Perfect SLI Scaling w/ rFactor 2 - Great Job ISI!

[Spinelli]

Settings and PC Info
- ISI NSX and Portugal, 12:00 AM, Rain, no opponents, pit garage
- GPU: 2x GTX 970 (SLI)
- Nvidia Driver: v353.30
- Nvidia Profile: All default except using "prefer maximum performance" and max pre-rendered frames @ 1.

NOTE: The overall framerates for these tests aren't nearly as high as could be due to the tests being performed while my GPU, RAM, and CPU were at the following slow default speeds:
- GPU @ 1315/7000 MHz core/memory (instead of 1500/7500 MHz)
- CPU @ 3.6 GHz (instead of 4.5 GHz)
- RAM @ 16-16-16-39 2T timings (instead of 11-11-11-26 1T)

TEST #1

Default Compatibility Bits - 0x02402005 (Sanctum, UAZ Racing 4x4, rFactor 2[rFactor2 Mod Mode.exe], etc.)
With the default SLI comp. bits I had some real weird results...
- Cockpit camera, both GPUs around 55% usage and poor SLI scaling and framerates
- TV cam, same thing, both GPUs around 55% usage and around 75 fps
- Chase cam (or any other cam like the cam near the wheels, side panel, etc.) and the framerates rise to 145 fps with both GPUs showing 98% usage. Wow!
- Then I go back to TV cam and, unlike the first time, the GPU usage stays at 98% and framerates in the 140s

Basically cockpit cam always SLI scaled poorly, any other cam SLI scaled great, and TV cam SLI scaled poor if the previous cam was cockpit but SLI scaled great if the previous cam was a cam that SLI scales great (e.g. chase cam)...Weird.


Using SLI Compatibility Bits: 0x02D04005
I just tried the SLI compatibility bits that I've seen others mention as of RF2 b982 called "0x02D04005 (LEGO Batman 2, LEGO Pirates of the Caribbean: The Video Game, LEGO Lord of the Rings, LEGO Star Wars III: The Clone Wars, LEGO Harry Potter: Years 1-4, LEGO Star Wars Saga, LEGO Harry Potter: Years 5-7)" and I'm happy to report that all issues are now gone and the GPU usage is at 98% for both GPUs and no matter which view is used including the cockpit view where I'm getting over 130 fps

TEST #2 Both Reflections @ "High" - everything else same as above

Default SLI Compatibility Bits
- GPU Usage = 55% (both GPUs)
- Framerate = 60 fps

Note: Always poor SLI scaling no matter the view

SLI Compatibility Bits 0x02D04005 (LEGO Batman 2, etc.)
- GPU usage = 98% (both GPUs)
- Framerate = 120 fps

Note: Wow! Each GPU's usage stays in the mid-to-high 90s % no matter what view is used and framerates are doubled. FANTASTIC!



TEST #3 - Max Graphics - except FXAA (disabled) and AA (Lvl 2)

Default Compatibility SLI Bits
- GPU Usage = 55% (both GPUs)
- Framerate = 42 fps

Note: Always poor SLI scaling no matter the view, not only that but there was major graphics corruption (solid coloured pink stuff) on the outer monitors.

SLI Compatibility Bits 0x02D04005 (LEGO Batman 2, etc.)

- GPU Usage = 98% (both GPUs)
- Framerate = 88 fps

Note: Even with all graphics settings maxed-out, SLI scaling remains pretty-much perfect (and no matter the view). The graphics corruption was also gone. FANTASTIC!



TEST #4 - Older Content
Same gfx settings as Test #3 from my OP (all graphics @ max except FXAA disabled and AA @ Lvl 2)
- Historic Eve F3
- Historic Spa
- In garage
- Night time
- Rain

Results
1x GPU = 60 fps
2x GPU (SLI) w/ default RF2 comp. bits = 52 fps
2x GPU (SLI) w/ "Lego" comp. bits = 103 fps



Conclusion: Until Nvidia sets it as default for rFactor 2, try SLI Compatibility Bits 0x02D04005 (LEGO Batman 2, etc.). The SLI experience, for me, goes from "broken" to great!

