Live Performance Benchmarking Comparison for rFactor 2

And your correct on this, I did a few more test with it on. The 8 extra FPS average is gone. I ran 4 tests so it was just off. Perhaps I hit F11 a bit early or late and it skewed the average.

 

AMD users need to download a program called "Radeon Pro" (every AMD user should have this anyway, it's AMD's NVidia Inspector, it can even inject SMAA and Sweet FX into games ) and then they need to set Flip Queue Size to "1". Flip Queue Size = Max Pre-Rendered Frames.


According to the Radeon Pro Manual:

Homepage --> http://www.radeonpro.info/
VERY good user guide that explains almost every setting, pics included --> http://www.radeonpro.info/manual/

 

Clean install: Frames: 9766 - Time: 68000ms - Avg: 143.618 - Min: 117 - Max: 164
Core: 1278 Mhz max 1265 Mhz min
Ram: 1752 Mhz

Checked with GPUz

You min is what seems low for some reason. Is your card down clocking cause of heat? Thermal limit is set to 80, was getting close to that in benchmarks.

 

Thanks for confirming it's not the install.

 

Example results:

CPU: i5-4670K @ 4.4GHz
rFactor 2 Build: 860
Graphics driver version: 335.23

GPU: EVGA 780Ti (manual overclock) - 1240 core/1871 mem (+220/+240)

Time: 67601ms - Min: 111 - Max: 177 - Avg: 145.604

 

TechAde - do you have triples? I'd be curious what your framerate is with the settings from this test + multi view! That would really influence my decision whether to purchase a GTX 980 or not

 

Not currently set up but let me see what I can do.

Sent from my SM-G900F using Tapatalk

 

I see nobody is listening to what am i saying about possible speed/performance hacks in the drivers. That was always a topic and never will be 100% clear what is going on behind the scenes.

Why did you think a special benchmark tool such as the Futuremark and others use standard settings without AA and AF on a standard test, while even than there is no prove that there is not any cheating with the driver going on, especially when new products reaching the market.

Don't use a tool, don't use AF/AA and even than it doesn't excludes cheating. I'm just saying but this benchmark is useless. Use Firestrike and compare GPU's, than look what rf2 does with the differnet GPU's on raw settings, and than go and blame the GPU manufacturers or ISI, but this way you never will find out who is responsible for anything



CPU @4000 Mhz - GPU standard:

1.- Frames, Time (ms), Min, Max, Avg
9216, 67377, 111, 158, 136.783

2.- Frames, Time (ms), Min, Max, Avg
9270, 67361, 112, 159, 137.617

CPU @4000 - GPU OC 1050Mhz:

1.- Frames, Time (ms), Min, Max, Avg
10033, 67658, 121, 171, 148.290

2.- Frames, Time (ms), Min, Max, Avg
9990, 67267, 121, 171, 148.513

 

New AMD driver here.

Going to download and see if there's a difference.

http://forums.guru3d.com/showthread.php?t=393371

 

Speed I'm going to disable AA and AF and do 2 more test with them on and off before I update the driver, then do it again just to get the difference between them all.

 

Ok, then make sure both setups (NVIDIA and AMD)...

A. are using max pre-rendered frames / flip queue size @ "1" (AMD users need to download Radeon Pro for this, see a few posts above, post #42)
B. Have texture quality set to max quality (NVida, set texture quality to "High Quality" and in AMD also do the same and set A.I. Catalyst to ("Off"/"None")
C. Have all texture optimizations disabled (also done in control panel for both games)
D. Disable "Surface Format Optimizations" for AMD users

There, if everyone does the above 3 then that should technicwlly give the most even, fair and "apples-to-apples" comparison possible between NVIDIA and AMD.

After that, using 16xAF and just regular 4xMSAA (or preferably 8xMSAA, more GPU load, lower framerates, less chance of CPU bottlenecks - we should really be using 8xMSAA) will be as apples-to-apples as possible.

Done

I will get pics that show all these settings in NVidia Control Panel, NVidia inspector, AMD Catalyst Control Centre and Radeon Pro in the next hour or so. Then hopefully from there we can finally solve this once and for all and get a long-lasting base benchmark setup.

 

No homie that doesn't works and isn't that easy, than you have to compare image quality also and that isn't the case as this benchmarks are done. You are blind believing the settings and numbers would generate the same image quality for the same cost of performance that way, but i don't don't. Transfer G live to industry managment and you will understand what am i talking about. Raw power is the only way to compare without any filtering and special effects. After that you can start to search in details, everything else is unprofessionell.

A simple example for you, two different cars, similar setup, same tire size but different tire manufacturer = result different in detail, apart from cheating and the general difference of the cars !

 

Single and dual card benches have been run!

