NEW Benchmark comparison between AMD & Nvidia

Discussion in 'General Discussion' started by John, May 2, 2014.

  1. Tosch

    Tosch Registered

    Joined:
    Oct 5, 2010
    Messages:
    1,872
    Likes Received:
    51
    It is somewhat strange but the bench results I see with the replay file have nothing to do with what I see in a "live" session. Here are my results with 20 AI (Honda) at Silverstone International, all settings at max and myself at grid position 21 letting the AI drive (Ctrl + I).

    Frames: 11973 - Time: 161477ms - Avg: 74.147 - Min: 51 - Max: 96

    After 2 Laps I escaped from the race and made a second try to compare the results (restart race).

    Frames: 8470 - Time: 159293ms - Avg: 53.172 - Min: 39 - Max: 68

    I have no idea why the frame rate is so different. :confused:
     
  2. G.Bird

    G.Bird Registered

    Joined:
    Apr 20, 2013
    Messages:
    52
    Likes Received:
    0
    I tried what Tosch has done and think the same.
    2014-05-03 20:23:22 - rFactor2
    Frames: 8796 - Time: 164488ms - Avg: 53.475 - Min: 38 - Max: 72

    2014-05-03 20:33:39 - rFactor2
    Frames: 9247 - Time: 165876ms - Avg: 55.746 - Min: 39 - Max: 72
     
  3. speed1

    speed1 Banned

    Joined:
    Jul 26, 2012
    Messages:
    1,858
    Likes Received:
    0
    Current setup and the maximum of the CPU 3,2@4ghz, ddr3 1600MHz.

    5760x1080 - level 8 aa - gfx/max - multiview/off - fxaa/off - sync/off

    Min 42, Max 71, Avg 56

    More is not possible and it also makes no difference on fps with level 4 aa, because the result is similar to. :)
     
    Last edited by a moderator: May 3, 2014
  4. Tony H

    Tony H Banned

    Joined:
    Oct 5, 2010
    Messages:
    82
    Likes Received:
    1
    I tried it for u John, with your settings its pulling over 2.1gb of vram at 1920x1080, so not sure how the 2gb cards are performing like that? When I use 3 screens, 5760x1080 it only jumps to 2.7gb...
    my results
    1920x1080
    Frames: 8350 - Time: 157093ms - Avg: 53.153 - Min: 43 - Max: 69

    1920x1080 (windowed)
    Frames: 7928 - Time: 156906ms - Avg: 50.527 - Min: 41 - Max: 65
     
  5. Gupster

    Gupster Registered

    Joined:
    May 19, 2013
    Messages:
    404
    Likes Received:
    25
    Frames, Time (ms), Min, Max, Avg
    7961, 156199, 40, 69, 50.967








    <-----Specs in profile :)
     
    Last edited by a moderator: May 6, 2014
  6. xAcid9

    xAcid9 Registered

    Joined:
    Jan 13, 2012
    Messages:
    73
    Likes Received:
    1
    Fresh PLR, RadeonPro off, default CCC


    GPU at Stock | 850/1250
    2014-05-04 14:02:23 - rFactor2
    Frames: 7846 - Time: 155984ms - Avg: 50.300 - Min: 38 - Max: 65

    GPU Overclocked | 1100/1500
    2014-05-04 14:06:22 - rFactor2
    Frames: 10064 - Time: 157125ms - Avg: 64.051 - Min: 46 - Max: 84
     
  7. xAcid9

    xAcid9 Registered

    Joined:
    Jan 13, 2012
    Messages:
    73
    Likes Received:
    1
    Just for the lulz, i loaded up Catalyst 14.1 and rerun the test with the same settings.

    Fresh PLR, RadeonPro off, Default CCC and cleaned CBASH folder.

    GPU at Stock | 850/1250
    2014-05-04 14:36:18 - rFactor2
    Frames: 3811 - Time: 156031ms - Avg: 24.425 - Min: 17 - Max: 39

    GPU Overclocked | 1100/1500
    2014-05-04 14:40:33 - rFactor2
    Frames: 4205 - Time: 157859ms - Avg: 26.638 - Min: 17 - Max: 43

    Because older driver was bad at handling CPU overhead, overclocking only give marginal performance increased.
     
    Last edited by a moderator: May 4, 2014
  8. John

    John Registered

    Joined:
    Oct 5, 2010
    Messages:
    1,085
    Likes Received:
    113
    That is quite the jump in performance. Double the performance.
     
  9. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,583
    Likes Received:
    101
    I compared my results to yours and they are within reasonable range (your beeing about 6x% better after OC than mine). Question is, are we have something not right, what slow us down a bit, or other guys are faster than should be? ;)
     
  10. xAcid9

    xAcid9 Registered

    Joined:
    Jan 13, 2012
    Messages:
    73
    Likes Received:
    1
    Well, their drivers prior to 14.4 was that bad for rF2.
     
  11. speed1

    speed1 Banned

    Joined:
    Jul 26, 2012
    Messages:
    1,858
    Likes Received:
    0
    Tosch is right, the replay thing is nonsense, because it's not a live session and don't replicates the real performance. I get for example stutter with the replay, it distorts the results only.