- Check out my positive Stock Car Extreme results here

 

[Spinelli]

Again, it is not my claims. There is a thread with 275 posts from many people all doing the same tests proving that rFactor 2, for whatever reason (I have no idea why, it's really strange), loses a lot of framerates when not using PCI-E 3.0 @ 16x. I don't understand what your problem is.

The PCI-E thing has nothing to do with SLI, those are two different things we're discussing and you're combining them. The PCI-E thing with RF2 happens with 1x GPU, 2x GPU, 1x screen, 3x screen, etc. SLI or no SLI has nothing to do with it.

 

[Ari Antero]

Again, it is not my claims. There is a thread with 275 posts from many people all doing the same tests proving that rFactor 2, for whatever reason (I have no idea why, it's really strange), loses a lot of framerates when not using PCI-E 3.0 @ 16x. I don't understand what your problem is.

The PCI-E thing has nothing to do with SLI, those are two different things we're discussing and you're combining them. The PCI-E thing with RF2 happens with 1x GPU, 2x GPU, 1x screen, 3x screen, etc. SLI or no SLI has nothing to do with it.

Cut out the crap man. Stop talking about that old and hilarious post 275 from ISI forums. .It is not me who is discussing about PCI-E 3.0 @ 16x. (single card) it is you who are doing that. All my posts are about SLI and not about single card PCI-E 3.0 @ 16x.
i7-5930k> Two way SLI GTX 980Ti > PCI-E 3.0 @ x16 x16 vs PCI-E 3.0 @ x8 x8 gives same fps in rFactor2. There is nothing mysterious with rFactor2 it is like any other game with PCI-E 3.0 @ x16 x16 vs PCI-E 3.0 @ x8 x8.

My problem is you bold statement: "Great Jobb ISI!!!" when in reality ISI has done almost nothing to improve rFactor2 SLI performance. NVidia profile (rFactor2 Mod Mode.exe, rFactor2.exe) is created 2011-09-22 and that there is no updated working SLI profile after all these years (+4 years). SLI Compatibility Bits 0x02D04005 is otta56`s finding not ISI`s. There is no guarantee that 0x02D04005 (LEGO Batman 2, etc) is going to work with next build. Every time when new build is released are users forced to spend hours to figure out which SLI Compatibility Bits is going to work this time and if there is SLI Compatibility Bits which is going to work at all. Only a person who is completely loyal to ISI regardless of if they suck or not can make statement: Good Work ISI!!!

The end.

 

[Spinelli]

If I was so loyal to ISI then why would I (and so many others) expose that it looses so many frames-per-second when not running PCI-E 3.0 @ 16x? If I was some loyal fanboy I would say the opposite and pretend RF2 is perfect no matter what.

A) Regarding the RF2 & PCI-E Anomaly (regardless of SLI or no SLI)
You are correct when you say the PCI-E thread is not current. Maybe the anomoly doesn't exist anymore but up until recently, the PCI-E anomoly definitely existed for a long time. If the PCI-E anomaly doesn't happen anymore then that's great. I will resurrect that thread in order to do new tests. Hopefully you are correct so people without PCI-E 3.0 @ 16x don't loose a lot of performance.

B) Regarding the RF2 & PCI-E Anomaly Specifically with SLI
It was no different with SLI, the PCI-E anomoly still happened as can be seen in the tests done in the thread linked to. Again though, if the issue doesn't happen anymore then that's great. We need to redo tests with b982.

C) Regarding RF2 SLI Performance/Scaling (regardless of any PCI-E anomalies there may or may not be)
ISI stated they made some improvements from their side in their coding as of b982. This very same build is where we finally find an SLI bits that works very good for RF2. It works for most systems, single screen, triples, HDR, both reflection options at full, etc. etc. Before b982 and ISI's code updates, there was no SLI bits that offered 70-100 % scaling improvements over 1 GPU, but instead it was usually negative scaling or very small improvements (e.g 10 or 15 % which I would hardly call an improvement) unless you used 3D Vision where SLI always worked perfectly (without even changing the default SLI bits).

