View Single Post
  #150  
Old 01-05-2009, 04:56 AM
cavedude's Avatar
cavedude
The PEQ Dude
 
Join Date: Apr 2003
Location: -
Posts: 1,988
Default

Quote:
Originally Posted by Itchybottom View Post
But Left4Dead lacks Reno and Flabby to harass. . .

I don't mean to stir the pot (if I am); I have to wonder [a question directed at cavedude or FatherNitWit], is the BIOS set to shut down the machine when it reaches a temperature zone on the processor or chipset? Or is it the case it literally get so hot it hard locks?

Discussing the former, these Pentium4-class machines operate at pretty extreme temperatures, and perhaps the BIOS setting needs to be updated. Pentium4 (Northwood and Prescott, not sure about the older Willamette core) are rated all the way up to 75C and 65C operating temperatures. Intel chipsets from that era (865, 875, 850, E7210) are rated between 65c-80c, while SiS chipset (645, 655) are rated for 110c (who needs a heater when you have a SiS P4!) I can vouch for this in testing: I've got a 3GHz Prescott still on my network running fanless with a 0.16cw rated heatsink, and a single 120mm (but 28cfm) exhaust in the power supply, that lives permanently in a 28c room (second bedroom, only one A/C vent unblocked, otherwise no air flow. it's full of folding machines) that hasn't skipped a beat since late 2004. Hard drives are also fault tolerant up to 40-55c. PC2700-PC3200 runs all the way up to 60c in some cases, and not that it matters, but some rambus modules go even higher. MOST ATX consumer-class power supplies are also usually rated at 75c with all of the rails loaded.

Either way, I'm just trying to wrap my head around how this datacenter (or office, or whatever it is) is heating up so much it's shutting you down. To me, this situation is unconceivable unless a human is hitting the emergency shut off switch for the grid or your server hardware is simply failing.

The only other thought I have is those hot-as-hell rack mounted APC and TrippLite units somehow pooping out. I've burned myself on those shiny APC Symmetra LX pieces of... but I've never seen them shut down due to heat. I have however seen CAT5E skin melt all over a TrippLite 3U unit, short out, put on a dazzling array of black/dark gray smoke, accompanied by a high-pitched noise with a few sparks, a loud bang, then proceed blow all of the line conditioners in the building before we could get the power off to that rack and a class C extinguisher in there. But I'm sure that doesn't apply here. RIGHT?
I don't have details, but I do know we are getting kernel core dumps due to excessive heat. I'm hoping the machine will then shuts itself off to prevent damage. But, I don't know if it does or not.

As for how it happens, it's part of a server farm with a malfunctioning (or possibly even non-functioning) AC system... Heatsinks and fans merely dissipate heat, they don't magically make it disappear. That heat has to go somewhere, and it will begin filling the room. That's where the AC comes in. Its job isn't really to keep the room cool, but rather to replace the hot air dissipated by the servers with cool (or at least cooler) air. If the AC is not functioning that room will become hot as hell (I speak from personal experience) and the internal temps of the servers will skyrocket since they will no longer be able to dissipate any heat, since there will be no place to dissipate it to

I have 6 personal machines that when put in the same closed room raises the air temperature at least 5 degrees in the winter time. (Granted, it is a small room) Still, if 10 machines can do that, imagine what a couple hundred can do!