    I recommend to set the same terms and conditions of the replay and run the test for 2-3 times. The average should form the result.

    People should try real conditions no records, none of us drives in replays, and thus only the real-time results counts for me. I need to know the performance of the live session, not the replays.

    It could lead to complications with the replays, depending on the system I think.

    In any case I am very satisfied on the result of the old base with the 290 is accelerated.
     
    Last edited by a moderator: May 4, 2014
  12. Ari Antero

    Ari Antero Registered

    Joined:
    Jul 27, 2012
    Messages:
    1,882
    Likes Received:
    825
    I am sorry to hear that you have problems with your GTX 780tTi
    To find out if you CPU is bottle necking download Unigine Valley Benchmark 1.0 here: http://www.guru3d.com/files_details/unigine_valley_benchmark_download.html
    You can look here what others scores with similar system as you have here: http://forums.evga.com/tm.aspx?m=1858694&mpage=31

    Many claims also here that SLI dont work and it is just nonsense, SLI is working fine with rFactor2 and it gives you GPU loads +90% and you get much more high FPS then with any single GPU card.

    I have water cooled PC with Zotac 780GTX SLI what makes me able to push limits and I have done a bit OC with my cards :rolleyes:

    My Valley benchmark score and Valley benchmark settings I used:

    Direct3D11
    Extreme HD
    Mode 1920x1080 8x AA fullscreen

    FPS 122,8
    Score 5242

    Min FPS 40.8
    Max FPS: 189.8

    GPU 1394 MHZ
    Memory 3550 MHZ

    :p
     
    Last edited by a moderator: May 5, 2014
  13. John

    John Registered

    Joined:
    Oct 5, 2010
    Messages:
    1,085
    Likes Received:
    113

    Sure. But THIS bench is a COMPARISON bench between representative GPU's. It is intended purely for interest sake.
     
  14. paarma

    paarma Registered

    Joined:
    Mar 20, 2012
    Messages:
    24
    Likes Received:
    0
    With the orginal benchmark my framerate results were very close between realtime "live" benchmark and replay one.
     
  15. speed1

    speed1 Banned

    Joined:
    Jul 26, 2012
    Messages:
    1,858
    Likes Received:
    0
    Yeah but like I said for me, it's only the real-time performance counts and it seems there are problems with replays.

    However, how is the CCC levels of AA with in game level 8 to be understood ?

    2x - level 1
    2xEQ - level 2
    4x - level 3
    4xEQ - level 4
    8x - level 5
    8xEQ - level 6
    ??? - level 7
    ??? - level 8

    Level 5 looks better and runs better than level 4 here while level 8 does not cost more performance. :confused:
     
  16. SMOK3Y

    SMOK3Y Registered

    Joined:
    Jun 23, 2012
    Messages:
    80
    Likes Received:
    0
    quoting myself lol, is there some benefit to a fresh PLR on every install?
     
  17. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,583
    Likes Received:
    101
    Depends, on what have you changed in PLR compared to the default one :)


    For previous benchmark (Clio @ Loch), done on B342, changes to fps were very minimal between the replay and real time bench. With replay, you always have exactly the same conditions. With real time, differences will mostly be more noticable.

    You can compare scores using replay benchmark and then extrapolate with your real time score to see, what you could have with a different hardware.
     
  18. SMOK3Y

    SMOK3Y Registered

    Joined:
    Jun 23, 2012
    Messages:
    80
    Likes Received:
    0
    just button (CSW) wheel, FOV, seat etc just stuff like that. havent really changed any ini file settings like in rF1
     
  19. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    27
    Even if the replay performance does not reflect the real-time live gameplay performance, it is irrelevant for the purpose of comparative benchmarking and it is not desirable to use live benchmarking over replay because live benchmarking can change wildly in results if you or the AI driving your car ends up behind or in front of the pack which alters the rendering scene which affects the benchmarking results consistency, no?

    I also think that Tosh was refering to live benchmarking results (please correct me if i'm wrong Tosh) when he gave his results in the post your quoted him on. I can see how you might have thought he was inferring they were from replay based on his initial sentence of "It is somewhat strange but the bench results I see with the replay file have nothing to do with what I see in a "live" session."

    Again, doesn't matter if you get stutter in the replay so long as your benchmarking performance is consistent with each repeat.
     
    Last edited by a moderator: May 4, 2014
  20. speed1

    speed1 Banned

    Joined:
    Jul 26, 2012
    Messages:
    1,858
    Likes Received:
    0
    The differences in a short session like the bench in real-time is zero believe me. I did multible benchmarks since the current driver is out, and the most tests were performed 2-3 times, but where is the logic if one has issues to drive the bench right when trying to compare real-time performance as it's the only one important to count. I would accept it if rf2 would have a bench routinely integrated which is optimized and protected for external influences. Replays were never a good basis for reliable tests imho, and to be honest, this comparison lags.

    I'm only doing this to show the developer the current status of their software, and because I had fun. :)
     

Share This Page