miguel883

I have had a MSI 6800GT since 3 months ago. My temperatures had always been within the reasonable, about 55C idle and not over 72C under load. I even remember having overclocked it to 420 / 1100 and it never surpassed 78C.

But tonight, after updating the drivers to forceware 66.97 i checked out the idle temperature and it was around 64C. That scared the shit out of me, and I proceeded to test it with RTHDRIBL, Doom3 and Half-Life 2. My fears were confirmed and I saw it reach 95C (!) after a few minutes running RTHDRIBL. Half-Life 2 gave me no less than 92C. Everything at stock speed.

All kinds of self-induced paranoia have begun to torture me. Did i hold the fan blades too rudely while cleaning the card one month ago? I don't think that's the answer, I remember checking temperatures and even overclocking after that. Does the driver version have anything to do with it (that IS paranoia)? No, they don't, apparently. The room temperature is in the mid-high 20's and obviously it hasn't varied much in one month, though my CPU temperatures are now some 4 or 5 degrees higher than they were a few weeks ago.

I really don't know what to do. Truth is I haven't experimented any artifacts or hangups while playing, which is strange with such temperatures. And i've touched the card's heatsink and while it was surely hot, I had no trouble keeping my fingers on it. It occured to me that the card probes could be measuring inaccurately, but why would that happen all of a sudden? What in the world could make a GPU that worked normally rise 10C idle and 20C under load? I have considered getting a NV Silencer 5 and some Arctic Silver, but before that I thought I could use some advice.

People who use these on their 6800GT seem to be within 60's under load, mainly depending on air flow inside the case, i guess. I have two intake 80mm fans and one outtake 120mm fan. So, theoretically, if I purchase the new heatsink and thermal paste and still see temperatures around 80's (they'll probably drop at least a bit) with no artifacts, could I assume the temperature probes are fucked up?

Any thoughts, theories, suggestions about this matter could noticeably ease my worried mind

Thanks a lot

mkk

Try stepping back to the previous driver and see if the values return to the level where they used to be. If so then perhaps in the newer driver they changed the way that the probe was read, wich could have been done to show off values that are closer to reality. These cards draw a lot of power and seem to be able to handle heat quite well. If you see no artifacting at all then you can safely assume that there is no real issue at hand.

If you switched a cooler that exausts heat directly from the card at the back of the system like the Arctic Cooling nV Silencer series then at least it would help the rest of the system, and probably the card itself as well.

No_Style

I would suggest going back to the latest nVidia posted ForceWare which is 66.93.

__________________

Geminiwave

yeah I'd agree with nostyle..

though it is worth noting that both ATI and nVidia have said that their GPU\VPUs are rated to go up to 100C without a problem. thats right 100C! not 100F. So I wouldnt worry about it. Most of the time those probes dont report things properly.

Also check the version history of the newer drivers...find out if they have any notes about new thermal readings.

__________________

miguel883

Yes, I know this card can endure (at least in theory) temperatures that high, but I'm afraid it could die much before it was expected to. What really worries me is the fact that I haven't been getting such temperatures from the beginning, which means the fan could have deteriorated. And so the card could be running near enough the frying point to decrease its lifespan, and far enough not to notice any anomalies. I have a carpet in the room and it's well known that doesn't exactly help dust issues. The fan didn't seem too dirty, though. Is there a way to measure the GPU fan speed?

Now I'm back to forceware 66.31 and it's about the same. So I guess i'll go for the Silencer and the Arctic Silver for some reassurance. At least that way I'll know I have possibly the best air cooling solution for my card, and this means I'll be able to forget those crazy temperature readings.

I have thought of a way of determining whether the temperatures are right. I could forget about them for a while and try to CAREFULLY overclock the card to the speed it was able to reach with any artifacts, then see what happens. I should be able to detect serious image artifacts before frying it, is that right?

