I have a 9800pro 128mb with a VGA Silencer on it. I installed the catalyst 5.6 drivers as soon as ATI released them a couple weeks ago. Shortly thereafter I had several crashes in 3d apps / games and noticed that this was from my cpu + vidcard reaching nearly 70C!
These two components are frequently around the same load temp in my system because of the layout of my motherboard (MSI K8N Neo Plat). The mem slots are above the cpu socket so the cpu socket is placed down between then and the AGP slot. The distance between agp slot and cpu socket is probably no more than 4 or 5 cm and the fins on the slk-948u on my CPU come within 1cm or so of the rear bracket of the vga silencer. Basically all the heat radiating off the back of my vid card is rising up directly under the cpu heatsink and likewise the fan on the cpu heatsink is blowing air out the top and bottom of the fins, the bottom being directly on to the rear of the gpu. This is in an Antec P160 so there is a 120mm exhaust pretty much directly behind the CPU which seems to be doing a good job so far.
A 30min run of CPUburn showed the usual max load temp of ~58C on the cpu (an a64 3000+ newcastle OC'd to 230x10 and cooled by an SLK-948U w/ SilenX 92mm). This pretty much narrowed things down for me to the vid card so i tried some other games / apps and nearly everything 3d that i ran on it caused the system to climb up to 70c pretty quickly and then lock up. I looked at the mounting of the vga silencer and it looked fine which got my thinking of when the problem start and I identified the drivers. Since switching back to the 5.5 catalyst yesterday the problem seems to be gone.
This still strikes me as strange. Has anyone else had this kind of problem with a driver release? I can't help but think I must be missing something here...
Catalyst 5.6 significantly increasing heat production?
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
Shining Arcanine
- Friend of SPCR
- Posts: 502
- Joined: Sat Oct 23, 2004 2:02 pm
Hmmm, now that you mention it, the graphics core temperature on my 9800XT has been a few degrees warmer lately. I had just chalked it up to being summer.
I just found this post on the Rage3D forums about an X850XL also getting hotter after installing the 5.6, so it is not just us.
http://www.rage3d.com/board/showthread.php?t=33819860
I just found this post on the Rage3D forums about an X850XL also getting hotter after installing the 5.6, so it is not just us.
http://www.rage3d.com/board/showthread.php?t=33819860
Good to hear its not just me. I've been trying a few more things here and it definitely seems to be excess heat from the gpu causing my problems.
I've also noticed the problem being more of an issue in OpenGL based 3d apps vs Direct3d based ones... interesting that this driver release is touting a great increase in OpenGL performance at the same time. I'll be contacting ATi about it later today probably to see if they have any tweaks for the driver or something that could improve this situation while maintaining the new features etc in the driver release.
Sounds like a possibility. I'm pretty sure the main problem in my particular case is my CPU being "cooled" by air which is considerably warmer due to the extra heat from the GPU which wasn't previously an issue. Using the 5.6's with the side panel off my case and a high output 120mm fan flushing a lot of the warm air from the area in question still yields CPU temps a little higher than normal so there could be some extra heat there. Of course we're also into the summer months so that has an effect too. I'm still experimenting and so far it definitely seems like the gpu itself is heating up considerably faster with the 5.6s. I expect the CPU would still be handling it just fine without the added GPU heat that seems to now be present (as evidenced by the fact that it handles cpuburn / prime95 / etc just fine when the GPU isn't loaded).lenny wrote:Is it possible that with more efficient drivers the CPU has to work harder, thereby resulting in the crash? Just speculating here... I have the same video card so I guess I'll hold off on the update.
I've also noticed the problem being more of an issue in OpenGL based 3d apps vs Direct3d based ones... interesting that this driver release is touting a great increase in OpenGL performance at the same time. I'll be contacting ATi about it later today probably to see if they have any tweaks for the driver or something that could improve this situation while maintaining the new features etc in the driver release.
