[Official] DX9 vs DX11 performance difference

Discussion in 'Technical & Support' started by 4thworld, May 4, 2017.

Thread Status:
Not open for further replies.
  1. Trevor Greenfield

    Trevor Greenfield Registered

    Joined:
    Apr 19, 2017
    Messages:
    37
    Likes Received:
    11
    I don't know how to see the build #, but the time stamp in the upper right says it was from yesterday. In discord hey said it was a content update. The beta updates thread did not have any information in it for yesterday. Why is it suddenly a drop in fps? I have no idea but I just redid the tests, fresh system reboot, no extra processes running and I come up with the lower numbers. Also, as I said, the GPU usage in cockpit is now maxxing, where it was only at first 68% then 78%. But that did not result in higher fps for me using my everything at minimum test.

    2017-05-12 10:51:19 - rFactor2
    Frames: 4415 - Time: 30000ms - Avg: 147.167 - Min: 135 - Max: 164

    2017-05-15 15:59:17 - rFactor2
    Frames: 3988 - Time: 30000ms - Avg: 132.933 - Min: 120 - Max: 152

    2017-05-16 08:35:51 - rFactor2
    Frames: 3972 - Time: 30000ms - Avg: 132.400 - Min: 120 - Max: 153

    These are my DX9 numbers from my first test
    2017-05-04 15:01:20 - rFactor2
    Frames: 7032 - Time: 30000ms - Avg: 234.400 - Min: 225 - Max: 246

    Look I understand we all have different configurations. A lot of you guys have much better video cards than I do, with more video memory. I would love to see some benchmarks from GPUs/memory the same or worse than mine. My system is definitely video memory limited, but I have dealt with it through DX9 into DX11, and there are certain tracks and configurations in different sims that stress that more than others. For example, iRacing is the best for me in DX11. AC is roughly the same as rfactor2.

    The problem is that for Studio397, if they release, as they have stated, DX11 and phase out DX9, as it stands it would be barely playable on 1080p using low settings, and at 5760x1080 on my triples, using the same test, I get 90fps. With just lvl1 AA that goes to 75fps. That doesn't give me enough overhead to turn any options on like AF, turn up detail, and certainly not enough to watch trackside cameras with even one AI.

    So just on FPS alone right now as it stands if Studio 397 releases DX11 and phases out DX9, it will be unplayable for me unless I buy another video card. To me, that means the minimum to run the game will be more than a r270 2GB, or perhaps it means no 2GB cards will be able to satisfactorily run the game. Can you see how much that limits the customer base? Also, it really struck me as odd how as soon as I joined Discord i was told how "the DX11 beta is already better than DX9 for most of us". Then I read the forums, and that is clearly not the case. Most people have worse performance than DX9, some are the same. Its all over the board frankly. Personally, I feel like if I don't speak up for people in a similar situation as me, this is going to turn into a situation where the devs don't give a shit because they only test using newer more expensive hardware, and their primary beta testing group obviously does not provide low-end data for them to evaluate. I don't see anyone else saying what I am saying so that is why I am saying it.

    Further, there are a number of bugs and issues that stand out. I believe the new UI could fix some of them. But the game currently has a bug that when viewing the standings video fps drops to 12. In fullscreen mode, it halves the fps from standard cockpit. People have mentioned that above 4x AA doesn't do anything.

    Basically, I'm at the point where the devs continue to release betas so I will continue to test and provide feedback in the hopes they can make it better. Nobody has thanked me or asked me to provide continual feedback, I'm doing this because I have owned rfactor 1 since it was released and bought rf2 with the hopes it would be rf1 reincarnated. There's no doubt for those who have spent more money on better video cards, DX11 is a boon, but to say this game in its current state is the same as DX9 for everyone is wrong, and to set the bar higher than a 4 year old video card can provide is cutting your market very shallow.
     

    Attached Files:

  2. Trevor Greenfield

    Trevor Greenfield Registered

    Joined:
    Apr 19, 2017
    Messages:
    37
    Likes Received:
    11
    MASSIVE increase in FPS today 5/16/17. Congrats and a pizza slice owed to whomever unlocked today's improvement.

