Unigine Valley: Lower FPS during benchmark than normal run?

[VanGoghComplex]

So I just got a new card, and am trying to benchmark it to compare to the scores my old setup got. While my FPS stays in the 80s during the normal rendering (outside the benchmark), when I F9 to start a bench, it stays there for a few seconds and then "throttles" to 30 or 60 FPS, almost like it's forcing VSYNC. I have VSYNC turned off in both my driver and Unigine, though.

Needless to say, this forced lower FPS is throwing my scores off, so I can't see if I went forward or backward from my Crossfire setup to my new Nivida 970.

[rotNdude]

Don't mess with the buttons when you're running a benchmark program. Just let the benchmark run and see the results when it's done running.

[yberkurko]

Początkowo opublikowane przez rotNdude:

Don't mess with the buttons when you're running a benchmark program. Just let the benchmark run and see the results when it's done running.

F9 is quick key to start benchmark in Unigine.

[VanGoghComplex]

Początkowo opublikowane przez yberkurko:

Początkowo opublikowane przez rotNdude:

Don't mess with the buttons when you're running a benchmark program. Just let the benchmark run and see the results when it's done running.

F9 is quick key to start benchmark in Unigine.

Yeah, that.

[rotNdude]

What are you using to monitor the frame rate outside of the benchmark? I guess I'm confused regarding what you are comparing. If you're looking for results from a benchmark, just let the benchmark run and let it take over full screen.

[VanGoghComplex]

Are you familiar with Unigine Valley specifically? It has built in frame monitoring, I'm not using anything else.

So Unigine Valley runs, and begins rendering the scenes right away. You can just let it do this for as long as you like, it'll utilize the GPU to the same extent as when you begin the benchmark. The only difference during the benchmark is that it's playing its set of 18 scenes (the same ones as the aforementioned) on a timer, while keeping a running average of framerates and developing a quantifiable "score" for use in comparisons.

My question is this: during the "normal" rendering (when the program is running and drawing scene but not benchmarking) I get 80some fps. When I start the benchmark, the first few seconds of the first scene stays at 80, then drops to 30 or 60 for the rest of the benchmark. This is irritating because I'm trying to get an apples-to-apples score to compare to my old video card setup, as I just changed them, but my old cards never "throttled" like this. And I do not have vsync on in the benchmark or the drivers.

[rotNdude]

Oh yeah, now I understand. Like I said previously, let the benchmark run until it's completed and then compare the score. I think when you tell that software to start the benchmarking, you don't worry about the FPS meter, you worry about the final score.

[VanGoghComplex]

Początkowo opublikowane przez rotNdude:

Oh yeah, now I understand. Like I said previously, let the benchmark run until it's completed and then compare the score. I think when you tell that software to start the benchmarking, you don't worry about the FPS meter, you worry about the final score.

I am getting a lower score with my Nvidia GTX 970 than I got with a single Radeon HD 7870, because it is throttling with my new card, and it did not throttle with my old one. The average fps recorded by the benchmark is a determining factor in the final score.

[rotNdude]

OK, posting 101 in this forum area is to give us all your computer specs. Do it!

[VanGoghComplex]

On it!

CPU: AMD FX 8350 OC'd to 4.7GHz
Mobo: Asus Crosshair V Formula-Z
GPU: MSI Nvidia GTX 970 Gaming 4G
RAM: Crucial Ballistix 16Gb 1600MHz
SSDs: Sandisk 128Gb (OS) and two Corsair Force GT 240Gb (games)
Windows 8.1

[rotNdude]

No power supply? Did you remove the AMD video card drivers before you installed the NVidia card?

[VanGoghComplex]

Haha, power supply is a Corsair RM850. Yes, I scrubbed the AMD before installing the new card and drivers.

[yberkurko]

I would say it's on the software side at least... as if the throttling starts only after starting benchmark run and not on free run. Driver issue seems most likely. Have you tried with some older version (if you are on newest at the moment) of driver if the issue persists?

[VanGoghComplex]

Figured it out. I run fraps in the background not to use the onscreen overlay, but because it reports to my Logitech G13 LCD panel. Apparently, F9 is the "start benchmark" key in Unigine, and it is also the "start video recording" key in Fraps. It was throttling me because of the vid recording.

[yberkurko]

LOL. Good that you figured it out. Seemed total mystery :-)