Effect of heat on system??

[ksv]

I left home for a week at a conference and came back to find that someone had closed the cabinet door on my new system despite everyone being told to never do that. The system works on BOINC constantly, including GPUGRID. As it turned out the Internet connection was inoperable for most of the time I was gone; things got reconnected the day before I returned. I expect little computational work happened during most of the week.

Things were probably chugging along for a good 24-30 hours before I got home.

I found the GPU at 87 deg. C. Don't know what the cpu was at. The cabinet atmosphere was quite hot (only passive ventilation). I know the nvidia module runs in adaptive mode.

The door no longer exists.

The system's basics are a Supermicro board, Xeon E3-1240v2 cpu (Noctua cooler, little that that might help under the conditions), Crucial M4 SSD, Seagate HDD, GTX 560 Ti graphics card, 16 Gb ValueRAM, Nexus Prominent R case (very open).

Anyone have thoughts on the amount damage this may have caused, and what clues might exist concerning that? I run Linux, BTW...

 

[Patrick]

I hate to say it but I have had mixed results with this. One test manufacturers use is to heat components and run them hard for weeks to speed up failures. This was not that long but it is going to depend at lot on the actual components. Does everything still work well? Usually you will see damaged components in DC scenarios fairly quickly IIRC from Folding@Home.

 

[ksv]

I really see nothing unusual. The compute errors I can track down occurred on other computers as well. All seems normal at this point.

I'm waiting to see whether there are failures in apps elsewhere. Can't see anything in system logs, though I don't know that there *would* be anything there anyway.

 

[mobilenvidia]

nVidia GPUs start to throttle at around 102degC, so you were running hot but within limits.
If there is no pixelation then chances are you are fine.

CPU's start getting unhappy around 95degC.
Most modern CPU's throttle at a certain point, still best to avoid it.

 

[cactus]

I would guess no significant harm has been done to your system. The GPU at 87C is not bad and your system would have shutdown if the CPU hit its thermal threshold. Capacitors would be the first thing to go in high heat(look for any bulging tops). The worst I have seen was a Cisco 6509-E(6 48p GbE cards and a sup) in a small network closet hit 115F+ intake temperature. The switch was fine once we went and opened the door, that was two summers ago.

 

[ksv]

Well, I'll be checking my capacitors, but feel more comfortable based on everyone's input.

I appreciate all the thoughts! Thanks!

 

[ari2asem]

very old topic .....but i have new addiction, it is called gpugrid (my first addiction was all about storage)....any info about your gpugrid-build/setup?