[Official] DX9 vs DX11 performance difference

[Daniele Vidimari]

Indycar - 16 AI @ Nola
Starting lap, starting position: 8th
Same conditions & settings
DX9 AA level 8
DX11 AA level 5
DX11 PP none

Vsync: both Video

AMD FX6300 - R9 270X - 8GB Ram - Win 8.1

 

[Mulero]

Track: Nola
Car: Indycar
Resolution: 1920 * 1080, All to Max, 35 FOV
AA: Level 8 in DX9, Level 5 in DX11 (no PP)
Conditions: 1 full lap from the finish line to the finish line.
Build ID: 1802039
Aggressive Threading OFF

 

[whatdahello]

I want to give insight, where the AI is jumping half a car forward-backward randomly. (like when there's lag in multiplayer)
It can happen with and without the agressive threading.

Both of the performance logs are from the start of the race. [And maybe there's a little stuttering in the second log as well, but maybe not too serious]
DX11 Atlanta Sunny
DX11 Howston G4 10 AI

SPECS:
i5 3570
8gb ram
GTX 750Ti 2GB

Win7
NVidia Driver 381.89
1280x1024 res.

 

[4thworld]

Track: Indianapolis Oval
Car: Indycar
Resolution: 6010 * 1080
Ingame Settings: All to Max
AA DX9: Level 1
AA DX11: Level 1 no PP
Conditions: 1 full lap from the finish line to the finish line.

DX9 BuildID: 1796904
DX11 BuildID: 1802039

https://docs.google.com/spreadsheets/d/1KWcaI0qUE8NoSO1N772bw_sIE_KcpnlrfC6WU2-QW7M/edit?usp=sharing

I am surprised about my DX11 benchmark because even that the FPS is compareable to DX9, the game feels so much worser in terms of performance. My Steering Input is 2 or 3 tenth of the second delayed even my FFB feels a couple of tenth off. In DX9 my input is instant. My FFB is spot on and if I turn my steering wheel the game reacts immediatly. In DX11 everything feels more like "Slow-Motion".

My Computer:
Nvidia Driver 378.92
| CPU: Intel Core i7 3770K @ 4,2GHz | 1 GPU: GeForce® GTX 1070 G1 Gaming 8G
| RAM: 2x8GB @ 1600 MHz| OS: Win 10 Home 64bit Version 1703 (Build 15063.250)
| Sound: Creative Soundblaster Recon3D
| Wheel: Fanatec Porsche 911 GT3 RS V2 | Pedals: Clubsport Pedals V1 | Shifter: Thrustmaster TH8A
| Monitor: Samsung SyncMaster BX2331 + 2x Samsung SyncMaster S23B300 (Triple Screen 6010x1080)

Can you try turning Aggressive Threading on in the config?

 

[4thworld]

@Mulero we just released another cpu-focused update, give it a shot whenever you can

 

[Mulero]

@4thworld Clouds are having a big impact on DX11 so I've repeated the tests with sunny weather. No improvement in the latest build.

Track: Nola
Car: Indycar
Resolution: 1920 * 1080, All to Max, 35 FOV
AA: Level 8 in DX9, Level 5 in DX11 (no PP)
Conditions: 1 full lap from the finish line to the finish line.
Build ID: 1802039 & 1802266

 

[qusimano]

Performance it's improved again, Thanks, anyway i still getting more fps in external view than in cockpit, this is strange right? In Nola i can Get 90+ Fps from trackside cam, and no more then 60 down to 54 in cockpit view,if i disable mirror it goes up to 70 (USF2000)

 

[4thworld]

we saw from your reports that there is a slow down rendering clouds, we're working on it.
Maybe that's the reason why you're slower in cockpit view.
Thanks for the feedback again, but let's try to keep the thread structured with the benchmark.
You can edit your post and notify us.

thanks again!

 

[Rui Santos]

In Assetto Corsa (DX11) external view has more FPS than cockpit, guess it's normal because there are more details on screen in cockpit...