    2017-05-16 10:03:54 - rFactor2
    Frames: 5663 - Time: 30000ms - Avg: 188.767 - Min: 169 - Max: 214

    Still a little ways off the DX9 target but not that far:
    2017-05-04 15:01:20 - rFactor2
    Frames: 7032 - Time: 30000ms - Avg: 234.400 - Min: 225 - Max: 246

    Also, here is PP testing for the same test:

    2017-05-16 10:03:54 - rFactor2 - NO PP
    Frames: 5663 - Time: 30000ms - Avg: 188.767 - Min: 169 - Max: 214

    2017-05-16 10:55:39 - rFactor2 - PP LOW
    Frames: 5621 - Time: 30000ms - Avg: 187.367 - Min: 168 - Max: 211

    2017-05-16 10:58:13 - rFactor2 - PP MED
    Frames: 3994 - Time: 30000ms - Avg: 133.133 - Min: 123 - Max: 145

    2017-05-16 11:03:05 - rFactor2 - PP HIGH
    Frames: 3783 - Time: 30000ms - Avg: 126.100 - Min: 117 - Max: 137

    2017-05-16 11:06:05 - rFactor2 - PP ULTRA
    Frames: 3164 - Time: 30000ms - Avg: 105.467 - Min: 99 - Max: 113
     

    Attached Files:

    Last edited: May 16, 2017
  3. Mulero

    Mulero Registered

    Joined:
    Jan 16, 2012
    Messages:
    351
    Likes Received:
    153
    No change in performance here although the feeling before seeing the numbers was a big decline. Control + F is affecting more and more to the performance.

    Track: Nola
    Car: Indycar
    Resolution: 1920x1080, 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: 1825438
    Aggressive Threading OFF

    FPS.jpg
     

    Attached Files:

    SPASKIS likes this.
  4. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    Great finding Mulero. I had the same feeling the other day. I had two very different results when I repeated a test which made me suspicious about something interfering.

    However it was too late so I didn't compare fraps benchmark with and without ctrl+f.

    It is clear now that in terms of benchmark ctrl+f should not be used. I am not sure if devs are wiling to have from now on both figures as Mulero provided for latest two builds.

    As I have already said several times, with this continuous updating beta we are not able to check for things related to older (obsolete) builds. Things like the presumably different performance of ctrl+f between build 1825438 and 1818133. I am updating now to the latest beta. I will check how ctrl+f affects my computer and report.

    I granted for 100% sure that the possible negative effect in performance of a log building tool had been checked and minimised until neglectable. Maybe it only happens in some computers and relates to the system has performance in general. However it is clear that if you are developing software for normal computers you should previously test it in normal computers. I hope it is the case and the person/s working in improving dx11 performance are not trying to improve it using super PCs.

    I will finally ask one question to the devs in case they want to answer. Since latest two updates there haven't been so many people reporting fps and perflogs. I understand that when S397 has announced that todays beta should significantly increase performance it is based on those logs. Based on this info, do you expect some of the users increasing performance and others not or it should be a general improvement to everybody motivated by some bug or problem that has been detected thanks to the information provided by reporters?

    Thanks in advance
     
  5. Trevor Greenfield

    Trevor Greenfield Registered

    Joined:
    Apr 19, 2017
    Messages:
    37
    Likes Received:
    11
    I personally have noticed Ctrl-f actually improves my fps by about 2fps according to fraps running at the same time.
     
  6. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    upload_2017-5-16_23-3-22.png
     

    Attached Files:

  7. corsa

    corsa Registered

    Joined:
    Oct 4, 2010
    Messages:
    93
    Likes Received:
    5
    To see the build number, in Steam right-click on your game and go to "Properties", then click the "Local Files" tab, and you will see the build ID at the bottom of the window.
     
    Trevor Greenfield and demerzel like this.
  8. patchedupdemon

    patchedupdemon Registered

    Joined:
    Jan 3, 2017
    Messages:
    1,602
    Likes Received:
    1,151
    Rf2 need to up the minimum requirement specs,can't expect it to run on potatoes for ever
     
    FS7 likes this.
  9. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    If you check my results two posts above, you will see that "apparently" cockpit performance has maintained like previous build around initial beta levels, being the important increase in trackside performance, above levels not seen before in dx11, the most significant change.

