My friend and I are building a rig for rFactor2. Please can you fill me in on the current state of SLI support in rFactor2? Our current plan is to go GTX970 for 1080p and then add another in SLI for 3 x 1080p a bit down the line. If SLI support is still very bad, it might be better to go 960 for 1080p now and then replace with a 980TI (or equivalent) when we go 3 x 1080p Thanks
You will be just fine with 2x 970, I'm having those and running 3x 1080 just fine. Now with multiview and hdr (which wasn't possible in previous builds), all max settings except some reflections shadows, because I'm using high antialiasing for nicest picture. It never goes below 60, where I have it locked. Some unoptimized third party tracks could give you some fps drops tho.
I'm pretty sure SLI is still broken in 2D mode (SLI works brilliantly with Nvidia 3D Vision). Go with a 960 now and a 980Ti/TitanX or R9 390/390x down the line.
SLI isn't half as broken in "2D" mode as some say I have 2 GTX 560's (I know, not exactly powerhouses) and running all settings on medium (aside of track & environment reflections) I got roughly 45 FPS running in single GPU mode, whilst with SLI enabled I get well over 60 FPS
cards these days are gd enough to not go sli and still get good fps just make sure you have 4 gig card. spend $$$ somewhere else or save them for an update in the future. 900 series so you can run the rift if you decide to go that way instead of triples. i get over 60 fps with a gtx 680 4gig card.
Test the game with absolute maxed-out graphics settings, and make sure not to use any high-quality forms of AA (eg. SGSSAA, "xS" combined modes, supersampling, transparency supersampling, etc.) which may give a false impression of better SLI scaling. Maybe even do the test with.RF2 set to night time as well as rain for an even more graphically demanding scenario from the game's perspective. I'd love to know you're benchmark results (single GPU VS SLI).
It's not broken, there was bad perfomance when used with multiview but that problem is gone now so it works fine with multiview enabled. Without multiview even better.
Quite a range of opinions! We will be using 144hz monitors, we're happy to turn down the details to get over 120FPS. Spinelli - Please would you be able to do a quick test with your 780TIs in 2D? Something like a fraps report on a replay of a lap. I would be so grateful, you have exactly the kind of system that we'll be setting up.
I've done many benchmarks before, and the SLI was definitely bad with single screen and triple screens (w/ multiview of course, non-multiview doesn't exist in my mind ). Again, the exception is Nvidia 3D Vision where SLI would perform brilliantly even with triple screens, multiview, full reflection settings, HDR, etc. http://isiforums.net/f/showthread.php/21360-RFactor-2-Benchmark-780-Ti-1x-2x-GPU-(SLI)-1x-3x-monitors-(Surround)-2-3D-Vision Remember, that's an old build but I don't think much has changed as I frequently do tests. SLI scaling results for above "Maxed Graphics" benchmark (mid-August 2015) 2D Single Screen - Avg: Negative - Min: Negative Triple Screen (w/ multivew) - Avg: Negative - Min: Negative 3D (3D Vision) Single Screen - Avg: 48% - Min: 68% Triple Screen (w/ multivew) - Avg: 72% - Min: 88% P.S. It's the same in GSCE (check the few GSCE benchmarks included in each graph). The 3D SLI framerates (once doubled) are almost 2x the framerate of the 2D SLI benchmarks. Or in other words, the 3D SLI framerates are almost the same - per eye - as the 2D SLI framerates rather than being about only half or so. [HR][/HR] April 15, 2015 Benchmark - Nvidia driver v347.88 - B946, yes? - Newly updated Malaysia circuit, night, rain, full/max graphics, single screen 2D 1x GPU: - Avg: 71 - Min: 41 - Max: 111 2x GPU (SLI): - Avg: 52 - Min: 36 - Max: 69 SLI Scaling: Avg: Negative - Min: Negative - Max: Negative 3D (3D Vision) 1x GPU: - Avg: 91 - Min: 51 - Max: 142 2x GPU (SLI): - Avg: 137 - Min: 86 - Max: 197 SLI Scaling: Avg: 51% - Min: 69% - Max: 39% Notes: - The SLI scaling in 3D mode - when using a single screen - is only 51% (avg fps) and 69% (min) - respectable, still definitely worth it, but nothing great. Also, nothing has changed as the SLI scaling results in my old benchmark are almost identical - Avg. 48%, Min. 68%. HOWEVER, notice the fantastic SLI scaling results in the triple screen (w/ multiview) benchmark results - Avg. 72%, Min. 88%. - The 3D framerates are normally reported per eye rather than total, I therefore multiplied the FPS by 2 for these 3D results in order to show the raw power of what the card is truly rendering in order to give a much better representation and comparison of performance (3D FPS have not been doubled in my old benchmark). - I did each benchmark 3 times, all results were extremely consistent - within 1 - 1.5 fps
A lot of work has been done on Multiview/HDR performance in the last few builds. This perf boost isn't associated specifically with SLI though. Anyone running Multiview/HDR will see these optimizations even on a single card. With these optimizations it puts Multiview in reach for slower single cards now. Reading around here it seems like reflections is what really kills SLI scalability.
