I recently purchased an Asus DK gts 250 gpu and the temps are insane. It idles at 60C. I ran a FurMark gpu stress test and after 174 seconds it got up to 99C and showed no signs of stabilizing. I stopped the test at that point so I don't know how high it would top at..not sure it would top :\
I'm wondering if it's even worth trying to reseat the cooler or if it needs to go straight to Asus.
FurMark score was 3463 @ 1280x1024 on XP Pro.
3DMark06 score was 13268. (SM 2.0 6218, 3.0 6501, CPU 2738)
Are scores at least ok? If it's performing ok I guess I'll give the heatsink a go. Maybe it wasn't installed properly. I've never seen temps act like that. Yes, the fan is spinning. Yes, it sounds fine.
40% is supposed to be the "fix" for a hot running 4870 so 36% should not be yielding this high of a result. I set the fan speed to 45% and ran the test. It reached 90C in 70seconds and was still steadily climbing. I increased the fan speed to 55% and it got to 90C in 53seconds (obviously not fully cooled down from previous test).
Btw another odd thing.. GPU-Z is showing my core clock at 740mhz while Asus specs say 738mhz and the memory clock is showing at 1100mhz while Asus stated clock is 2200mhz..
I know that. It still should top out somewhere in the 70C. The fact that it can reach 99C and still climb at a steady temp is well beyond normal. Even if I up my fan to 65% it never drops below 60C and this is also not normal.
I ordered the same card from Newegg. We'll see how it does when it gets here.
About your clock settings... Showing 740 when ASUS states 738 is fine. The memory is DDR, so it's 2200MHz effective, or 2x1100. They are both set correctly.
No, it should not top out in FurMark at 70 something. FurMark will push air-cooled GPUs up to 85C and above rather easily. That's what it does. That's what it's designed to do. It pushes them to max temp in order to allow ppl to test stability for over-clocked settings.
Contact the makers of FurMark if you don't believe me.
When I did a google search that's what people were reporting their top temps at. Nobody is idling at 60 and climbing to 100c. People recommending leaving that stress test up for a good 30minutes. My card will fry or my computer will turn off if I try to leave it on that long. Temps never max out. The temps aren't ok. :\
Ok, I see what you're saying now. It's not flattening out at all. Not my best night around these parts... Sorry about that.
I'll definitely agree that it should begin to flatten out at some point above 80C at least, so there is surely something awry with that card. I wouldn't bother opening it up and checking it out to see if it's seated correctly, though, as I believe that would void the manufacturer warranty. ASUS is very strict when it comes to their warranty terms, whereas companies like XFX and eVGA are far more liberal.
Also have to agree with Dekasav and suggest monitoring during an actual game or two.
The idle temp is normal if you keep your GPU fab speed on auto. I have the 8800 GTS (same chip as yours) and it idles at 60 C and goes to 78 - 80 playing Crysis. However, I use nTune to manually set the fan to 100% prior to starting Crysis and at full load it won't go above 66 C.
I just did a test where I manually set the fan and it made no difference..
Anyway, I found out some disturbing things about the Asus 512nb gts 250 that Newegg is selling. The GTS 250 is almost identical to the 9800gtx+. This I knew. The few differences are the LENGTH (gts 250 is a lot shorter) and that it uses less power so it only has one pci-e connector while the 9800gtx+ has two. Well the GTS 250 I got is the same length as the 9800gtx+ and has two connectors. Apparently Newegg somehow got sent a bunch of Asus 9800GTX+ marked as the GTS 250. Anyway I bought a different one and sure enough it's shorter and has only one connector. Also all the reviews you see online show one connector and the shorter version. I dunno what's up with the Newegg 512mb cards. The 1gb Newegg version is the real GTS 250. I'd go elsewhere for those wanting the 512mb version.
Also I still think my last card had cooling issues. If you look online to various reviews, like this one just for example http://techgage.com/article/asus_en9800gtx_512mb_dark_k..., you'll see that temps just shouldn't get this hot. Their card maxes out at 65C using 3dMark and mine went up to 81C. My idle is almost their max. It's not a problem with case cooling as my HDD idles at 32C and my CPU at about 39C and never hits 50C even after 7+ hours of Prime95.
Performance-wise things seem to be fine. After fixing my CPU clock back up to stock, results are as follows:
3DMark06 score was 13778. (SM 2.0 6499, 3.0 6619, CPU 2893)
Old one probably had a bad fit or thermal compound seal between the GPU and heatsink.
Glad the new one's working fine. If it still seems to heat up a little too much under stress, you can try using eVGA's Precision utility to lock-in an increased fan speed so it auto-applies whenever your system reboots.
As RazberyBandid said, bad thermal compound seal or dying vent... Happend to me a while back...: My GT 6600 was overheating, normal idle temps was 45 degres Celsium, max temp was never above 65. Then my cooler started failing... after some time idle/max temps started increasing... soon after my GPU went SUPERNOVA XD
Smell of burned electronic -.-
So check your GPU cooling, disasemble cooler from GPU. Reaplly thermal compound and put your cooler back on.
Also first make sure if your vent is spping at all.