NEW Benchmark comparison between AMD & Nvidia

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

  1. speed1

    speed1 Banned

    Joined:
    Jul 26, 2012
    Messages:
    1,858
    Likes Received:
    0
    Only for the correctness and fair comparison. I think a replay is nonsense, there are too many factors may have an influence on the result and only the live performance counts in my opinion.

    Here the current results of an R290 with the 14.7 RC3 Beta driver on a rather old base. The result is always similar, no matter how many runs. Grid position last. Min - 45 Max - 79 Avg - 63
     
  2. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,583
    Likes Received:
    101
    Only replay can give consistent scores. Live benchmark won't guarantee that by design. We already discussed that :)


    @DrR1pper
    680 vs 770 is the same with the same clock. It's the same chip basically.

    Will check CPU utilization and let you know here later. GPU utilization is at 98% (I'm using GPU-Z to monitor that stuff).
     
  3. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    +1


    I don't disagree and stock 680 clock to 770 clock is 6-7% fps increase in games. However i still had a 6-7% increase over yours when we both had the same core clock because my memory was clocked 23% higher. Mem clock performance increase doesn't scale as linearly as core clock i presume.

    Okido. I'm going to do more tests soon...dinner first.
     
  4. speed1

    speed1 Banned

    Joined:
    Jul 26, 2012
    Messages:
    1,858
    Likes Received:
    0
    I don't agree and I wasn't aware of that people play replays. Did we play replays or are we using the sim in real time. Isn't the logic somewhat off an gives people a distorted impression of the current performance of AMD with rf2. I hardly think that's a good promotion for rf2.

    The replay performance isn't the same as the real time performance, just for the record, and you can discuss what you wan't, youre wrong. I'm not intrested in replay performance.
     
  5. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    How is using a replay not the best way to compare card performances? It holds everything constant within the repeatable benchmark. The only reason that would make it bad is if the relative performance ratios between cards in a replay benchmark did not translate to the real time/live gameplay. I don't know of any instances where this has been shown to be true.

    The except to this is probably AI causing possible CPU bottlenecking effect. But since we know that the replay benchmark method doesn't cause this (well, at least when i use "max pre-render frames = 1) it shouldn't be an issue at all.
     
    Last edited by a moderator: Sep 25, 2014
  6. speed1

    speed1 Banned

    Joined:
    Jul 26, 2012
    Messages:
    1,858
    Likes Received:
    0
    It is true. The performance of the replay isn't the same as in real time, and what am i doing with the simulation, driving in real time or just playing around with replays ? It is nonsense to compare replay performance. Just the real time performance counts, that is what matters, nothing else.
     
  7. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    Mate....we all know that. But it's irrelavent for good comparative gpu benchmarking methodology because replay benchmarks are so easily repeatable and the exact rendering scenes are identical with each pass. Live benchmarks are not so easy identically repeatable as well (especially if you add AI into the scene, they take different paths each time and this affects the end result).

    I really think you're missing the whole point of these relative performance benchmark comparisons. It's not to see what the actual performance will be when driving live but simply to compare gpu performance between different cards in rf2. If you could compare thes cards in a live state with exactly the same precision as a replay benchmark, you'd find the same performance ratios between cards as you would with the replay benchmark method. But again, you can't get the same precision and hence why we use the replay benchmark method.
     
  8. speed1

    speed1 Banned

    Joined:
    Jul 26, 2012
    Messages:
    1,858
    Likes Received:
    0
    I can not understand this logic homie because I also do not compare data routines of a car but live performance and those are always different, no matter if the engine or the map is the same.
     
  9. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,583
    Likes Received:
    101
    It goes up and down by a lot. During the benchmark, I saw it to be as low as 20% but also a few times it hit 93%.
    I am checking in Task manager, how much CPU % rfactor2.exe takes (have two active displays). There was no a typical value but most of the time utilisation was between 30 and 73%.
    GPU was 98% all the time.
    I am using driver 344.11.
     
  10. realkman666

    realkman666 Registered

    Joined:
    Jun 28, 2013
    Messages:
    919
    Likes Received:
    0
    So you guys are comparing replays then? Were there serious benchmarks done somewhere else, even if outdated?
     
