Score too low and overclocking having negative results

[Ex6tenze]

Heyo, I've done some benchmarks and got some rather... Troubling results.

My hardware is a staggering 3K points under the average. https://www.3dmark.com/3dm/53991120?

Finding some suggestions online, I swapped over to "maximum performance" in the Control Panel, which boosted it up to 13k average. Still roughly 1k short. https://www.3dmark.com/3dm/54008345?

Doing some overclocking to boost my GPU as it ran stock, I instead got a backwards result. I used a toned-down version of an overclock from a friend, who has a 2070 themselves. https://www.3dmark.com/3dm/54027127?

It's immensely peculiar and I can't seem to even get near the average. I know Ryzen isn't too splendid for raw gaming performance, but the difference between the 3900x and 9900k was only marginal. Not enough for me to score closer to my friend's 3700x - 2070 build as opposed to the average. Any clue, and any way I can fix this?

[Migz - DH]

Hi.
Your 13,693 overall score, 14,111 graphics score, seems to be pretty close to where most systems with your configuration sit in Time Spy according to 3DMARK's page. Although it's not pushing the upper bounds, most systems can't, and I wouldn't necessarily say that that's a problem.

That your system dips down to 12,596 overall, 12,703 graphics score, when you overclock the GPU, however, is bad. The GPU's average temp hit 75C in this test vs 72C in the higher-scoring one.

Have you tried to run the test with all of your fans maxed out and the case open? If not, try that to determine if you have a thermal problem. (Also, if you didn't already know, you can use GPU-Z's Sensors to help determine if there's a performance cap being hit.)

Edit: Oh, you might try using Time Spy Extreme rather than regular Time Spy btw as that will likely be a better test for your hardware, because the Extreme version will tax everything more.

[Ex6tenze]

Originally posted by Migz - DH:

Hi.
Your 13,693 overall score, 14,111 graphics score, seems to be pretty close to where most systems with your configuration sit in Time Spy according to 3DMARK's page. Although it's not pushing the upper bounds, most systems can't, and I wouldn't necessarily say that that's a problem.

That your system dips down to 12,596 overall, 12,703 graphics score, when you overclock the GPU, however, is bad. The GPU's average temp hit 75C in this test vs 72C in the higher-scoring one.

Have you tried to run the test with all of your fans maxed out and the case open? If not, try that to determine if you have a thermal problem. (Also, if you didn't already know, you can use GPU-Z's Sensors to help determine if there's a performance cap being hit.)

Edit: Oh, you might try using Time Spy Extreme rather than regular Time Spy btw as that will likely be a better test for your hardware, because the Extreme version will tax everything more.

I have done as you said, and ran the Extreme test with all components running at stock speed. Average of about 7k, but I end up at the lower spectrum of 5.5k yet again. https://www.3dmark.com/3dm/54142598?

The 13.6k test was a one-off, and I can't seem to reproduce the scores anymore.

[Migz - DH]

Ouch. Ya, 5630 seems really low for that setup in TSE.
The 73C temp isn't bad, but that's at the lower clocks again.

Did you try overclocking in TSE with the fans up and case open?

[Ex6tenze]

Originally posted by Migz - DH:

Ouch. Ya, 5630 seems really low for that setup in TSE.
The 73C temp isn't bad, but that's at the lower clocks again.

Did you try overclocking in TSE with the fans up and case open?

I've tried a light OC of 100 on the core and 800 on the memory, and got just over 6000. I'm not going higher yet, as the GPU temp was hitting 83/84C on the top end. GPU fans only up to 41%, so I'll look into the BIOS for that.

Here are the results. https://www.3dmark.com/3dm/54156246?

[Migz - DH]

Try again once you're able to max out the GPU fan speed.

[Ex6tenze]

Originally posted by Migz - DH:

Try again once you're able to max out the GPU fan speed.

I struggle with that too, I can't get them higher than 1900 RPM, while others report them going up to >3000.

[Ex6tenze]

Originally posted by Ex6tenze:

Originally posted by Migz - DH:

Try again once you're able to max out the GPU fan speed.

I struggle with that too, I can't get them higher than 1900 RPM, while others report them going up to >3000.

Nevermind, I managed to adjust the entire fan curve. They're now at 4000RPM.

[Migz - DH]

Good! I expect that'll help some. :-) (no, you won't want to play with them that high, but at least you can check how much temps are affecting your score.)

[Ex6tenze]

Originally posted by Migz - DH:

Good! I expect that'll help some. :-) (no, you won't want to play with them that high, but at least you can check how much temps are affecting your score.)

Not... All that much, unfortunately. Still with the 100 core clock and 800 memory clock.

https://www.3dmark.com/3dm/54157535?

[Migz - DH]

61C ave temp but still well under average score for similar systems.
I think you can remove GPU temps from the list of possible causes at least.
I'm not sure what the issue would be though. Perhaps someone else will have ideas for you.

[Ex6tenze]

Originally posted by Migz - DH:

61C ave temp but still well under average score for similar systems.
I think you can remove GPU temps from the list of possible causes at least.
I'm not sure what the issue would be though. Perhaps someone else will have ideas for you.

We've been looking into it more. Perhaps it's the fact that I use a single cable splitting into 8+8 pin cables, rather than two separate 8 pin cables?

[Ex6tenze]

Here's another comparison benchmark to someone with the same system, but lower clocks. They manage to get where I can't, somehow.

[Migz - DH]

What does GPU-Z say about the performance cap?
And, I suppose it's very possible about the power cables. Is your power supply not powerful enough to run a 2080Ti? How much power does the card draw during use for you?

[Ex6tenze]

It's a Corsair 850W. On forums, the consensus was "Power hungry cards should get separate cables.". And considering the 2080ti is the beefiest of the 2000 series, it's a worthwhile try.

GPU-Z registers a nice hit of the full TDP in Extreme, and no odd artifacts on separate modules during the benchmark.

[Migz - DH]

And, how about the "PerfCap Reason"? Does anything show in that GPU-Z Sensor line?