The LEGO SLI bits has been around for a long time, therefore the RF2-SLI situation has massively improved almost certainly because of ISI's work (RF2 SLI code updates as of b982).

 

[Ari Antero]

If I was so loyal to ISI then why would I (and so many others) expose that it looses so many frames-per-second when not running when not running PCI-E 3.0 @ 16x? it was not possible? If I was some loyal fanboy I would say the opposite and pretend RF2 is perfect no matter what.

Post 275 ISI forums is old and when rFactor2 SLI was totally broken and It was only possible to use SLI with same performance as single card if you was able to greate new SLI profile (single card) from the scratch to you specified hardware. It is also common knowledge that PCI-E 3.0 @ 16x single card vs PCI-E 3.0 @ 8x single card that PCI-E 3.0 @ 16x is big improvement. I rely don`t see any reason to test PCI-E 3.0 @ 16x single card vs PCI-E 3.0 @ 8x sinle card and this is why I claim that post 275 is nonsense.
No you are not a fanboy in my mind. I know that you like to help people, you are a good guy.

A) Regarding the RF2 & PCI-E Anomaly (regardless of SLI or no SLI)
You are correct when you say the PCI-E thread is not current. Maybe the anomoly doesn't exist anymore but up until recently, the PCI-E anomoly definitely existed for a long time. If the PCI-E anomaly doesn't happen anymore then that's great. I will resurrect that thread in order to do new tests. Hopefully you are correct so people without PCI-E 3.0 @ 16x don't loose a lot of performance.

I can confirm to you that my PC i7-5930k> Two way SLI GTX 980Ti >rFactor2 >PCI-E 3.0 @ x16 x16 vs PCI-E 3.0 @ x8 x8 have equal performance.
BF4 and GTA V> PCI-E 3.0 @ x16 x16 vs PCI-E 3.0 @ x8 x8. > PCI-E 3.0 @ x16 x16 + 15-20 fps.

C) Regarding RF2 SLI Performance/Scaling (regardless of any PCI-E anomalies there may or may not be)
ISI stated they made some improvements from their side in their coding as of b982. This very same build is where we finally find an SLI bits that works very good for RF2. It works for most systems, single screen, triples, HDR, both reflection options at full, etc. etc. Before b982 and ISI's code updates, there was no SLI bits that offered 70-100 % scaling improvements over 1 GPU, but instead it was usually negative scaling or very small improvements (e.g 10 or 15 % which I would hardly call an improvement) unless you used 3D Vision where SLI always worked perfectly (without even changing the default SLI bits).

The LEGO SLI bits has been around for a long time, therefore the RF2-SLI situation has massively improved almost certainly because of ISI's work (RF2 SLI code updates as of b982).

Yes, I agree with you that ISI has done lot of improvements but the problem with ISI is that they release half done work all the time, like HDR, reflection options, triples (multiview UI) also 80% of all content etc. is not updated.
About the SLI Compatibility Bits 0x02D04005 (LEGO, etc). SLI profiles are much more then only working SLI Compatibility Bits. On the bottom of Nvidia inspector rFactor2.exe profile is Undefined settings which are more important then SLI Compatibility Bits. Undefined settings can only be done by NVIDIA and to make this to happen must ISI submit the latest build of the rF2 executables to Nvidia so that an updated working SLI profile can be created and if not, there is no guarantee that 0x02D04005 (LEGO Batman 2, etc) is going to work with next build. Every time when new build is released are users forced to spend hours to figure out which SLI Compatibility Bits is going to work this time and if there is SLI Compatibility Bits which is going to work at all. It is ISI and their communication and relationship with NVIDIA which is the reason to rFactor2 SLI problems.