 

[4thworld]

In this specific thread the most important thing is to keep in mind the comparison between DX9 and DX11, if your DX11 FPS are < than the DX9 post here, we'll keep update until we reach a point where performance in DX11 are at the same level or better than DX9.
If your FPS is >= than dx11, then this is not the right place to post.

Again : let's keep this structured as instructed in the first post!
thanks again.

 

[gor]

Check ReadMe.txt inside zip

 

[Mulero]

@4thworld Small improvement with the last Build. Much more fluid, FPS are half that in DX9.

Track: Nola
Car: Indycar
Resolution: 1920 * 1080, All to Max, 35 FOV
AA: Level 8 in DX9, Level 5 in DX11 (no PP)
Conditions: 1 full lap from the finish line to the finish line.
Weather: Sunny
Build ID: 1803126

 

[demerzel]

Here is my DX11 performance for the latest build. I already sent my DX9 performance. I used the same settings.

 

[Ferrari64]

Can you try turning Aggressive Threading on in the config?

Yes, I did it. Tbh, I couldn't tell a difference, Still a big gap between me turning the wheel and the game noticing it. On top of that the FPS got a little bit worse and even with over 130 FPS I had stuttering. Experience is not really smooth at the moment. If you want me to do a second full Benchmark run with Agrressive Threading enabled, just let me know

 

[corsa]

Intel i7–6700HQ 2.7 GHz - 16 GB RAM
AMD Radeon Pro 455 - 2GB VRAM
AMD Drivers 17.4.3

Track: Nola
Car: USF2000
No AI
Clear weather
2 laps from end of pit lane

Build 1803126

Resolution: 1920 x 1080
Gfx settings: All to max except Shadow Blur at Optimal

DX9 AA Level 8
DX11 AA Level 5, PP None

DX9: Average 71.278 fps
DX11, Aggressive Threads Off: Average 24.994 fps
DX11, Aggressive Threads On: Average 21.882 fps
DX11, Aggressive Threads Off, Shadows reduced from Max to Low: Average 56.311 fps

Benchmarks and Performance Logs attached

 

[BoothJoe]

Today's runs:
I74770 24ram Sapphire RX480 8ram

My previous DX 9 on NOLA with USF2000 was around 140+ FPS. Today with DX 11 I get upwards of 70 with it sunny/clear (see perflog...clear). When I make it overcast, I lose half that FPS (see perflog...overcast). AA set to 5 post is None. I'm the only car on the track. Eyefinity triples 5900x1080.

 

[BoothJoe]

I forgot to mention that the recorder seems to turn off on its own after a short time for me so I can't even get a lap's worth of data.

 

[BlackGoku]

Hello, after the updates of 5 May in the DX11 I thought that I was going to get more FPS but my game is still unplayable and has ridiculous FPS compared with DX9.

GPU: R9 280 3 GB VRAM
CPU: A6-3650 overclocked to 3,9 ghz
RAM: 12 GB
OS: Windows 10 Pro x64
Controls: Logitech G27

Track: Indianapolis - Oval
Car: Dallara DW12

Here I post my LOG recorded with the CTRL + F and Shift + Space:

https://www.dropbox.com/s/s3bjqbfw5fntx6o/PerfLog_2017-5-6-1-59-12-.txt?dl=0

In DX9 I had 60 fps with V-Sync activated in full graphics with the same track and car, but after switching to DX11 I am getting 16 fps minimum and 20 maximum with the same settings, full graphics + None in the Post Processing Effects and disabled Agressive Threading. The game is totally unplayable. I am getting the same fps from the first DX11 build.

I MUST SAY that I could notice that switching to external camera (Delete key) gave me +30 fps, something maybe will help you to fix it. I say this because maybe you can see in this log that I had those 30 fps improvement and I wanted to say why it happened.