    The reason why I quote "apparently" is because I have found something that could ruin the whole benchmark so far. At least regarding trackside performance.

    Before getting to it, I would say that it could be the same issue that Mulero has reported before, although he was blaming ctrl+f as the culprit of a significant fps loss.

    According to my findings Ctrl+F provokes a massive stuttering while it isn't hurting fps values being logged. Although the apparent framerate seems to be 10-15 fps, the values being logged are much higher.

    What I have seen today is that trackside is showing two different behaviors. One with very good fps and another one with a much worse one. If you check the frapslog.txt I uploaded, you will see at the end of the file that there are several runs on Monaco with today's beta. You will see that cockpit runs are all of them pretty much the same while trackside runs show two groups of results. In the graphs I have posted, I have used best result.

    I can't tell for sure what's going on between two cases but it would seem that what is being rendered is simply not the same. Like LOD not properly working or something like that.

    I have been able to reproduce the following behaviour:
    - After loading the track I hit race.
    - I switch to trackside view (AvPag)
    - I activate AI (I) and time accelerate (Ctrl+x) until AI is about to cross the finish line.
    - I start the benchmark with fraps (F11) when AI crosses the FL and end it the same way when the AI finished the lap.
    - I switch to cockpit view (Ins), time accelerate as before and benchmark the lap.
    - I switch back to trackside view time accelerate as before and benchmark the lap.

    After doing this the second trackside benchmark shows considerably worse performance compared to first one. I don't know what could be happening in between.

    I have been already over two hours with it and after checking that both situations could occur in the same outing only by switching cameras I have decided to stop and report. It is more productive to let others know about it and try to figure out altogether.

    Good night
     
  10. Emery

    Emery Registered

    Joined:
    Oct 24, 2010
    Messages:
    3,035
    Likes Received:
    1,654
    Build 1825438 showing minimal improvement over previous two builds. Main thing I notice is GPU at 100% constantly now instead of 99%.

    settings: Level 4 AA, PPE=off, resolution 5760x1080 multiview, no sync, aggressive threads=off
    system: single GTX 1080, 8 GB VRAM; i7-4770 @3.4 GHz, 32 GB RAM
    circuit: Atlanta Motorsports Park (AMP)
    car: AC Cobra
    view: cockpit camera
    weather: overcast

    DX9, Avg 88 fps, Min 75 fps, Max 101 fps
    Build 1813052, Avg 85 fps (-4%), Min 75 fps (+0%), Max 97 fps (-4%)
    Build 1818133, Avg 85 fps (-4%), Min 74 fps (-1%), Max 97 fps (-4%)
    Build 1825438, Avg 86 fps (-3%), Min 75 fps (+0%), Max 98 fps (-3%)
     

    Attached Files:

  11. Emery

    Emery Registered

    Joined:
    Oct 24, 2010
    Messages:
    3,035
    Likes Received:
    1,654
    Just because we have Monaco added as updated content this time...

    settings: Level 4 AA, PPE=off, resolution 5760x1080 multiview, no sync, aggressive threads=off
    system: single GTX 1080, 8 GB VRAM; i7-4770 @3.4 GHz, 32 GB RAM
    circuit: Monaco 1966
    car: AC Cobra
    view: cockpit camera
    weather: overcast

    DX9, Avg 94 fps, Min 79 fps, Max 122 fps
    Build 1825438, Avg 92 fps (-2%), Min 76 fps (-4%), Max 119 fps (-2%)
     

    Attached Files:

  12. corsa

    corsa Registered

    Joined:
    Oct 4, 2010
    Messages:
    93
    Likes Received:
    5
    The lastest DX11 build is another step backwards for my performance. To keep things even with all my tests so far, I've stayed on drivers 17.4.3 but am thinking about updating to 17.5.1 to see if there is any difference.