  11. Prodigy

    Prodigy Registered

    Joined:
    Nov 2, 2012
    Messages:
    948
    Likes Received:
    64
    We need a benchmark that is always the same, consistent and repeatable, that's why we use replay system, it's easy. Making two identical live runs is not easy and never same. Ofcourse the replay is not performing the same as live race, but we are not benchmarking how one card can handle difference between replay and live session, we are benchmarking how more cards are performing on same base. Otherwise you can say same thing for the 3Dmark (probably most popular benchmarking software) that it always use the same scene, or why can't we play with those characters during testing scenes. There's are reason why we can't move them..
     
  12. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    There was and it produced the following graph but i can't seem to find the original thread.

    [​IMG]
     
  13. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    +1

    Better said than i.
     
  14. realkman666

    realkman666 Registered

    Joined:
    Jun 28, 2013
    Messages:
    919
    Likes Received:
    0
    You still end up with data that is useless when you're playing. The quantity of cars and world detail, reflections, etc. is so much higher in replays and jumps around a lot more. I find it ridiculous.
     
  15. realkman666

    realkman666 Registered

    Joined:
    Jun 28, 2013
    Messages:
    919
    Likes Received:
    0
    Holy cow.
     
  16. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,286
    Likes Received:
    29
    lol

    In any case, that benchmark is no good to compare your own results to now because of the much new builds and possibly different gfx effects and performance (not to mention the horrid amd performance at the time). But it's certainly useful as a gauge for nvidia cards.

    Absolutely correct but we want to know how the cards perform to relative to one another in rfactor 2 and if it matches what we see in other games as well. For that, using a replay benchmark is the perfect solution.
     
  17. realkman666

    realkman666 Registered

    Joined:
    Jun 28, 2013
    Messages:
    919
    Likes Received:
    0
    Still, that's scary. xD
     
  18. BlaringFiddle5

    BlaringFiddle5 Registered

    Joined:
    Sep 3, 2012
    Messages:
    427
    Likes Received:
    2
    i7 3770
    16 GB RAM
    EVGA Nvidia GTX 680 2GB
    Nvidia GeForce 340.52
    rFactor2: Build 860

    Without max rendered-frames = 1 (1080p)
    Frames, Time (ms), Min, Max, Avg
    8301, 153723, 43, 72, 53.1000

    with max rendered-frames = 1 (1080p)
    Frames, Time (ms), Min, Max, Avg
    8857, 156906, 45, 75, 56.448

    with max rendered-frames = 1 (5760x1080 - no multiview)
    Frames, Time (ms), Min, Max, Avg
    4822, 156843, 22, 44, 30.744

    with max rendered-frames = 1 (5760x1080 - multiview)
    Frames, Time (ms), Min, Max, Avg
    3462, 156859, 14, 31, 22.071

    Installing 344.11 nvidia drivers and then will test again...

    EDIT: Installed 3.44.11

    with max rendered-frames = 1 (1080p)
    Frames, Time (ms), Min, Max, Avg
    8878, 156906, 45, 75, 56.582

    with max rendered-frames = 1 (5760x1080 - no multiview)
    Frames, Time (ms), Min, Max, Avg
    5097, 168309, 21, 44, 30.284

    with max rendered-frames = 1 (5760x1080 - multiview)
    Frames, Time (ms), Min, Max, Avg
    3477, 156921, 13, 39, 22.158
     
    Last edited by a moderator: Sep 26, 2014
  19. Spinelli

    Spinelli Banned

    Joined:
    Jan 28, 2012
    Messages:
    5,290
    Likes Received:
    31
    We need a brand new chart/graph (damn, I always forget the difference, lol) for the new build. And on page 1 we need a note saying that as of post # xx, results are for rF2 build xxx. Either that or a new thread should be started because we cannot be comparing to old rF2 builds too much data being crossed and causing too much speculation and uncertainty.
     
  20. John

    John Registered

    Joined:
    Oct 5, 2010
    Messages:
    1,085
    Likes Received:
    113
    LOL. Really? Go to the first post of this thread, look at the first line of the first post. See a link and click on it.

    Sorry, just had a chuckle there.

    I would love someone to make a chart using the results in this thread.
     

Share This Page