Here I post my settings in game:

http://i.imgur.com/8v1GDI7.jpg
http://i.imgur.com/Mt3rC7I.jpg

As you can see, the game says I have -1043 MB of VRAM (WTF?) and 11 GB of RAM instead of the 12 GB I have and DX9 dettects it.

Finally, I must say this is not a CPU related issue, because I know is not the best but as I already said I was having 100 fps full graphics with DX9 and 60 fps with v-sync on, and I also can play Assetto Corsa 24 cars ultra 60 fps and Project CARS high 50-60 fps.

 

[suryaprihadi]

Track NOLA C. Car : GTR500 2013.
MAX setting, but in DX11 AA is L4 ( not L5 ), to match DX9, so it will be apple to apple DX9 vs DX11.
No Post Processing in DX11 as instructed.

Test are : Single 2560x1440 and multiview angled 7680x1440
For single 2560x1440, a single GPU and SLI 2 way GTX980 test.

NOTE : Since RF2 DX11 is new, Nvidia driver 378.92 I use is the default DX11 SLI bits 0X000000 , which is weird, because it does not allow SLI ( 2nd GPU load max 3% only and Nvidia SLI Indicator not running ) if resolution is 2560x1440 or 4K, ........ but SLI will work in 7680x1440.
So I ended up using DX11 SLI bits 0X80000F5 as reccomended by other members, for single 2560x1440 monitor in 2 way SLI.

I did have problem with DX11 Performance Log txt file. Sometime it cuts off at 40 seconds, sometime it does not create the text file. Sometime it works the whole 2 minutes of the test.
Sometime it produce a tone when LEFT-SHIFT+SPACE pressed, most of the time no tone.

I attached Afterburner/RivaTuner log files and also RF2 DX11 Performance Log, not FRAPS because in each I have prepared a graphic view of 1 complete pit to 1 round of the track. Where in SLI troubleshooting a graphic view is a better* tool ( * I discovered a 2 seconds freeze occuring ) and the very same test can be replicated by me to compare to future update of the DX11. The Afterburner logging is 2 data points per second.

I also have compiled GPU-s Load & FPS DX9 & DX11 in 1 excel file for all the test, to save your viewing time.

Thanks for this effort. Really appreciate it.

 

[Daniel Surace]

Hello, after the updates of 5 May in the DX11 I thought that I was going to get more FPS but my game is still unplayable and has ridiculous FPS compared with DX9.

GPU: R9 280 3 GB VRAM
CPU: A6-3650 overclocked to 3,9 ghz
RAM: 12 GB
OS: Windows 10 Pro x64
Controls: Logitech G27

Track: Indianapolis - Oval
Car: Dallara DW12

Here I post my LOG recorded with the CTRL + F and Shift + Space:

https://www.dropbox.com/s/s3bjqbfw5fntx6o/PerfLog_2017-5-6-1-59-12-.txt?dl=0

In DX9 I had 60 fps with V-Sync activated in full graphics with the same track and car, but after switching to DX11 I am getting 16 fps minimum and 20 maximum with the same settings, full graphics + None in the Post Processing Effects and disabled Agressive Threading. The game is totally unplayable. I am getting the same fps from the first DX11 build.

I MUST SAY that I could notice that switching to external camera (Delete key) gave me +30 fps, something maybe will help you to fix it. I say this because maybe you can see in this log that I had those 30 fps improvement and I wanted to say why it happened.

Here I post my settings in game:

http://i.imgur.com/8v1GDI7.jpg
http://i.imgur.com/Mt3rC7I.jpg

As you can see, the game says I have -1043 MB of VRAM (WTF?) and 11 GB of RAM instead of the 12 GB I have and DX9 dettects it.

Finally, I must say this is not a CPU related issue, because I know is not the best but as I already said I was having 100 fps full graphics with DX9 and 60 fps with v-sync on, and I also can play Assetto Corsa 24 cars ultra 60 fps and Project CARS high 50-60 fps.

Did you try clear sky instead of overcast? Many people report poor frames with clouds. And I get double frames with clear sky.