CPU: i7-4930K overlocked to 4.3GHz
rFactor 2 Build: 860
Graphics driver version: 344.11
GPUs: 2x GTX 780 Overclocked (1130 GPU1/1120 GPU2 boost)
Full loop custom water cooling


1920*1080 - Single
Min 109 | Max 160 | Avg 137.976

1920*1080 - SLI
Min 120 | Max 177 | Avg 152.567 (+10.1% min +10.6% avg)

------------------------------------------

5900*1080/Multiview/HDR - single
Min 52 | Max 76 | Avg 66.057

5900*1080/Multiview/HDR - SLI
Min 48 | Max 69 | Avg 60.655 (-8.3% min -8.9% avg)

----------------------------------------

5900*1080/Multiview/HDR/Fov31 Texture sharpening 0 (my normal settings) - single
Min 49 | Max 74 | Avg 63.631

5900*1080/Multiview/HDR/Fov31 Texture sharpening 0 (my normal settings) - SLI
Min 47 | Max 72 | Avg 61.398 (-4.3% min -3.6% avg)


Yet again clear negative scaling with triple screen setup. ISI, your customers have shown time and time again negative scaling with high end multi GPU setups. It's 2014 we shouldn't have to put up with this. You keep saying that DX11 offers you nothing that you can't do in DX9. Well I beg to differ because I've never seen negative scaling like this in any of the DX11 games I own. If you really believe DX9 is fine then please work with AMD and Nvidia for the good of your customers to sort this mess out. Thanks Stephen.

 

Here you go with some more OC on the non X 290 GPU which is still air cooled without any voltage modification.

CPU @4000Mhz - GPU@1075Mhz - MEM@1400Mhz - 14.9 driver:

1.- Frames, Time (ms), Min, Max, Avg
10178, 67626, 122, 174, 150.504

2.- Frames, Time (ms), Min, Max, Avg
10142, 67377, 122, 174, 150.526

 

............and now i like to see more realistic benchmarks please, driven by some solid logics.

 

But those settings I listed in post #52 will at least brings things as equal as possible between AMD and NVidia, and on top of that - regardless of any other back-end AA/AF optimizations/differences - almos every single user will use 16xAF (or at least 8XAF) and AA regardless of GPU brand. AA and AF is too relevant in this comparison to just fully dismiss them because AMD and NVidia may do them slightly different.

 

No homie, i can't agree, the world is everything but not true and honest. I'm willing to help the developer but we aren't at this way.

Let me see some firestrike results of differnet GPU's, which is much more believable to me as any amateur benchmark, to beginn with. This actually is just a di.k measurement in my opinion.

 

I mean this in the most respectful way so don't get me wrong. I'm very glad to be getting decent performance for a single monitor. Its when I try triple monitors is where it kills me.

This is the first time I've never been able to solve such lopsided support in a piece of hardware. For years I've cured or at least band-aid crossfire support in a game with a combination of modified drivers and radeon pro. I actually like the challenge I've learn a lot digging deep. But I shouldn't have to change over to nvidia card just because ISI and AMD can't come together.

I7 920 @ 3.5GHZ
Gigabyte AMD 7990 6GB @ 1000 core/1500 mem
AMD Catalyst Driver 14.9

This is with no AA or AF

2014-09-29 17:33:32 - rFactor2
Frames: 9836 - Time: 68094ms - Avg: 144.447 - Min: 115 - Max: 165

This is with 4xAA and 16xAF

2014-09-29 17:40:14 - rFactor2
Frames: 8364 - Time: 67689ms - Avg: 123.565 - Min: 101 - Max: 141

This is with 8xAA and 16xAF
2014-09-29 17:50:06 - rFactor2
Frames: 8173 - Time: 67673ms - Avg: 120.772 - Min: 98 - Max: 140

I used the newest driver out today but its based off the driver I originally had dated aug 27th as they both give me identical FPS so the No AA/AF vs 4xAA/16xAF is irrelevant.
The September 9th driver is actually newer, but it has crossfire issues with Open gl so I'll wait for them to fix that in another driver release.

Using radeon pro had no effect on the fps either. Flipped queuing had no effect at all. However I like how I can increase the LOD of the textures though makes the tracks look crisp.

BTW Far cry 3 has max frame rendering adjustment built into the game which works for either card. Changing it had no effect there either.

 

Forget it my friend this test will never be anyone deliver weaken and strengthen in detail and no developer will benefit from, regardless of whether if it is the manufacturer of the gpu's or the software developer in the case of ISI.

First it needs to provide a basis for, and to start anywhere, such as to use a professionell benchmark software, than it needs to go step by step trough the settings, features and whatever the software offers to show off where something is to improve. Or what is the reason to do this and put all the effort in this, i can't see a solid logic, just kids playing around with a pc.