NEW Benchmark comparison between AMD & Nvidia - 2.0

CPU @4000:

290 standard:

Frames, Time (ms), Min, Max, Avg
13004, 200212, 52, 81, 64.951

 

CPU: i5-2500K overlocked to 4.2GHz
rFactor 2 Build: 860
Graphics driver version: 340.52

GPU: GTX 670 Stock 1137 core/3055 mem

1st Bench Run:
Frames: 11718 - Time: 200056ms - Min: 48 - Max: 72 - Avg: 58.5741


GPU: GTX 670 Overclocked 1215 core/3348 mem (+78 core, +293 mem)

1st Bench Run:
Frames: 12372 - Time: 199978ms - Min: 50 - Max: 77 - Avg: 61.867

 

This is very intresting and with other builds i have never noticed that rf2 doesn't knows my CPU. The most interesting is the indicated CPU frequency changes with each restart of rfactor2.

View attachment 14391
View attachment 14392
View attachment 14393
View attachment 14394

edit: and she is sleeping during the process. I taken this shoot shortly before the finishing line on the second turn.

View attachment 14395

Looks like as if rf2 not only don't likes the AMD GPU's, it also don't likes AMD CPU's as it seems. What a mess.

 

CPU: i7-3820 @ 3.80GHz
rFactor 2 Build: 860 64bit exe
Graphics driver version: 344.11 (PNY GTX670 stock speed)

2014-09-28 11:59:46 - rFactor2
Frames: 11981 - Time: 200312ms - Avg: 59.812 - Min: 50 - Max: 73

2014-09-28 12:04:55 - rFactor2
Frames: 11946 - Time: 200343ms - Avg: 59.628 - Min: 50 - Max: 73

 

Oh ya, nevermind, ignore me, lol.

 

Ah s***!

 

Exactly, because of Nvidia's stupid method of ever changing dynamic clocks, most people don't truly know what clock their GPU is actually running at, and that's really all that matters.

Using the graph of Nvidia Inspector, EVGA Precision, or MSI Afterburner, you can see a very usefull chart that will show you the history and you can see how much the GPU is fluctuating and to what clocks. You may discover that it's not even constantly pegged at 993/1087/1189 MHz, in your case.

 

Speed1, can you do a 1 lap benchmark (like we did before with AI driving 1 outlap recording from pit exit to finish line at whatever track and car combo) and do this bench twice. Then go and overclock your gpu and bench twice again (but don't close rf2 between the benchmarks!). Tell me what your scores were please for the different gpu clocks.

 

Ya. Specific AA levels should be instruced to be set in the GPU control panel for this test rather than rFactor 2 "levels". "Levels" can be different. Possibly even different GPUs from the same manufacturer with different forms of AA available (real old GPUs) may therefore be using a different AA setting for the same "level". Same with the future - AMD/NVidia sometimes adds/removes AA types, and those could affect what the rF2 "levels" correspond to.

Should just be set in control panel to remove any doubt and chance of disparity.

 

Right guys, since there was shown to be a very significant inconsistency with Timpie's result on a 3rd repeat, i don't think we can rely on replay benchmarks at all (at least not with the current build if that is the cause of the problem). I'm going to do tests on both my cards again but with a solo live benchmark method with the AI completing 1 outlap (from pit exit to crossing the finish line) with the Civic at Silverstone.

I will make a new thread soon with all the details required so it's all consistent for everyone, etc.

I think we should also do it without any rf2 AA settings since there seems to be confusion about which level of AA in game corresponds to the equivalent AA level in AMD and Nvidia cards. What do you guys think?

I want to keep the benchmark method as simple and as little time consuming as possible, so not sure about doing 2 sets of repeat benchmarks (so a total of 4 runs would be required), one for no AA at all and one for 4x AA in the AMD/Nvidia control panel?

 

It doesn't hurt to check, you may be suprised. Before I got a custom BIOS with a dead-stable 3D-gaming clock, I too was well below temp and power as well, and there were still some fluctuations. Sometimes just a few here and therefore, but sometimes it was bad and could cause sudden fps slowdowns and very specific moments, or just generally lower avg fps. It depended on the game, fps, VSync on or off, etc.

 

Nice, I agree. In my opinion, and as you suggestsd, AA should be set, and in the control panel. How about 8X instead, in order to put some more load on the GPU? Up to you, but ya, at least 4x, and from control panel.

 

Yes no problem but actually it makes no sense anymore if you look at my report below.

What did you guys think about this ?

I even would go further and don't use any AF filtering or AA samples to eliminate the possibilty of any speed/performance hacks of the drivers by the manufacturer. Actually it even would need to test rf2 in steps, display/gfx feature on/off and from low to high settings. This could help to sort out the weak points in relation to the GPU manufacturer and ISI software.

 

This is too much speed. As nice as it would be to have a full breakdown of performance for each graphics card with rf2 gfx settings at many different settings....i think it's too messy and time consuming. We want to keep it as short and least time consuming as possible enough so people will be willing to do the benchmark.

 

How about we make a current benchmark lap with the current build as well. It only makes sense.

 

Isn't the latest build (860) what we have been using to benchmark with this whole time?

 

I tried with the Benchmark recommended settings, and then with the settings I use.

My system is AMD Phenom II X4 970 @ 3.6ghz
AMD HD 6950 2 GB with bios switch set to 6970 14.4 driver.
8 GB DDR 2 800 memory

Benchmark recommended settings result:
Frames, Time (ms), Min, Max, Avg
6584, 199947, 26, 44, 32.929

Result with MY normal settings:
Frames, Time (ms), Min, Max, Avg
12755, 204283, 51, 79, 62.438

Quite a difference, and to me, I can not really tell much of a difference in the quality. So my settings almost double my FPS., with very little quality difference.

 

Ya, I was going to say, show a picture of the Nvdia and AMD control panels showing which settings to select to disable all the optimizations and stuff, and on top of that, provide a 100% default player.json file in the benchmark download package, but I agree with Dr.R1pper, it's just too much for some and might turn people off.

 

Speed1, you need to use some sort of monitor program to monitor and log your CPU clocks, and for each core/thread while doing these tests. As DrR1pper said, something seems really messed up with your CPU as even the overclock didn't do anything in the Firestrike CPU score...

Also, what version of Windows are you running? Set your windows power mode to "High performance", and if using Windows 7, then please download a program which will allow you to edit and disable "core parking" in the power settings for the high performance profile. You don't need to do the "core parking" fix with Windows 8 or 8.1, those operating systems are much, much smarter when it comes core parking.