Thanks for your help

Geminiwave

yeah you should be able to run the thing with artifacts even without frying it. It just isnt suggested(nor is overclocking REALLY..)

honestly heat isnt going to fry your card nearly as fast as overclocking is. Overclocking is the main killer even if you keep the temps super low.

That being said, most people happily overclock their video cards for years and have no problem.

Some of us arent so lucky.

now that being said: there is a way of detecting fan speed. You got a motherboard monitor thingy already right? And it checks CPU speeds. Well disconnect the fan PLUG from the geforce card and plug that plug into the motherboard where one of the sensors will pick it up(usually at either the ones that say "PSU" "CPU" or "Case1" anyway stick it in there and you'll get readings. I've done that for my radeon. It works really well.

__________________

noko

I wonder if the newer drivers slowed the fan speed down more? Since the default HSF has a variable speed fan. I use Expert tool and force a set % speed for 2d and then for 3d operations. I can dramatically change tempertures with ExpertTool by changing fan speed. I usually set 3d at 100% fan speed and 2d at 75%. Right now in 2d I have 32c on my GPU . Well it helps when blowing 2c air over my video card with my A/C cooled case. I am now usually under 50c when playing 3d games.

__________________
Q6600 at 3ghz, 8gb DDRII, GTX 260 (216), Vista 64

miguel883

The fact that certain driver versions could change the temperature reading sounds a bit strange to me. I mean, in theory, the thermal probe should be measuring temperatures and then displaying them through the GPU properties panel or whatever, but without any software interference. Why would a new driver corrupt the information? Anyway I have already gone back to Forceware 66.31 and the nVIDIA control panel shows the same temperatures. Maybe it's something you can't fix just by reinstalling drivers. I was about to format the HD and install a Windows XP with built-in SP2 so it directly detects my 160gb in a single partition, so I guess it would be a great way of checking this

I have just downloaded ExperTool (thanks, noko!) and well, surprise! It says in desktop conditions the GPU temperature is about 53C and ambient about 42C, while the nVIDIA control panel still is in the low 60's for the GPU and the same, 42, for ambient. Do you think ambient temperature for the GPU can differ noticeably from, for example, case temperature? It seems logical. I get 33C on the motherboard with PcProbe and 24C for the case with its built-in thermal sensor.

I don't know, what's more trustworthy, ExperTool or the control panel?

I apologize for any potential signs of shameful ignorance or incorrect use of English

noko

I would say ExpertTool in your case. Maybe the driver you loaded was a driver for a particular version of card with different temperture sensor. Anyways try adjusting your fan speed manually, I am able to shave off around 4c easily with it.

__________________
Q6600 at 3ghz, 8gb DDRII, GTX 260 (216), Vista 64

mike2h

nvidias drivers have been having some issues with detecting correct fan speeds & temps with various manafacturers.
this seems to be mostly due to various manafacturers varying from the ref design. also seems like nvidia trying to correct the problem is possibly creating more issues. jmo.
anyway, noko has it right. there is also a very good tool that is called speedfan. believe there is a link to newest ver in the news section.
if you are still having issues after the clean install & are seriously considering an aftermafket cooler i would take a look at zalman new solution. think it has much better buid quality over the silencer.

__________________

MiDKnYtE

you can use rivatuner to monitor temps also.

http://www.guru3d.com/rivatuner/

1- Start Rivatuner and go to the settings tab.
2- Select "Run at Windows startup".
3- Choose "via Startup folder" under the pull-down menu.
4- Select "Send to tray on close".
5- Click OK.

Now, Rivatuner always starts when you start Windows.


6- Go to the "Power user" tab.
7- Expand \Monitoring.
8- Click on the left side of "BackgroundMonitoring".
9- Double click on the right side of "BackgroundMonitoring" on the darker area.
10- Enter 1.
11- Expand \Core Temperature.
12- Click on the left side of "ShowInIcon".
13- Double click on the right side of ShowInIcon.
14- Enter 1.
15- Click OK.