HD 5970 GPU fluctuation and problems

Spandy87

Distinguished
Jun 19, 2010
14
0
18,510
I'm having a problem with my HD 5970.

I've noticed this in game and when running stability tests. My gpu usage dips from 95%+ to 70%. This is both with AND without overclocking.

furmarktest1.jpg


System specs.

CPU: Q6600 @ 3.5ghz
850 watt PSU
4gb ram
Sapphire HD 5970 (currently OC'd putting 1.72 volts into it, but the fluctuation happens with no OC).
Gigabyte EP45-UD3P mobo (with pci-e 2.0 @ x16)
Old-ass SATA hard drive @7200, apparently I've gotten lower than comparable results in HD tests, but I don't know if that relates to this problem specifically.

Interestingly enough, when using Furmark, my GPU usage fluctuates consistently (not temperature related as far as I can tell), but when I run the stability test in "extreme burning" mode the GPU's do not fluctuate and remain consistently above 95%.

furmarktest2.jpg


Also, games like bad company 2 and crysis still run like crap, like 30fps! It seems like I'm not getting the most out of my card.

Oh and I have the latest 10.6 drivers (and I've done the driver sweeper stuff as well).

Can someone explain to me what's going on?

What should I do about it?

What should my next upgrade be to maximize my 5970 (there might be a serious bottleneck somewhere)?
 

Spandy87

Distinguished
Jun 19, 2010
14
0
18,510
It's a Raidmax RX-850SS

About the drivers, I have tried the 10.5 but not the 10.4. I'll give it a go.

Also, about the temps, if you look closely at those screenshots, you see that the usage is stable in the second picture, which also has higher temps. So I am reluctant to think that's causing it. They've also never gone above 95 degrees.
 

notty22

Distinguished
Ati has these cards throttle around that temp, if you read the 5870 launch review at Anands they talk about the cards throttling, which is also different than what took place in ati's last generation. These cards need to run cooler, do what you have to do with the fan speed to keep them at least under 90c.
edit: You might want to check gpu-z, it offers more temp sensors, maybe your vrm's are heating up, in that top photo. You mention increasing the V ?
 

Spandy87

Distinguished
Jun 19, 2010
14
0
18,510
Yeah but like I said, it does this on stock voltage.

Also, look closely at the temperatures in those pictures.

It fluctuates THE MOST at lower temperatures, and almost not at all at 90+ so it can't be temp throttling.


 

G-REG

Distinguished
Jun 20, 2010
2
0
18,510
Spandy87 - you are NOT alone. I just received my new Digital Storm pc and I've been having the EXACT same problems with the HD 5970 that you, and a number of other people are having.

I have been trying to receive help from Digital Storm, Electronic Arts, and AMD. So far the only responses I've been getting has been from Digital Storm. EA and AMD will probably take another month or so before I get any response.

This is the Digital Storm forum I'm posting on:
http://www.digitalstormonline.com/ [...] 733#127733

I have been going through a myriad of solutions to get the problems fixed to no avail.

However, using benchmarks, such as Furmark and 3dmark06, have resulted in some interesting data about my new, "killer" HD 5970 card. I found that my card can handle 89C temps (maximum) - (minimum was 51C) just fine via Furmark strenuous "Burn" stability testing. So your worries of the card heating up to 92C maybe unfounded. Although, it seems if you can fry an egg on your computer components then you would think the electronics can fry as well. I too, am extremely nervous about these high temps, but I also said that about other cards I've owned over the years, and those cards performed time and again satisfactorily.

Have you found a fix yet? I'm assuming you would post your fix and results if you did.

I'm rather hesitant on using fixes like overclocking and downclocking or forcing directx versions. Changing the cards performance outside of GPU driver functions have always got me into other dilemmas and problems in the past. I have found that waiting for the right GPU driver that fixes the vid card bugs gets the job done better. I must admit though, this is the first time where I've experienced a GPU manufacture (AMD) that could not fix a vid card's bugs with any driver they come out with. Even the latest drivers (version 10.6) don't seem to be working for me. I'm wondering if this has anything to do with the fact that there are so few people who were able to purchase the HD 5970 and not enough people complaining about the issues of this card in comparison to the other AMD cards, thereby AMD only responding to the "squeeky wheel" or should I say the "squeekiest wheel". Perhaps, as more "suckers" (at least thats the way I feel about myself at this moment) purchase this card, then theres more of a chance AMD will fix it's bugs through a good driver. I am soooo jealous of those people who have purchased any other AMD vid card besides the 5970, because their systems are not having the same problems as we are. Quite often I get responses to my forum requests for help like, "I'm using the latest ATI driver and I'm not having any problems at all. You must be doing something wrong." And I find these people have 5870 or 4870 cards. NOT EVEN IN tHE SAME BALLPARK!!

