3DMark
3DMark > General Discussions > Topic Details
cave hermit Dec 20, 2019 @ 5:13pm
Do my results seem right for my hardware?
I recently upgraded to a RTX 2070 Super, and I wanted to benchmark my machine to make sure things are running as they should, as I was getting some split second freezes in CS:GO and other games, and this was my first time upgrading a GPU (mistakes were made, for starters I tried to jam the new graphics card in with no success not realizing there was a protective cap on the connector I had to remove. Then I bent the pci lock not realizing there were screws holding it in place, and then on my first bootup after install it couldn't reach the BIOs and just made triple beeping noises at me until I reinserted the GPU again)

Here is my score: https://www.3dmark.com/spy/9779451

Does this seem like the kind of result I should be getting with this new card? I tried the compare function, and compared to the best score that also used a i7-6700 I'm about 1000 points behind. People using better CPUs got much higher scores.
< >
Showing 1-6 of 6 comments
UL_Jarnis  [developer] Dec 21, 2019 @ 1:12am 
Just click "Show Result Details" to open the fold that shows how the score compares to similar hardware on the 3dmark.com page. Your score is highlighting the tallest bar, so it falls roughly to what this hardware is expected to perform.

Highest ones (which you are ~1000 points behind) are always overclocked results. It would also tell that the specific card you have could be pushed by up to 1000 points but that may require things like replacing cooling with a waterblock and obviously overclocking. Not something I'd worry about for day-to-day system. Highest results are always "tweaked" to get the maximum out of the hardware, with overclock that *barely* stays stable enough to get through one benchmark test and not the same as running with day-to-day settings.
Last edited by UL_Jarnis; Dec 21, 2019 @ 1:14am
cave hermit Jan 7, 2020 @ 11:42am 
So after using DDU and fiddling around with nvidia control panel in an effort to get rid of stuttering in some games, I felt as if perhaps I had made performance worse somehow so I ran 3dmark again. This time I got slightly worse results compared to my previous run: https://www.3dmark.com/compare/spy/10075509/spy/9779451

Given this comparison, is this just random variation I shouldn't worry about, or did I actually do something to mess up my system? I probably should mention that the first time I did the test I was on a 1080p monitor, but this time I ran the test on a 2 monitor setup with the primary being 1440p. Also this time when I ran the test I had just finished playing a game.
UL_Jarnis  [developer] Jan 8, 2020 @ 1:31am 
This is within margin of error (up to 3% is considered normal variance). To conclusively answer the question, you would need multiple runs.

If you run this same test five times and take an average, that would give some idea but without 5 runs from "before" it is still just a suggestion that something affected the performance, not a guarantee.

One of the runs was done with 30mhz lower clocks which would alone probably explain most of the difference. 30mhz lower clocks might come just from cold machine vs. machine that has been up for hours and has run heavy 3D for a while.

To verify if this is just thermal throttling, you could simply shut down the whole PC for 15-30 minutes to ensure it cools down, then restart and re-run.

cave hermit Jan 8, 2020 @ 6:57am 
Well I shut down my PC overnight, and when I woke up turned it on and ran the benchmark again: https://www.3dmark.com/compare/spy/10085584/spy/9779451#

This time the results were a bit better than my previous run, but my very first run was still better. It looks like the clock was a bit higher than it was during yesterday's run, but it still wasn't as high as the first run. However another variable has been added in the form of a Nvidia driver update which I applied last night.
Last edited by cave hermit; Jan 8, 2020 @ 7:01am
cave hermit Jan 9, 2020 @ 1:36pm 
https://www.3dmark.com/compare/spy/10103833/spy/10085584

Somehow my results are getting worse. This is a comparison of today's benchmark, taken after I removed and reseated my GPU, to yesterday's benchmark. Is this still within margin of error? I should probably mention when I removed the GPU from the PC it was still a bit warm.

Overall my score has decayed from 8608 at best the first run to 8349

Edit: Just tested it again and this time results were 8501.
Last edited by cave hermit; Jan 9, 2020 @ 2:51pm
UL_Jarnis  [developer] Jan 9, 2020 @ 11:28pm 
Within 3% is within margin of error. Score is still slightly lower but so little that this could be simply due to variance introduced by background programs and thermal throttling (modern GPUs have a clock speed of IT DEPENDS, ie. they boost as much as they can until they hit thermal and power limits and this can vary between runs)

For more accurate scoring, run 5 or 10 runs and take an average.
< >
Showing 1-6 of 6 comments
Per page: 15 30 50

3DMark > General Discussions > Topic Details
Date Posted: Dec 20, 2019 @ 5:13pm
Posts: 6