    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, no clouds
    3 laps from end of pit lane

    DX9 Build 1807979
    DX11 Build 1813052 / 1825438

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

    DX9 AA Level 8
    DX11 AA Level 5, PP None

    DX9: Average 71.738 fps

    DX11 Build 1813052, Aggressive Threads Off: Average 25.948 fps
    DX11 Build 1825438, Aggressive Threads Off: Average 22.834

    DX11 Build 1813052, Aggressive Threads Off, Shadows reduced from Max to High, Soft Particles reduced to Low, Shadow Blur from Optimal to Fast: Average 66.240 fps
    DX11 Build 1825438, Aggressive Threads Off, Shadows reduced from Max to High, Soft Particles reduced to Low, Shadow Blur from Optimal to Fast, Road Reflections from High to Low: Average 49.147 fps



    Benchmarks and Performance Logs attached
     

    Attached Files:

  13. stonec

    stonec Registered

    Joined:
    Jun 19, 2012
    Messages:
    3,399
    Likes Received:
    1,488
    The first explanation I thought when I read this is the trackside camera groups. There are 3 or 4 camera groups on most tracks which you can cycle through with page down button. The default camera view is cycling these groups in some sort of random fashion so that you get a different camera view on different laps. So this cycled camera will not measure consistent FPS from lap to lap.

    After latest couple of updates, my FPS is basically within 5% of DX9 version on Monaco, with the added better looks (single screen / 1080p / GTX 660 Ti).
     
  14. knackko

    knackko Registered

    Joined:
    Oct 6, 2010
    Messages:
    142
    Likes Received:
    65
    Quick test yesterday with the last beta build, I didn't make a trace of ctrl+F but before it was difficult with my GTX750Ti to have AAx4 without stuttering (even with a correct FPS), now I can configure AA to x5 and no stuttering ! nice update :)
     
  15. whitmore

    whitmore Registered

    Joined:
    Feb 26, 2012
    Messages:
    552
    Likes Received:
    142
    For me may 16 build it is better. Now better thanDX9 and good enough to start using PP
    Replay remains a problem FPS 14 with bad stuttering.

    As reported using inbuilt CTRL - F hurts frame rate so tested without for best output.

    Test details same as all of my other posted test results
    NIssan GT500 Silverstone 1.6 Max AA No PP May 16 Build - Video card software set to Adaptive Multisampling
    In game settings all max.
    DX9
    Frames 10862 Time (ms) 60000 Min 136 Max 202 Avg 181

    DX11
    Frames 14134 Time (ms) 60000 Min 191 Max 261 Avg 235
     
  16. Jeferson Richart

    Jeferson Richart Registered

    Joined:
    Oct 23, 2013
    Messages:
    174
    Likes Received:
    34
    Best build so far. The stutters I had on some tracks, were gone.
     
  17. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    I already know about different groups within trackside camera. However, what I am describing has nothing to do with it.

    There is something going on that is affecting performance in trackside cameras.
     
  18. Emery

    Emery Registered

    Joined:
    Oct 24, 2010
    Messages:
    3,035
    Likes Received:
    1,654
    Replay monitor definitely stutters while fullscreen replay is fine.
     
  19. SPASKIS

    SPASKIS Registered

    Joined:
    Sep 7, 2011
    Messages:
    3,155
    Likes Received:
    1,426
    I may add that whatever that could be happening in this sense could be unrelated to dx11 itself. I have the suspicion that it could also happen with regular dx9 version or even with latest non steam build 1098.

    If you check my benchmarks for B1098 monaco shows worse results compared to cockpit view. Looking at different builds behaviour in the same graph, trackside (when running smoothest) always beats cockpit view.
    View attachment 7743
    I am 99% sure that if I repeat the benchmark for B1098 trackside Monaco I will get better results than the posted ones.
     
    Emery likes this.
  20. Trevor Greenfield

    Trevor Greenfield Registered

    Joined:
    Apr 19, 2017
    Messages:
    37
    Likes Received:
    11
    Build 1828309
    1 fps increase on avg and min

    2017-05-17 11:11:54 - rFactor2
    Frames: 5682 - Time: 30000ms - Avg: 189.400 - Min: 170 - Max: 214
     
Thread Status:
Not open for further replies.

Share This Page