Please post any new developments you may have.

Thank you so much for your posts.

"SO MUCH FOR THE FASTEST VID CARD IN THE WORLD!"
 

Spandy87

Distinguished
Jun 19, 2010
14
0
18,510
I tried this. I initially suspected that this was the problem because it seemed obviously a PSU issue. But now I'm not so sure. I even used the molex to pci-e provided by sapphire incase there was a cable issue. Still haven't found a fix for this problem.
 

Spandy87

Distinguished
Jun 19, 2010
14
0
18,510
Okay so heres some new info on this issue.

I reset all of my stuff to stock clocks, and ran the stability test on furmark and other similar programs (furmark is the most consistent).

I got very high gpu usage fluctuations, along with lower overall gpu usage. It went to a maximum of 85% for both cores. I got about 100 fps at low res on stability only mode.

Then I went and overclocked my GPU, I got higher overall usage, but also higher spikes, going from 70-90%. Not much real difference though.

Then I reset my gpu clock to stock settings, and overclocked the hell out of my cpu.


This is where it got interesting. I got higher overall usage and a lower differential in fluctuation with an overclocked cpu. Naturally I got way higher fps, but that was to be expected.

I don't really understand what is going on here, because you would think that since my cpu is the limiting factor here, that my gpu would be doing most of the work.

Yet it is clear by these tests that the more I OC my cpu, the more USAGE I get out of my gpu.

Of course I get better performance with an OC'd cpu and I understand what that does, but I'm talking strictly about the gpu USAGE only.

Is this just a function of my cpu bottleneck that I'm somehow now understanding correctly?

*note, I have my Q6600 at 3.6ghz

**Also, what is the max temperature that I can safely run a Q6600 at? it seems I can keep cranking my cpu but I'm afraid to break something. core-temp says low:48, high:52
 

Spandy87

Distinguished
Jun 19, 2010
14
0
18,510
Well, I think you're right, its an outdated cpu and I should upgrade.

But riddle me this, in a game like dirt2 I get only up to 50% gpu usage!

Its not THAT cpu intensive is it?

Even when my fps dips to like 40 (which it does occasionally), my gpu load isn't more than 60%, and they're NOT hot. They're at like 75c.
 

Spandy87

Distinguished
Jun 19, 2010
14
0
18,510
If I'm getting like 21559 gpu score (performance) in vantage, is that okay?

I see my gpu usage on average stay around 65%, but then in certain situations in the gpu tests my usage spikes up to 99%.

My fps also goes up at those times of course.

Is this anything other than just my cpu bottlenecking my card?

Tell me if I'm completely wrong about this, but here is how I understand bottleneck.

My CPU runs at a certain speed. That speed limits my fps to a certain point. My GPU has the potential to run at much higher fps. Whenever there is a situation that reaches to my CPU's limit in a CPU intensive test of some kind, my GPU usage will go down. This is because there is no reason for it to run at its full potential when it won't reach those FPS anyways due to the limit of the CPU.

Without getting too technical, is this roughly how my GPU usage would be related to my cpu bottleneck?

If so, explain this to me. These tests are run at 1280x1024, and I thought that at lower resolutions, the cpu can come closer to the GPU's limits. Thus the GPU usage should be higher.

Ultimately my goal is to have this card have higher average GPU usage.
 
Raising the resolution allows the GPU to stretch its legs more, and you might get a higher usage out of it.

Download the AMD fusion utility for desktops, run that, go to advanced mode, and run MAX PERFORMANCE. That will disable every unnecessary service to run windows, and you should be able to get more out of your CPU. Once you're run, this will restore every closed service.
 

Spandy87

Distinguished
Jun 19, 2010
14
0
18,510
Severe fluctuation fixed!

Installing 10.4 drivers again fixed the issue with furmark. My gpus now run at 95% in furmark!

I also found that with these drivers Bad company 2 runs at an average of 55-60fps!

Thanks for all the help!
 

Jester99

Distinguished
Aug 9, 2010
6
0
18,510
I am running my 5970 with 10.7 drivers. with all benchmarks and games im getting 100% utilisation of both GPUs. 10.4 proved to do nothing..

I also run a varying number of clock speeds on my 5970. Normally its 800mhz core with 1100mhz memory @ 1084 mV ... if i push the voltage up a bit i can get 920mhz core and 1250mhz memory stable ... but im not pushing it anymore until i get my water cooling block attached (the fan is like a hair dryer on max speed when i turn it up)