What's the best way to do continuous GPU temp logging?

v0id

Member
May 30, 2003
162
6
81
What's a good program to use to (easily) measure GPU temps and record them to a log file every, say, 5 seconds or so?

I'm having problems playing games - after about 20-30 minutes the system will often freeze completely. Sometimes leading up to that I see some graphical corruption (blue lines blipping across the screen, blue dots, etc) so I think there are some problems with one or both of my graphics cards. I'm running 2x 4850 512MB cards in crossfire. There's not a whole lot of room in the case so they might just be getting too hot. I would like to turn the temp logging software on and then go play Crysis or something, to see what the temps are leading up to a crash. I don't really know what an acceptable temp range is for these cards, so if anybody can help me out there that would be appreciated as well.

Just for fun, here's a pic of the inside of my case. It's not too dusty or anything, so I'm sort of surprised the cards are having problems. If that's actually what it is.

img1137c.jpg
 

v0id

Member
May 30, 2003
162
6
81
Toyota, the bracket on the bottom card appears to be poorly machined and it is not possible to fully seat the card while it is screwed in.

Soulkeeper, the heatsink is a Scythe Mugen II. It's supposed to work pretty nicely (or at least it was when I bought it), but it requires a tricky back plate installation which wasn't much fun.

Thanks everybody. I will try gpu-z or RivaTuner and come back with my results.
 

lavaheadache

Diamond Member
Jan 28, 2005
6,893
14
81
The air flow for that bottom card is severly hampered due to the network card blocking the air flow. I had a similar problem with my trifre setup. My bottom card is an hd4870 with the egg cooler similar to yours. My case is an antec 900 with the psu mounted on the botom, and only leaves about 5mm between the fan and psu. The card was not breathing well at all and crashing out randomly in around 20 minutes or so. For about a week i stuck a 90 mm case fan directly in front of the card resting on the psu and it blew directly across the card. Problem solved. I thought it looked tacky and eventually moved my 4870 X2 into that slot which the fan extended way past the psu defeating the issue.

I was monitoring the temps real time using a second monitor and gpu-z or CCC. I could literally watch the temps rise or drop immediately by manipulating the location of the 90 mm fan.

Hope this helps

lava
 
Last edited:

Zap

Elite Member
Oct 13, 1999
22,377
2
81
I will try gpu-z or RivaTuner and come back with my results.

They both work. You might want to choose based on how they present the data. RivaTuner gives you a line graph while GPU-Z is a text file in 1 second increments. With RivaTuner you can limit what kind of data it keeps while GPU-Z just logs everything.
 

lavaheadache

Diamond Member
Jan 28, 2005
6,893
14
81
back with a little photo shoot of my situation



Cards in question

SSPX0454.jpg



The original overheating problem

SSPX0453.jpg



how it got remedied


SSPX0449.jpg





Your cards are of similar design to mine. I suggest you try swapping them as i did
 

Nathelion

Senior member
Jan 30, 2006
697
1
0
The reason that lower card won't screw in is because the mobo and backplate are bending from the weight of the CPU heatsink. I would recommend taking some of the heatsink weight off the mobo with rubber bands, steel wire, or zip ties.
 

v0id

Member
May 30, 2003
162
6
81
Wow, lots of good suggestions here! Thanks everybody. I will try them when I get home from work.

I sorta figured that the wireless card was contributing to the heat problem. I was considering buying a wireless USB dongle, though I would rather not if I don't have to. I had not considered switching the two cards, although that sounds like a great idea since the other card has a mostly-sealed slot exhaust system and is probably less vulnerable to overheating.

Although you can't see it in the picture, there's also a 120mm fan at the front of the case blowing air over the 2 HDDs. However, this probably blocks a significant portion of the air flow making it back to the GPUs. Would it help to move the HDDs up to the suspended HDD cage you can see on the right side of the picture?
 

v0id

Member
May 30, 2003
162
6
81
OK, I tried RivaTuner and got some unexpected results. I ran Chronicles of Riddick: Assault on Dark Athena at 1680x1050 max details for about 20 minutes before the system froze up, which was about what I expected. Here's what RivaTuner says:

(GPU 0 is the full slot cooler card in the upper slot in the earlier pic, GPU 1 is the card with the... orb? style cooler in the lower slot, immediately above the wireless NIC)

riva.png


As you can see, the game wasn't using the lower card at all. I don't know if this is a game-specific issue or if there's some other problem I'm having - as far as I know the second card is recognized just fine by Windows and the ATI Catalyst drivers. Anyway, the game still crashed. Maybe I have other problems besides insufficient airflow to the second card. Do the temps for the card that was being used (GPU0) look too high to anybody? The card was at 100% utilization and temps were 73C core/82C memory at the time of the crash.