Ya, check out my old "Reduced Graphics" benchmark in my link. It was the only scenario where there was any sort of postitive 2D SLI scaling, and even then, it was only in single screen mode (but that may have changed due to multiview optimizations).
Spinelli - Thank you! Your testing is so helpful, exactly what I needed to know. So for now, especially with 3 monitors, it looks like SLI is a waste of time. Looks like we'll go for a single card setup.
For what it's worth, a 970 GTX can run triples in multiview, but you have to turn down AA & you won't be getting 120 fps. Go with a single 980GTX would be my advice if you want the best shot at 120 fps. With multiview disabled, the 970 GTX runs triples very nicely without turning down any settings. Can't remember if it's 120 fps nice, but it's close.
With new Nvidia driver in Inspector I noticed rFactor2.exe in SLi compatability bits. Can't recall seeing that before.
Hi, you wouldn't mind posting some shots of your graphics settings and Nvidia inspector profile. I spent some time fiddling last night to arrive at the conclusion that SLI still doesn't work on my rig. I ended up setting the SLI compatability bits to 0x00000000, as I've had set before as the official Nvidia profile has always given me catastrophic results, and setting all the GPU/SLI options to one card, force single card AFR etc. This has probably made things as good as I've ever had them. Was testing in the DW12 at Silverstone in wet and dry conditions with 29AI and getting around 40FPS at the start rising to 60+ after the first couple of corners. This was with close to maxed out settings: environment/road reflections at low, shawdow blur optimal, shadows high, track detail high, max 30 cars visable, everything else maxed, 5900*1080, multiview on, HDR on, AA level 3+FXAA.
Hmmm, I'm having some different settings, using AA through NVinspector (4xMSAA + Supersampling), and frame rate limiter at 58fps which also reduces input lag very noticable. The other in-game video settings are pretty much same, except I'm not using shadow blur. I've tried the same session, Silverstone + DW12 + 29AI In the garage I'm having around 70-75 fps, and driving outside pit on the track I'm having 95-115 fps, mostly staying around 105, except when passing near pits, it goes to around 85-90. NVinspector settings http://i.imgur.com/cvhsX2o.jpg http://i.imgur.com/2H0YvG1.jpg Video settings http://i.imgur.com/EulZpFo.jpg http://i.imgur.com/k2WdKGA.jpg Session Settings http://i.imgur.com/SBFnp7P.jpg Maybe you have messed up something with NVinpector and compatibility bits..
Wow, thanks for your settings Prodigy - they have made an incredible difference Minus the frame rate limit/force vsync in practice I was getting 100+ FPS in the pit garage and over 170 FPS down the hanger straight. At race start it was around 70FPS but once the first few corners were out of the way it was soon back to 90+ and well over once the field had spread out some more. Given this I went back and played around with some more settings. Road and environment reflections are the killer, having either of these set to high immediately tanked my FPS to 30. Even shadows at ultra wasn't much of a hit. I have since settled on the following which gives around 50 FPS at race start, climbing to over 60 once the first few corners are out of the way. Back to running adaptive vsync, which I prefer to eliminate tearing. It's not quite running as well as it should, still a little stutter at times with lots of cars around you even with the in game FPS meter pegged at 60 but otherwise smooth. View attachment 16732 View attachment 16733 View attachment 16734 View attachment 16735 Here's a little video I made to show the FPS because if someone told me they were getting this sort of performance I wouldn't believe them. This was done with your original settings. I had to set the res back to 5760*1080 for this so the counter doesn't get obscured by the bezels, if anyone was wondering why the tyres don't line up.
I'm glad I've helped Yes, reflections are killer, I've set them as low because from cockpit view they don't really matter much, but it's still there on other cars, rather then completely off. And MSAA4x + SuperSampling gives nice picture, maybe not as nice like MSAA4x + SGSSAA4x, but this combination is more FPS hungry. FXAA I've never really used because it tends to blur the picture.. As for V-Sync, I've tried lot off combinations until found that V-Sync ON + Framerate Limiter at 58hz gives me best solution, smooth, no tearing and reduced input lag comparing to Vsync alone. If I set only frame limiter, it doesn't work as good as combined with v-sync. But I guess that also depends on peoples monitors or tv's, so each to it's own. Enjoy your RF2 my friend