Test on X800XL
I tested on AGP X800XL Card. Both Catalyst 5.5 and Catalyst 5.6 have the same temperature. I use 3DMark05 Demo, with 2xAA and 16xanimorphic
Room temp 29C
Environ temp 53C
GPU temp 63C
Room temp 29C
Environ temp 53C
GPU temp 63C
-
scotty6435
- Posts: 100
- Joined: Wed Nov 03, 2004 3:02 pm
- Location: Salisbury, UK
As I remember, ATI Cat 5.6 optimized by calculate less to perform the same, not more. This make faster fps with the same calculation speed. We should expect the same temp, cause the same voltage, same calculation speed, just the eye effect (fps) is faster.Tzupy wrote:If there is a significant increase of OpenGL performance with the new drivers, then you should also expect an increase in heat output, since the GPU load is now higher. Similar problems happened when Doom3 was launched, especially with ATi cards.
no underclocking. All in stock clock.scotty6435 wrote:Did you have your card underclocked for 2D or something before the change?
with 5.6
- idle /2D for more than 2 hour.
- run 3DMark05 test
- run 3DMark05 demo - see highest temp with ATI tray tool graph.
change driver to 5.5
- uninstall 5.6
- restart
- install 5.5
- restart
test with 5.5
- run 3DMark05 test
- run 3DMark05 Demo - see hightest temp with ATI tray tool graph
3DMark05 Demo takes ~10 minute, makes run 2D before, irrelevant. It may not reach steady state yet, but I think quite close.
btw, I use Zalman VF700 AlCu for my test.
Re: Test on X800XL
Same findings here with an X800XL, no temperature increase. Comparable temps too.widman wrote:I tested on AGP X800XL Card. Both Catalyst 5.5 and Catalyst 5.6 have the same temperature. I use 3DMark05 Demo, with 2xAA and 16xanimorphic
I do underclock it to ~200Mhz core (down from 400 stock) while 2Ding, but to tell the truth I do not notice any temperature drop: it still idles ~50C. Stock cooling lets me overclock to ~440MHz, so I'm looking forward to that VF700.scotty6435 wrote:Did you have your card underclocked for 2D or something before the change?
-
CX23882-19
- Posts: 80
- Joined: Mon Mar 21, 2005 11:18 am
- Location: Hessle
Those with "hot" cards, are you using standby or hibernate? Last night I put my PC in Hibernate and when it came out the GPU temperature rose to 60°C whereas normally it will be at 35-40°C. I wasn't even aware of it except that the fan was on and off like a yoyo. This was just sitting at the Windows desktop doing nothing and I let it sit for half and hour and it didn't budge.
After I rebooted the temperature began to slowly drop; so it seems that the card really was that hot for some reason and not some sort of glitch with the temperature reading. After a couple of minutes the temperature had dropped down to 40°C.
After I rebooted the temperature began to slowly drop; so it seems that the card really was that hot for some reason and not some sort of glitch with the temperature reading. After a couple of minutes the temperature had dropped down to 40°C.
Last edited by CX23882-19 on Tue Jun 28, 2005 8:00 am, edited 1 time in total.
Ahem, it is, from what I know, impossible for a GPU to stay hot during hibernation. Hibernation totally switches off one's pc, saving ram data to a file called hiberfil.sys. The entire pc switches off, and nothing at all is left running. Likewise, when using STR (s3) mode, one is running only one component of one's pc - the ram (and supplying current to it via one's mobo), and hence the GPU cannot be on.
The only sleep mod which could cause this is POS (s1) mode, which keeps things going (though not usually one's gpu).
The only sleep mod which could cause this is POS (s1) mode, which keeps things going (though not usually one's gpu).
-
CX23882-19
- Posts: 80
- Joined: Mon Mar 21, 2005 11:18 am
- Location: Hessle
"Ahem", I didn't say that the temperature was 60°C during hibernate or suspend, it was that temperature after coming out of it (although I guess I didn't make it that clear - ammended now) - i.e. after coming out, the temperature rises very quickly to 60°C, just as it does when you start a 3D app.Dominic wrote:Ahem, it is, from what I know, impossible for a GPU to stay hot during hibernation.
PS: If I was going to be anal then I could say that devices are kept powered during S3 and S4 modes, and even in S5 (soft-off) with modern motherboards. It's not relevant though as the heating isn't occuring during S3/S4, it's occuring after waking from it.
