NEW Benchmark comparison between AMD & Nvidia - 2.0

Ran it the once just out of curiosity.

CPU: i7-4930K overlocked to 4.3GHz
rFactor 2 Build: 860
Graphics driver version: 344.11

GPU: 2x GTX 780 Overclocked (1100 boost)
Full loop custom water cooling

1st Bench Run:
Frames: 17972 - Time: 199526ms - Avg: 90.074 - Min: 73 - Max: 116

Quite frankly pathetic. Looking at the other single card scores if this software had anywhere half reasonable scaling for multi GPU setups it should be smashing them out of the park. I see a few people mentioning 3DMark Firestrike. For reference my top score on that is just over 17000 points with a GPU score of over 21000.

 

ya, rF2 seems to have messed up SLI scaling - in 2D mode - for many people, but for others it seems to work fine. The issue seems to become worse when triple screens, multiview and maxed graphics are involved. Because it gets just about perfect SLI scaling in 3D, I probably get higher 3D framerates than you get in 2D mode, lol! Is the bad 2D-SLI scaling an NVidia or GMotor issues? Who knows. What I do know is that it's not just up to NVidia to make a profile; certain things need to be done on the game engine level as well. I'm not that ISI hasne't done this, I'm just saying it could be either (or both) issues.

 

Stephen, interesting SLI results. I was using GTX 460 SLI about 2 years ago in rf2 and it was working properly. My single gtx 460 was performing right relative to other 460 owners and when I enabled SLI, fps nearly doubled.

Is it possible the replay Problem extends to SLI?

You'll be able to find out in a little but. Look out for a new benchmark thread within the hour.

 

So, back to this, what is that for a tool homie ?

 

AMD 1100t @ 4.00 GHz
8Gb ram @1.6 GHz
GTX770 OC

Frames, Time (ms), Min, Max, Avg
13638, 209275, 52, 85, 65.168

I have the same problem of Speed1 with rFactor2 not recognizing my CPU...

 

So DrR1pper, what do you think about post #60? (Default and 4xMSAA)?

Core parking means that Windows effectively shuts off cores to save power.

Setting Windows to high performance mode changes many back-end settings with regards to CPU power, cores, hard drive stuff, GPU power, USB power, all sorts of power saving and performance stuff. I wouldn't worry about all that though; for now, just select the high performance mode and....and give me 5 minutes to update this post - I got something else for you

EDIT: http://isiforums.net/f/showthread.php/21982-CPU-Performance-Windows-Performance-Plan-Stuff?p=309954#post309954

 

Ok thank you.

 

I7 920
Gigabyte 7990 6GB
X58 UD5
Windows 7
latest rfactor build
Latest AMD drivers

I also enabled 4xAA thru the control panel instead of rfactor 2.

First run at stock CPU speeds 3.1ghz Turbo on. HT on.

2014-09-28 16:51:58 - rFactor2
Frames: 10492 - Time: 182786ms - Avg: 57.400 - Min: 48 - Max: 68

Second run at 3.5ghz Turbo off, with HT off.

2014-09-28 17:03:54 - rFactor2
Frames: 13479 - Time: 199915ms - Avg: 67.424 - Min: 56 - Max: 81

No GPU scaling whatsoever. It runs as if I have a single card. Looking at Stephen Bacchus results. A single 780 should be faster than a single 7970, on top of his 4.3 overclock I'd say nvidia has no SLI scaling either. Of course we knew this from the begining which is beside the point. I don't think multi GPU will ever work on an ISI based sim, both AMD, nvidia, and ISI care less to even comment on it.

I think a replay should be made from the current build though just to eliminate that as a threat.

 

Sorry, forgot to reply. Yep, i'm doing 4xMSAA from the nvidia control panel as well.

 

Post #67 for anyone with suspicions of under-performing CPUs

 

Ok, so with wanting to keep the benchmark short, sweet and simple, i think we should stick to just either no MSAA or 4x MSAA. I vote for 4x MSAA since i would think that most people run with some amount of AA enabled and if so, i reckon 4x MSAA would represent the average.

 

You can run 8x too. It doesn't make any difference in FPS as long as you have at least 2GB of ram on your card, especially running at 1080p.

 

Well, then it shouldn't matter if we use 4x then i guess. I've also spent the last 2 hours benchmarking, don't want to have to repeat if can be avoided too.

 

I think you are confusion anti aliasing with anistropic filtering. Anistropic filtering does indeed make 0 difference wether its on 4x or 16x, anti aliasing is quite another matter

 

I'll post the new bench thread soon. I just need to grab a bite to eat, not eaten in 24 hours.

 

I benched yesterday with 4x and 8x, no FPS difference. I can do it again if you want. Is at higher resolutions is when it starts to effect. 1080p doesn't hurt performance unless you run out of texture memory. Its when you go from multi sampling to Adapative or supersampling Anti-aliasing is when you notice a difference.

Now if we where to do this at 2500x that be a different story.

Regardless, 4x is fine. Especially if you have less than 2GB.

 

And was there a visual difference between 4x and 8x? Because obviously when nothing changes ingame and everything looks exactly the same it's normal that there isn't an FPS difference.
But i'm not all that clear on how the GPU AA levels (2x 4x 8x ...) compare to the RF2 menu AA levels (l1,l2,l3,...).
If for example l1=2x l2=4x l3=16x (and this is completely hypothetical) and you choose 8X in your GPU driver, would it even work when there isn't a RF2 AA mode that mirrors that exact setting?

 

It works like this,

There was a visual difference, but ISI does AA very horribly so it still amounts to having jaggies, just not as pronounced. If you load up MSI afterburner you can see the difference via texture memory usage.

Also the levels work like this,
1=2x, 2=2xEQ, 3=4x, 4=4EQ, 5=8x, 6=8xEQ, 7=16x, 8=16EQ.

Depends on what your card supports. That's why you see not everyone has the same amount of levels in game. You can force 16x too but the quality doesn't change but the engine uses the GPU as if it where 16x with high GPU utilization and texture memory usage.

 

Gotta laugh Ripper, all this because you got 17% percent performance lost when you can already run the sim maxed out for it to matter anyway.

I suppose it could matter in a full field of stock cars on a super speedway nose to tail.