So, got my new computer going well except the temps on one of the two GTX 260's is INSANELY high. GKrellem reports ambient GPU temp at 66C (been running full load, SETI@Home CUDA, for about an hour) and GPU 0 at 92C with GPU 1 at 54C. Idle temps are lower (around 50C for GPU 1 but around 60C for GPU 0). Using ATSC 840 (airflow is really good, with case open GPU 0 jumps to 100C) with ASUS P6T SE. The two cards are REALLY close together, so besides buying watercooled cards (which I want to do later, when I can afford it) is there anything I can do to lower these temps? It's still below the 105C that nvidia-settings has set as the default alarm level, but WAY higher than what I would like to see.
Check GPUz when you first bootup your computer and have a look at the clock speeds. If they are at full speed then your drivers may not be declocking the card on idle. I had this problem until I reinstalled latest 186.16 drivers with my GTX275. See http://forums.bit-tech.net/showthread.php?t=171589 for more help if the problem continues.
Should have mentioned I'm running Ubuntu so no GPU-Z. But the Nvidia X Server Settings shows it running at 625Mhz, I'll have to start it up next reboot and see what it says then (Adaptive Clocking is enabled).
Best I could do with this crappy camera... I know the cabling could use some work, but what I think the real problem is is that the cards are so damn close together. I let it run for about 7 hours doing SETI (door was closed, room temp was around 32C) and the top card his 103C, the bottom card was only 60C. Once I stopped SETI it dropped down pretty fast, currently at 73C and 52C (about 20 minutes of running idle).
ooo what coolers are they as there not stock. I suspect the hot air isn't being exhausted from the case, hence the high temps.
It's a South Korean brand (I live in Seoul), company named Emtek. I would think they work as the lower card keeps pretty cool... Besides fixing up the cabling, what else could I do to increase the airflow? The case is an ATCS 840 with no modifications made. I'm due a reboot soon (kernel upgrade) so I'll double check all the fan settings while I'm at it. I know it is pumping out a good amount of heat as I could feel the heat coming off the top fans when the temps were really high.
have you still got the stock nvidia coolers? I would throw them back on and have a look at temps again. the stock design just pushes cool air into the heatsink and exhaust it outside the case, where as now your just blowing alot of hot air around near the cards and not really exhausting alot. The nvidia stock cooler might run a little hotter at idle temps compared to aftermarket coolers, but it dose keep the card at stable temps under load. The stock cooler is a fantastic piece of design and i wouldn't be surprised if arctic cooling had something to do with it.
Looks like the top one is being starved of airflow, the fans are right up against the backside of the lower card, hence the difference between the two temperatures on the two cards. You can probably get a stock cooler on eBay or similar, which would probably help the top cards temps. Alternatively, try a different after-market cooler. Edit: Very unlikely, but at work we had a 9800GX2 that hit 120*C under desktop usage, but when we reseated it in the motherboard temps dropped down to 60*C. Just make sure the cards are seated correctly in the PCI-E sockets in the motherboard, if necessary take them right out and put them in again. Another test would be to swap the top and bottom cards, and see if the used-to-be-on-top card that would then be on bottom is still hotter. If it was, the used-to-be-on-top card might be defective.
I vote for that the top one is starving of cool air... it takes in cooled air and reject hot air which then takes back in, and does that continuously which end up not cooling your gpu. I don't think these heatsink was designed for SLI. Also some cable management would not hurt, but I don't expect to me the cure of your problem.
this is a old issue with sli.. I'm a big spot cooler myself- hang a 92 off the end blowing in.. do some wire management in that area for airflow- then stick a fan that fits on the bottom blowing up into the rig you'll have to clean it out for dust every so often.. I actually use a custom made pipe on my gpu using a dryer vent duct to pull air directly from outside the case- you could do something like that too with your setup.. like hang a 80 on the back.. buy the dryer vent duct and cut it to size, run it between an 80 on the bottom and the new one hanging off the cards- you'll have to reshape the duct but it's not too hard to do and works really well- I had really good airflow but my gpu would go over 70c in the summer loaded.. with the duct pulling outside air between a 92 and slim 80, lower 60's (which on the 260 is perfect, can oc away without any issues) but try a simple 2 fan setup first, since your just folding that will probably be enough
Ok, never did any kind of modifications, so not sure I fully understand. Dryer duct I know, I assume 80 means 80mm fan. If I understand correctly, you're saying 80mm over that vent on the bottom (pulling air inside the case), with a dryer duct over it. Other end of the duct pointed at the back of the cards with another 80mm fan blowing the air at the cards? If that's what you mean, I should probably raise the case up off the floor a little more? Since I'm not running SLI (four heads, hence the two cards) I could probably move the bottom card down to the third PCI-E slot... I assume it would still run at 16x as long as one of the three slots is unused (correct me if I'm wrong).
Ok, moved the bottom card to the 3rd slot (AFAIK it should still run at 16x) and temps are better but... The bottom card is now, at best, an inch off the PSU so it's temps are about 3C higher under load... not terrible but still not ideal. Going to run SETI overnight and see what happens. Turns out these cards are factory overclocked (pretty happy with them now, spent as much as stock clocked cards in the US cost). What are safe idle (current 47C) and load (testing to find max now) temps so I know what to look for when I try to push them more.
a nice idle temp for an air cooled 260 would be around 45-50 realistically so 47 is fine. under load i'd say the highest sustained temperature i'd accept would be 80 but my single 280 usually sits between 65-70 so thats fine. under prolonged testing you're likely to see much higher temps than even a decent length gaming session so take that into account.
my 260-216 idles at 45-50, and loads never pass 75. stock nVidia cooler, but with a 92mm fan pointing at the card's intake. SETI or Folding will both stress them pretty heavily. so hitting 80c if your case isn't properly ventilated is normal.
One is at 70C and the other at 71C after almost 7 hours straight (in a closed room too), with room temps at 31C. I think that's not too bad, might still look at doing some kind of spot cooling as well, as I'd like to OC them and see how far they go (already factory OC, but... more...).
Ran for another 12+ hours and had 71C and 74C, so I think this worked. I was looking at some of the PCI slot coolers (Antec Cyclone), anyone know if they really help that much? For 10-15USD even a couple degrees less would be worth it IMO.
Avoid a PCI slot cooler in your current config because then that would just be starved of air.. Get something like an Antec Spot Cool, which can attach to any motherboard screw. See here: Alternatively, use cable ties to hang an ordinary fan (e.g., 80mm) off the end of the top graphics card, to blow air between the two, or hang on the bottom card to blow air into the gap between the PSU and the bottom card.