I'll have to think about this. Hmm. Any more help is appreciated.
 

lavaheadache

Diamond Member
Jan 28, 2005
6,893
14
81
seems as though you have other issues going on, otherwise temps look fine. Dark athena scales very nicely using crossfire so that should not be the problem. Try running other games to see what is going on with the second card.
 

lavaheadache

Diamond Member
Jan 28, 2005
6,893
14
81
i would also try running each card individuallytoo to elimante that as a potential. Also, if you are oc'ing your cpu and ram and stuff, try returning them to stock until we can get to the bottom of this freezing issue
 

v0id

Member
May 30, 2003
162
6
81
Thanks guys. I bought Crysis during the Steam holiday sale so as soon as that is done downloading I'll run the timedemo until it crashes and see if the second card is utilized at all. I'll also try using each card individually and make sure they both work by themselves.

I'm running a Phenom II x3 720 BE on an ASUS M4A78T-E mainboard w/ 2x2GB G.Skill DDR3-1333. PSU is a 500w Silverstone. Everything is stock. A while back I did a bunch of testing with Prime95, MemTest, and a Kill-a-Watt meter, and I didn't turn up any problems. The CPU was stable for 24+ hours, no problems in MemTest through 6-7 runs (left on overnight), and I couldn't get the box to draw more than ~300w from the wall, even running futuremark vantage. So I'm pretty sure my problem is graphics subsystem-related. I just have figure out where the problem is exactly.
 

Hauk

Platinum Member
Nov 22, 2001
2,808
0
0
Crysis stresses like no other. As others have said, sufficient airflow across the cards is important. Does your case have a side panel fan? You need air hitting right between the cards, a 120mm fan or an Antec Spot Cool for example. There's sufficent room for airflow, prolly just not enough of it (no need to swap card locations).

Also, what's the rpm on that exaust fan? I assume that's your psu above. While it's got a decent size fan, that case exaust fan should be doing the work to exaust case heat (which you're gonna have with 4850 CF). With just a single 120mm exaust fan to work with, I'd be running at least a 1600rpm fan.

VRMs on 4850's get hot. Do either of those cards have VRM sinks? If not, I bet they're roasting. Several ways to install sinks, would like to see detail pics of the VRM areas. Sinking and an airflow increase.
 
Last edited:

eklock2000

Senior member
Jan 11, 2007
292
0
0
The manual for the M4A78T-E has some detailed setup in BIOS for enabling Xfire, Section 5.2.

I would try removing the NIC card and running tests.

Peace,

EK2K
 
Last edited:

v0id

Member
May 30, 2003
162
6
81
EK2K, Section 5.2 of the M4A78T-E manual appears to describe running the on-board GPU (Radeon 3XXX I think) in Crossfire with a discrete GPU. I am not making use of the integrated GPU and actually have it disabled it in the BIOS. I'll look at the BIOS again when I get home, but I'm pretty sure there's no need for further configuration.

Lots of good ideas, I'll do some more testing when I get home.
 

eklock2000

Senior member
Jan 11, 2007
292
0
0
EK2K, Section 5.2 of the M4A78T-E manual appears to describe running the on-board GPU (Radeon 3XXX I think) in Crossfire with a discrete GPU.

I didn't catch that...weird, who would do that? Seems like it would bottleneck something? Good luck with finding a solution.

EK2K
 

v0id

Member
May 30, 2003
162
6
81
OK, I moved some things around and I think I have it set up a little better now.

img1143z.jpg


I switched the locations of the two cards, and now the card with the open cooler has significantly more room to 'breathe'. I also moved the two HDDs to the upper bay, so there should be more un-obstructed air flow across both cards from the 120mm fan at the bottom front of the case. Unfortunately I can't remove the wireless NIC at the moment as I need it for connectivity, but I think it's going to be much less of a problem now since it's only blocking an unventilated portion of the sealed cooler on the lower card.

I have tested both cards individually with Crysis and they both do seem to work OK. Crossfire seems to work as well (higher FPS), although for some reason RivaTuner is not reporting any GPU activity for the card now in the upper slot. Even so, I can see temperatures for that card in the log rising during gameplay, so it's clear that it's being used. Perhaps it's just some random reporting issue with the card. Temps seem a bit lower on both cards, and I haven't gotten any games to crash yet so we'll see how it goes.