NEW Benchmark comparison between AMD & Nvidia - 2.0

What is this preoccupation with benchtests ? it is not 3DMark ?



Drive the thing !

 

That's true. The other possibility would be to start with everything low/off and than to run with high/max/on whatever is possible but both tests without AF and AA in use.

Let me come back to this after doing the solo bench.

 

Ok, but is a default player.json really necessary?

Good idea on the default Nvidia/AMD control panel 3D settings. Can you take a picture of that for me please and i'll do the same for Nvidia.

 

Solo lap benchmark on LochD./Civic - practice settings standard as it comes with the tracks first install - all other settings as the replay benchmark, except from AA is set to 4XMSAA in the CCC :

CPU @ 4000Mhz - GPU standard:

1. - Frames, Time (ms), Min, Max, Avg
10603, 84365, 107, 153, 125.680

2. - Frames, Time (ms), Min, Max, Avg
10554, 84646, 107, 152, 124.684

CPU @ 4000Mhz - GPU OC 10%:

1. - Frames, Time (ms), Min, Max, Avg
11391, 84256, 114, 164, 135.195

2. - Frames, Time (ms), Min, Max, Avg
11407, 84458, 117, 163, 135.061

 

Ok good. This is very interesting and telling.

So it would appear that for you, overclocking your GPU in the replay benchmarks didn't yield any difference (for whatever unknown reason) and in this live benchmark it seems to be working correctly.

Btw, i don't know if you made a mistake with stating (in the other thread) that you overclocked your CPU and not your GPU in the Firestrike benchmark but your performance increased 8% in rf2 with your 10% GPU overclock, identically to your 8% increase in Firestrike GPU performance.

 

DrR1pper, in relation to the firestrike bench, It should be no problem to test it again but that was the case yes, one test without GPU OC and one with 10% GPU OC but both tests where performed with the CPU @4000Mhz instead to 3800Mhz previously.

And yes i see the solo benchmark result is better with 10% GPU OC but it didn't does anything to the replay benchmark.

 

Those from you are pretty consistent but know that the inconsistency only arises when you quit out of the replay and reload it again. So technically there was only 1 chance of an inconsistency coming up in your 9 repeated runs and it was when you restarted rf2 (between runs 7 and 8).

The fact that Timpsie got an inconsistency though and how problematic the replay method has been shown over the last few days, i think it's just safer to use the solo live benchmark (and for sanity sake! )

Yes i have and it is superbly consistent. The lowest consistency i've found so far is 99.94%. It's pretty darn consistent.

 

Right, well without saying any of this it became all very confusing due to the scattered nature of the posts (and it still is).

If you want to show me you're cpu is affecting your gpu score, all you have to do is perform 2 firestrikes tests, both with constant gpu clocks and alternate the cpu speed. If no change....no cpu bottleneck.

 

Oh, i didn't see where you said that but ok.

 

Ah ok, my mistake. Sorry

Perhaps some people don't experience the problem whilst others do. The fact Timpsie did is enough cause for concern i think.

I'm running through all the solo live benchmarks now and will post results and testing method (etc) soon.

At least this way you'll also have the real performances of the cards to compare with instead of the consistently much lower performances from the replays.

 

I just tried to do another run just to see what results I got and again I got the 97fps avg one instead of 84fps. I'm not sure what the cause is but it certainly fluctuates in some cases. I didn't completely exit out of RF2 between benchmarks but did go back to the main menu and reloaded the replay.

 

Yeah....we're going to have to ditch the replay benchmarks. Thanks for mentioning the problem happens so easily for you.

 

Nope I don't see any difference apart from the fact that the benchmarks starts at 95fps instead of 82fps when waiting on the grid
And the faster runs happen randomly but have exactly the same results so I don't see how a background program or a driver setting could influence it in exactly the same way each time I get a faster benchmark.
Obviously there's some stuff running in the background (avast in gamemode, Steam, dropbox, couple more) and Nvidia settings are application controlled.

 

I was saying that I originally thought it would have been a good idea to do all that, but then I saw your message about making it quick and easy for everyone and realised you had a good point so ya, it may be too much for everyone.

How about we just tell everyone to click default in their NVidia/AMD global (or rF2 game specific) profile. Then from there just set AA to override application and to 4xMSAA (don't forget to click apply! lol). You think that would be easier and more consistent? Pre-rendered frames will also then be set to default (probably 2 or 3) instead of 1 for any users who have changed it, but must people who change it are NVidia users anyway and not so many AMD users (it seems way more rF2-NVidia people know about/use NVidia Inspector compared to the amount of rF2-AMD users who know about/use Radeon Pro).

So to summarize, "default", then 4xMSAA for everyone (and make sure rF2 AA is obviously set to "none")??...

 

When it happened to me it was very noticeable. It's not smooth at -16% performance and in the 2 times it jumped up 35%, not only was the fps showing it so but it was visually extremely silky smooth. The fps averages were 49 fps (-16%), 58 fps (expected) and 79 fps (+35%). Since it was around the 60 fps mark, was quite visually noticeable seeing the difference.

The two times +35 occured i had walked out of the room. However i check the benchmark excel spreadsheets first and when i saw it say avg 79 instead of avg 49, i went back in to watch the replay and sure enough it was running at a much higher framerate consistently throughout. Only when i quit out the reply and rejoined did it then suddenly (and depressingly) defer back to the avg 49 fps. So very strange.