PS. Turn on the heat in ISI forums about this and support this link : http://isiforums.net/f/showthread.php/24906-Create-A-New-Nvidia-SLI-Profile-For-Multi-GPU-Setups

 

[Skinny]

Ari I know what you are referring to with regards to performance (in general) with PCIE x16 v x8 etc etc as I too have watched and read many a test demonstrating the minimal performance increase between the two - both in single GPU and SLI/Xfire configs. My motherboard is only PCIE 2.0 yet my SLI configuration will let me play pCARS, Assetto Corsa, Battlefield 4 at 5760x1080 (or 5910x1080 bezel corrected) with all settings either maxed or on high at the very least and pull strong FPS (always above 60). When I run benchmarks on my rig, I score very well and often out perform similar setups so I know there is no underlying strangulation within my system. The same core rig with SLI'd GTX670's frequently out benchmarked the "pro" tests from many sites.

rF2 however (and I'm guessing other games based off Gmotor2) just hates my setup. Yes I am currently on Windows 10 and am aware of the current Nvidia drivers issues, but I had the exact same performance issues with rF2 back in Windows 7 through many driver versions. SLI on a single 1080 HDTV was fine but as soon as I moved to the triples, the game became unplayable.

Using default NVidia settings, I get a huge 20FPS in game (which I've seen others with SLI'd Titans also getting). I've tried every Inspector config you can imagine and the best I can get with full graphics is low 50's on the start with 10 AI's - well down on Spinelli's results. The only way I can achieve playable FPS in rF2 at the moment is by turning off Multiview.

Both Spinelli and I run the same core GPU i.e. GTX970's. Yes he has a much newer spec i7 but since my i5 never seems to get overly stressed I'd say it aint a bottleneck in my setup. I've got 16GB of good quality DDR3 RAM, a couple of good SSD's and a good 1000W PSU so sure that is all fine, the only major difference I can see is indeed the motherboards - and here we are back to the PCIE bandwidth issue

As I posted earlier, I'm planning an upgrade later this year anyway - the young un wants a PC rather than his laptop to play his Minecraft on so figured I'd treat myself and donate my spare parts I've always been a believer in getting a good mobo so why not spend a few extra quid and get one that has dual PCIE x16 lanes? Who knows, maybe by Christmas rF2 might be usable for me again

Sooo, Ari, Spinelli - once again, thank you both for your input on this thread towards my issues. Not sure you guys are ever gonna see eye to eye on this performance issue overall though Might just be one to leave in disagreement...

One thing I could say is that it's 2015, and no matter what the game, we shouldn't be needing programmes like Inspector to make any current game playable with what is becoming a very common setup i.e. multi GPU and multi monitor. No harm in tweaking stuff, but shouldn't be needed to simply make it work

 

[Ari Antero]

Hi Paul.

I had the same experience as you with my system when I tested NVIDIA drivers WHQL 353.52 and WHQL 355.60. Are you 100% sure that you are using WHQL 353.30?
With NVIDIA drivers WHQL 353.52 and WHQL 355.60 are only default SLI profiles working it is not posible to edit SLI Compatibility Bits at all, not only with rFactor2 but in all games. NVIDIA Inspector simply does not work with WHQL 353.52 and WHQL 355.60 in windows 10.
If you are using WHQL 353.52 or WHQL 355.60 use Display Driver Uninstaller (DDU) and scroll you system back to WHQL 353.30.
Display Driver Uninstaller (DDU): https://forums.geforce.com/default/...or-new-cpu-core-analyzer-updated-07-10-2015-/

 

[Skinny]

I'll have a check later tonight. Inspector must be working with the later drivers though as changing settings can have a drastic effect on my FPS. Default Nvidia gives me 20FPS and changing to Lego SLI bits bumps it up to 50 odd. That was with the 353.62 drivers anyway.

 

[Ari Antero]

Well, many has problem with 353.62 and 355.60.

https://forums.geforce.com/default/...eady-display-driver-feedback-thread-8-13-15-/

 

[Skinny]

Ok so did a full uninstall with DDU and installed 353.30 drivers and exactly the same issues experienced. Stock Nvidia settings = 20FPS, few Inspector tweaks achieves around the 50 mark again. Checked other games and they still play perfectly.

 

[Andrea Ventura]

I have the same problem since I upgraded to Windows10. All other games run smoothly :-(

 

[Spinelli]

Get rid of Windows 10 . It's so new, you're pretty much an alpha/beta tester.

