybody to have any insight into or control of it's product for warranty reasons. I can't find anything regarding fan speeds in the bios setup either and I kept the bios up to date until they stopped updating the bios.

Where is this "dell support" switch that is disabled by default? IIRC only very specific Dell MBs were capable of being supported.