DX11/DX9 low fps/stuttering with SLI setup and multiview

Full dx11 from the first day of May, no sli (780gtx@5760x1200), and never used rf2 since!

 

I use no post effect and dx11 is still an major visual upgrade, with any post effect on, old content even if working, do not look good
(from my point of view). When swapping between dx9 and dx11, the later is the only option.

To bad you can't get it to work properly, as to my understanding, with same details, aa level and no post effects normally similar fps levels seems to be obtained (or in some cases, even higher).

 

@Wergilius Unfortunately the background stuttering are not even a matter of content, pp or fps; with or without pp, my fps is steady @60... it's just time to fix the fullscreen!

 

Werg,
Have you tried with only 1 of your GTX1070 ? It can handle your currrent tripple screen.
If you have not, do try it and you probably see improvement.
Your screen resolution is similar to mine, but I run GTX1080Ti x 2.
For now, SLI is not working in RF2 the way we expect SLI to at least scale decently well.
So me for now use only 1 GTX1080Ti.

Let us know how the result is...

 

Unfortunately the asus strix 1070 only have 2 dp ports so I must use 2 cards to connect my monitors (3x2560x1440, 144Hz gsync).
I tried initially (not with rf2) to use 2 dp and 1 hdmi, but then you are limited to 60Hz and it did not run well at all on AC that I first
tested out the setup with.

I hope the SLI scaling will be better in the near future, for me it works ok at least with an rather high end pc, but further on when more content get adapted to dx11 it sure must improve to perform on an decent level as you enable post effect and so on.

 

Ohh that sucks with only 2 of DP connector. I use Dell 2716DG 144hz Nvidia G-Sync capable, but I capped it to 120 fps.
When I shop for the GTX1080Ti, first thing I do was to make sure 3 of DP ports available.
It seems DP is a more popular choice for fast monitors today.

 

Yes, i did not plan an triple setup and sli when I bought the first card and was not aware of the limits for different ports, so it is what it is

DP is the way to go, current hdmi is 60hz and dvi is 120hz, but no cards have triple dvi and new monitors do not support it as well.

 

Hello friends, I'm Michael and I love to play gaming on my Asus VG248QE gaming monitor. I like my monitor very much and I provides unique, breathtaking and amazing gaming experience. You can also try this.

 

Not satisfied with the performance I upgraded my computer (again), I have now 1080ti SLI and also an new MB with 80mm SLI spacing and 2x16 PCI slots.

The current compatibility bits:



My current in-game setting:



I have limited the fps to 90 in the json file.
Tested with the new GT3 and the radical with 15 AI car race, the fps stays close or at 90fps at all times even when close racing.

So right now it is working perfect, I just love this this sim, content may sometimes not be as polish as others but for sure is the closest to the real thing.

For the interested, many claims 2x16 do not improve performance (in tests), I made these test in Assetto Corsa using the internal bench mark (using same gfx settings) so it does seems to help in other sims as well:

2x1070, 2x8 PCI
FPS: AVG=86 MIN=58 MAX=177 VARIANCE=0 CPU=37%

2x1080ti, 2x8 PCI
FPS: AVG=118 MIN=26 MAX=217 VARIANCE=7 CPU=51%

2x1080ti 2x16 PCI
FPS: AVG=157 MIN=15 MAX=259 VARIANCE=32 CPU=74%


What I have noticed in rF2, that I do not understand why and sure looks like some performance bug or something. If I select any other setting than Low for Shadows, the fps in general drops but also as soon I get close to the pack, everything gets very slow. Even if fps says 60fps it is visual like 10fps. As soon the distance is like 5m or more, it goes back to be ok. Until I learn that low/quality work I used other settings for shadows as the cars was bad connected to the ground otherwise.

It is also not only rF2 having performance problems, with above in AC I still had stuttering issues (how is that possible??), changing some settings the performance got these values and now it runs very smooth, locked at 90fps:
FPS: AVG=192 MIN=21 MAX=292 VARIANCE=31 CPU=81%

(all test done at resolution 7680x1440, 3x27" g-sync monitors)

 

Shadows are CPU heavy in rF2, stuttering when close to AI is indicator of CPU running out of real time. It may be that your CPU is not running at full power or it's not strong enough. Especially if you run with triple screens or VR, CPU usage increases from rendering to many views. Regarding PCI, yes 16x is absolutely essential for rF2, I think one GPU @ 16x is better than SLI @ 8x.

 

I'm running an i7-7700k OC 14% (4.8 GHz something) so I for sure hope it would do the job. The visual lag is over an rather "long" time so I will log the load and/or use the cpu load indicator ingame to see if I close to any limit.

But I think (hope) there is another reason since my system should be able to manage an 15 car Ai race.

 

Just forget CPU, i7-7700k don`t bottleneck in rFactor 2.
When you are using Nvdia SLI never edit Nvidia Inspector settings if you don`t know what you are doing. Messing with Antialiasing settings gives pure performance and and bad looking game in rFactor 2.
Maximum pre-rendered frames = 1 > single card, 2 > two-way SLI, etc.

Use settings below and I think you get maximum GPU loads with your SLI.

 

I'll try your AA settings to see if it make any difference, my Nvidia profile is standard except for SLI bits (dx11) and power management. I have previously tested with both 1 (using now) and 2 pre-render frames and it doesn't seem to make any big impact in my configuration, it did for sure not affected the close range shadow "slow motion" when very close to other cars.

I should mention that with my settings it runs really smooth now (no post effects used, not sure it works ok with all content I use yet so have not bother tested it for a while) and looks ok after I change to quality shadow blur. But if an Mid settings or better could be used, that would lift the visual even higher

 

If you edit SLI comparability bits (DX10+DX11) to default 0x00000000 and edit Maximum pre-rendered frames to 1 your GPU loads should be something like card 1 99%, card 2 15% and with Maximum pre-rendered frames 2 GPU loads should be 99% with both cards. If your system don`t work like that there is something wrong with your drivers.

 

Thank you, that is an good test, I have noted the "one card high load bad performance" with default profile, drivers could for sure be one part of the problem.

 

Download MSI afterburner to monitor your GPU loads:
http://www.guru3d.com/files-details/msi-afterburner-beta-download.html

Remember to disable SLI and triples before new driver install and use DDU to make clean install:
https://forums.geforce.com/default/...or-new-cpu-core-analyzer-updated-12-03-2017-/

 

When you are using VSync, it is mandatory to enable triple buffering in the nV driver panel !

 

To check.

But triple buffering is not for direct x but for open gl.

At best it does not impact latency or performance, at worst though.

For months I have disabled, and vsync works perfectly. I use the vsync of rf2 "video" I think.

 

Got a link to backup your claim or is it just based on hearsay?
I my observation, triple buffering appears to be working in DirectX as well, resulting in better and more consistent framerates.

[edit]
Just did some searching about the DX vs. OGL issue and it seems, you are right.
At least, I found quite a few reports about the triple buffering option in nV's CP affecting OGL apps only.
I guess, I'll have to verify for myself again...
[/edit]

 

if i remember correct, triple buffering is valid for vsync only?

i'm using gsync (vsync off) so that setting should have no impact.

i tested only to change to 2 pre-rendering frames yesterday but did not notice any big changes, my other settings was already as suggested. However, i was busy racing but maybe today I will do an more systematic test with dx11bits to find the best setting.