 

[Ari Antero]

NVIDIA has huge problems at the moment with drivers, not only windows 10 drivers but to all OS. It is not my finding that 353.62 and 355.60 is broken and that 353.30 the best driver at the moment. Goeforce forums are full of rapports about it.
All other games run smoothly and If you guys are wondering why rFactor2 don`t it because NVidia profile (rFactor2 Mod Mode.exe, rFactor2.exe) is created 2011-09-22 and when ISI updates gmotor2, additional SLI optimizations etc. in 2015 and never submit the latest build of the rF2 executables to Nvidia so that an updated working SLI profile can be created it is ISI and their communication and relationship with NVIDIA which is the reason to rFactor2 SLI problems.

Turn on the heat in ISI forums and support this link :http://isiforums.net/f/showthread.php/24906-Create-A-New-Nvidia-SLI-Profile-For-Multi-GPU-Setups
OR
Get rid of rFactor2 it is more then 4 years old but you're still pretty much an alpha/beta tester.

 

[Skinny]

Well I would bin Windows 10 if it used to work good in Windows 7 but alas I had the same results in that OS too

 

[Andrea Ventura]

Rolled back to 353.30 and got the fps back.

 

[Ari Antero]

I think you guys should support WhiteShadows thread in ISI forums and Quote Ott56`s request in Nvidia forum.
http://isiforums.net/f/showthread.php/26985-New-NVIDIA-SLI-profile
I did it

 

[Skinny]

Well time to resurrect an oldish thread

So I finally got my upgrade sorted:

Asus X99-A
i7 5820K Haswell-E OC'd to 4.4GHz
16GB DDR4 Corsair Vengeance LPX 3000Mhz

So I hooked up the rest of my gear (as per previous posts), fresh install of Windows 10, latest Nvidia drivers and BANG - 30-40FPS lol - exactly where I was with the old rig. Tried the Lego SLI bits and that bumped me up to 70-80FPS alone on track but down to low 50's on the start line with a good few AI chums. Tried a few more tweaks here and there but little difference.

So I removed the latest drivers with DDU and installed 353.30 then re-applied the Lego SLI bits and WOOHOO, a lovely solid 80-120FPS in a pack of 20 AI cars!

Just wondering if others are still on the 353.30 drivers for solid SLI performance? Ran a benchmark in Dirt Rally and the earlier drivers don't seem to have affected performance at all in that game (aint tried AC and pCARS as yet).

Whilst I'm on, does anyone else play rF1 or Euro Truck Simulator 2 at all? With my old setup, rF1 ran like **** with anything other than 0x00000000 for SLI bits but nothing I tried got decent FPS in ETS2. Any suggestions welcomed

 

[Emery]

http://isiforums.net/f/showthread.php/26985-New-NVIDIA-SLI-profile

As suggested by Prodigy, Whiteshadow and Spinelli are testing 0x02F00403 with the latest driver 359.06. Whiteshadow confirmed that it works as well as Lego bits with the older 353.30.

Forthright suggests 0x00500005 (Windows System Assessment Tool), but is single screen rather than triples.

 

[Skinny]

Cheers for the link, I'll keep an eye on that thread

 

[Spinelli]

http://isiforums.net/f/showthread.php/26985-New-NVIDIA-SLI-profile

As suggested by Prodigy, Whiteshadow and Spinelli are testing 0x02F00403 with the latest driver 359.06. Whiteshadow confirmed that it works as well as Lego bits with the older 353.30.

Forthright suggests 0x00500005 (Windows System Assessment Tool), but is single screen rather than triples.

I'm going to be doing benchmark tests soon of 353.30 and the very latest NV driver. I'll probably only do the benchmark in triple screens this time and with one type of GFX settings since I don't have as much time. I'll do tests in single GPU and SLI, and as-well as 2D and 3D. I'll probably only use sims that have true triple screen support like RF1/SCE, RF2, and AC, so probably no PCars and R3E.

Going to do the tests right now. Look for results later or tomorrow.

 

[Ari Antero]

rFactor2 has same bad performance with latest Nvidia drivers as before. Game works ok with 353.30 driver but Nvidia 3D Vision don`t work.