Live Performance Benchmarking Comparison for rFactor 2

I'm not changing any compatibility bits unless I specifically state so.

When I did change compatibility bits I still had negative SlI scaling, just slightly improved, but still negative.

SimonV6, are you sure your rF2 settings are 100% identical to the settings in the OP? Both reflection options must be on "high" or "full" (can't remember), not "low" or "off". Shadow blur on "quality" and just about everything else on "max" or "full", not "high".

The fact that you only have 30% SLI scaling while others have much, much better SLI scaling (both reflections set to "off") leads me to believe that you shouldn't be using that other compatibility bit, but instead should just be using the default one.

Can you do a test with exactly as the OP says with regards to absolutely everything? So go into Nvidia Inspector and click the reset to profile defaults button, then click apply. Then go into the standard Nvidia control panel and click default for the rFactor 2 gfx settings profile. Then just do the 3 control panel changes listed in the OP in either the Nvidia control panel or Nvidia Inspector (Pre-Rendered frames @ 1, AA to override, AA amount to 4x (4xMSAA).

 

when you get negative, what load is being reported in GPU-Z 0.7.9?

 

For the most part, ranging from low 50s to low-mid 60s on both GPUs. In 3D mode it ranges - for the most part - from mid 80s to mid-high 90s on both GPUs

 

it is very strange that both cards are working at 60% and yet your fps goes down?

 

I don't think it's as simple as "60% x2 is more than 99% x1, therefore 60% x2 should have a higher framerate". In my opinion, if a GPU or 2 is/are only working at 60% each, then something just isn't right, well, unless there's a massive CPU bottleneck - obviously.

 

do you want to try my .nip profile and see if it makes any difference?


the sim does need loads more SLI optimization yet I guess

 

Are you sure you're using the correct settings as you're getting at least 40fps more than other 970 users?

Also your benchmark time is a couple of seconds short, it should be ~67000ms, pit exit line to start/finish line halfway along the straight, *not* the chequered line a bit earlier!

 

yes all correct setting this time, and pit lane to chequered I thought, anyway it wont make any difference I doubt?

setting didn't change from sli @ 254 to single @193, don't ask me why but I have double and triple checked this time

 

Both reflections set to full/high? Shadows to max/full? Circuit detail to max/full?

HDR enabled in the rF2 configure window? Resolution at 1920x1080 in the rF2 configure window?

Entire Nvidia profile back to default (don't adjust individual settings back to default, but make sure to press the default button so that everything goes back to default)? Then, make sure the NVidia global - rather than the specific rF2 profile - gfx options are also set back to fully default. Only then set AA to override and 4xMSAA, and pre-rendered frames 1.

Start time 9AM and the rest of the session settings identical? Same with the A.I. settings?

Are you fully disabling SLI by going into your Nvidia control panel GPU configuration and actually selecting "disable SLI"?

 

oh... I am not using a default .nip file in Nvidia Inspector, I couldn't even get sli to work when I first got these tow cards, so been plying around with that for a while, not much that is standard?

I did put all the setting that was asked for in Nvdia control panel, don't tell me i have to load the default .nip file into Nvidia Inspector now too and do all these damn test again lol

but yes to everything else you asked Spinelli

 

Well, if you're getting 40 fps average more than other people with the exact same card, on a single GPU, then something is surely different.

Can you upoad this magical profile for us? I would love to test it myself and compare it to my single-GPU and SLI results. 40 fps on a single GPU - with all gfx settings remaining identical - is a massive improvement.

 

ok changed the profile in Nvidia Inspector back to default, set x4 and 1 frame in control panel, did all the rest you said and it does drop to 175 in single, but now sli is broken (like it was first off) and I only get gpu load on one card, and 5% on other?

Frames, Time (ms), Min, Max, Avg
11727, 66940, 141, 200, 175.187 single card


Frames, Time (ms), Min, Max, Avg
16583, 67283, 179, 306, 246.466 sli now with "Sli compatibility bit" set to 0x00500005 all rest of .nip set at default

40% improvement in SLI

what texture filtering quality do you guys run this test in?

 

Texture filtering quality in the nvidia control panel? Default. Everything is default except AA at 4xMXSAA and pre-rendered frames @ 1.

With the default inspector profile your avg fps - in single card mode - went down from about 195 to 175. I wonder what settings were changed on that other profile...

 

can you just make sure that default texturing is "performance" as mine is? in control panel, I want to make sure my setting are the same as yours

 

I think i'm losing my edge. Thanks for pointing this out.

You're still using 1300 core (during load?) and 1900 mem Simon?

I can only get about 135fps with those clocks.

Even with your new-found reduction from 193 to 175, that's still 35% more fps than what other 970's can achieve for the same clock.

Something is definitely still not right, ignoring SLI for the moment.

Can you try my player.ini file? http://www.filedropper.com/player

 

seems i was on high performance texturing, now all back to default through NI, and these are what I get?

Frames, Time (ms), Min, Max, Avg
11806, 68126, 140, 198, 173.297 -

single card PCI_E 3.0 x16, profile set back to default in nVidia Inspector, only SLI Compatibility changed to 0x0050005 for SLI
game set at 1920x1080

all setting maxed out
FOV default, in car, mirrors on, wheel on, and HDR on
override AA x4 and x1 frame ahead x16 AF


344.16 drivers, GTX 970 1300/1900 cpu 4200


Frames, Time (ms), Min, Max, Avg
16583, 67283, 179, 306, 246.466 - SLI +42%fps, should be at least 80% IMHO


without SLI Compatibility changed to 0x0050005 SLI was not working, and so scores where almost the same

 

well I am doing my head in here lol,
I have all the game setting just as your pictures, reduced my res to 1920x1080
defaulted my .nip profile in NI and set up in control panel to x4 AA and +1 frames ahead
HDR is on
clocks are set at 1300 core and 1900 memory, i7 5930k is at 4200 turbo, and DDR4 is running at 2666Mhz
I run the test from pit to start line, was using pit to finsih before
not sure what else i can do?

any idea;s as to why i am scoring more in single?



what 3dMark firestrike score you getting with your setup? just to make sure our PC are performing the same

 

Simon, for an apples-to-apples comparison, please follow the instructions in the OP (original post). If you do that then you won't be using "performance" or "high performance" texture filtering because the instructions say to reset all settings back to default.

Forget NVidia Inspector for now. Just go into the Nvidia control panel and reset the rF2 profile back to default, as well as the global profile. Just choose click the reset/default button, rather than manually changing settings yourself back to default.

Also, in Inspector, if you threw out/replaced the stock NVidia profile (I believe it's a .nip file) and replaced it with another base profile, then the defaults will probably not be the true Nvidia-rF2 profile defaults. You need to make sure to reload the original Nvidia-rF2 base profile through Inspector. If you didn't save/back-up the original Nvidia-rF2 base profile, then one of us can upload it for you, or you can re-install your Nvidia drivers.

 

Call me crazy, but do you reckon the DDR4 system ram could be maybe 1 in a million chance causing the difference?

Everything tells me "ofc not" but provided he's done everything to the letter (i.e. software is all correct), it's the only other difference i can see from the hardware perspective.

 

Good idea. Let me check.

At load: 1300mhz core, 1900mhz mem

11,765 GPU score
8851 CPU score
4450 Combined score
9692 Firestrike score

There's a new benchmark available (verson 1.4.775) which i didn't bother updating to.