Random FPS Drop, Due to Card Overtemp? 4870's

This is a new thread regarding an issue with my system that came from here, since I don't believe it is so much an issue with Crossfire anymore as something else.

I did some data logging with my system using HWiNFO32 while I was playing Dragon Age: Origins and I came across something that I pretty much expected. Periodically my GPU Utilization goes from 99% down to ~50% on my #1 adapter and from 99% down to ~0% on my #0 adapter. That, is not normal by any means, HWiNFO32 also only shows my GPU fans running at 30% approximately 90% of the time, with the #0 (assumed primary video card) never reaching above 30%, and my secondary card (#1) will periodically hit 100% fan speed.

Honestly, I'm lost, I just did another data collection with the game simply running with my character sitting next to a fire. That way I could have a sustained time frame to collect data where none of the variables would be changing, i.e. loading screens and combat. During this time period the game still had drops in FPS equating to approximately 50% of the normal max. The data I've collected is attached in an excel spreadsheet format, or at least a .CSV. In order to upload the spreadsheet I had to change the extension to .doc, so if you take a look you'll have to change it back to .CSV.

Thanks

 

Yeah it looks like your GPU #2 is hitting over 100oC and your CPU core 2? (TEMP 1) is hitting 98 while the other cores are 50's and 60's. They should be much more even than that I would think. Unless that's what something looks like running soley off one core?

I'm not sure about the fan %, but I have always read that it doesn't report correctly for our machines. To be honest I didn't think it ever showed anything except 30% until I saw this log.

Have you done, or considered a GPU & CPU repaste? If you'd rather not do that then I'm not sure how well tech support will respond to high temp complaints, but maybe calling it in, and send them a log like that might help get it fixed.

Have you tried running with crossfire disabled? Except for running with less FPS, does the issue go away? Does one GPU and CPU temps become normal, or does the one CPU core still overheat?

 

All of those things I've considered, however, I got screwed out of my warranty because this was a replacement for my R1, which took damn near a year for them to finally put in. So, I have no warranty anymore. I am deployed currently however I am having my wife send me some things so I can essentially redo all the seating and cooling paste. Regardless though, it's still ridiculous.

 

A bit more testing and the results are fairly close to my original results. This time I used GPU-Z to record the data and have uploaded it here. I believe this data was recorded solely for my primary adapter, which, sure enough the GPU load would drop from ~100% to 0%. The temperatures are not reading nearly as high as HWiNFO32 shows them however, all approximate 80°C-90°C. This data shows the card also down-clocking as well. How can I stop this? ~85° is not great, but it isn't bad or reason to worry.

 

Run two instances of GPU-Z at the same time, one monitoring each card. Then, under the 'Sensors' tab, you can actually click in the boxes and you will see tiny words appear. If you click twice, you will see 'MAX' appear in the box. This will, of course, keep the maximum temp, clock, etc. You can then post screen shots of that instead that grueling .txt document. The sensor tab for HWinfo32 does this too.

 

I just did a complete reinstall of all my graphics drivers, after uninstalling everything, all the "tools" I've downloaded and then installed the 10.10 drivers, followed by the 10.10e hotfix from Guru3D.com. I suppose I'll run the Multi-GPU FurMark and provide a few screenshots as opposed to logs.

 

I am so confused by this system right now. GPU-Z isn't showing one of my adapters being used at all anymore, by ANY of the three drivers, 10.10 Dell, 10.10, and 10.11. I'm not sure happened but it's really pissing me off. The fans are running like crazy and the temperature is definitely raising, but it is also obvious in game now that the video cards aren't working properly. GPU-Z and HWiNFO32 show my secondary adapter at 0% usage and my primary is only reaching about 40% now. Also, GPU-Z is not displaying a value for the Default Clock/Memory speeds.

I'm beyond lost.

 

That happened to me once! Unfortunatly, I have no idea why it went away. Just one morning, I wanted to see if it was still messed up, after having wallowed in the pit of despair for a while. And it worked. I was overjoyed. The only thing I can think of that I did was leave it alone for a while. I have no idea why that would help, but in my case, hopefully yours as well, it did.

 

Is CrossFire enabled in CCC? Does GPU-Z say that CrossFire is off, but available? Try doing a complete power drain, to include the CMOS battery.

 

I believe what happened was the cord to my power supply came lose or something of the matter and the R2 just disables the secondary card, regardless, I installed the 10.11 drivers and just loaded into a game of DA again. Here are some SS's of the game I was playing. My GPU Loads drop periodically to ~0% or ~40%. I am assuming this is when the FPS drops to nearly half or less of what I normally get.


-- Primary Card (the game was closed for a bit and I don't know how to "pause" the data for the SS there, oh well)


-- Secondary Card

Also, for whatever reason, I can't get my "Default Clock and Memory" to show up as a proper value. It just stopped working one day, no matter what drivers I have installed.

 

Those temps are too high. No game should get them there. You probably have some thermal throttling going on. When is the last time you cleaned the gunk out of your heatsinks or applied new thermal paste?

 

The 2nd GPU is fine but the 1st is a little high really. But shouldn't cause such throttling. The 5870 thermic shut down limit is 114C. Not sure what the 4870 is.

And Turbod unless he is out of warranty then he can't apply different paste except Dell's crap stuff.

 

Cleaned it out probably about two weeks ago and I've only had the system for about 4 months. In regards to the thermal paste, I haven't done it because it should of been find out of the box, granted I wasn't expecting 60°C temps while playing the heaviest games, but still.

 

All of his GPU temps accept for one are above 95°C. Those are not good temperatures for regular gaming. A stress testing program like FurMark should be the only thing that ever gets them that high. The one 74°C temperature is because he didn't set it to keep the max temp, or I'm sure it would have been 95+ as well. When Dell technicians come out, they use whatever they've got, so I'm quite sure it doesn't matter. What game are you playing that gets your cards this hot? I believe there is an issue, and the laptop needs to be opened up.

 

I see what you mean now, yes. However, when I game I sometimes leave two GPU-z open to monitor temps and my 2nd GPU is always about 10C cooler or so.

My max temps for gaming are about 80C but most of the time they are about 74C tops.

EDIT: Having just got a Coolermaster Notepal U3 because I felt the Lappy needed to be a bit cooler, I noticed that Black Ops no longer has minor slow-downs the OP is describing in other games. Further reinforcing the temperature issue. But also having asked Dell about my temps, I quote:

''100C is fine''

In relation to my 5870m setup. You may want to call Dell and ask what temps you should be having.

 

Yea, the max temp for my slave card was 98 degrees. This is in Dragon Age: Origins with no AA, Graphics set to High, not extra. Also, these cards are rated to up to 120°C according to ATI, however, that is the chipset I suppose, not dells ized version using GDDR3