GTX 770 - performance lower than expected. Need help from 680/770/780/Titan owners

Discussion in 'Technical Archives' started by Marek Lesniak, Jul 6, 2013.

  1. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,584
    Likes Received:
    101
    First things first, so, here's my rig:
    - i5 2500k @ 4,5GHz
    - 8GB RAM
    - GTX 580 1,5GB VRAM @ stock speeds (772/2000)

    Yesterday I bought a GTX 770 from Gigabyte, the one with Windforce 3 cooler. It's factory overclocked to 1137MHz (+ nVidia Boost).
    As I went through a few reviews from techpowerup, guru3d etc. I was expecting a fps boost of about 45-50%.
    What I only gained, is like +10-12% vs my GTX 580... not a whole lot, right?

    So now I am wondering, is it the current state of rF 2 (older builds, like 90 were giving me expected fps gain when I was testing various nVidia cards) or something on my PC (even though other games have performance as should be on GTX 770).

    Could GTX 680, 770, 780 and Titan owners do some tests for me and report with fps you get?

    A few things to set up first:
    1. nVidia driver 320.49:
    - texture quality - set to "High Quality"
    - Maximum pre-rendered frames - set to "1"
    - Power Management Mode - set to "Prefer maximum performance"
    Everything else set to default (which typically means "Application controlled").
    NOTE: If you use nV Inspector or such, drop its settings to defaults as we don't want it to mess with drivers panel settings!


    2. rF 2 Config settings:
    - resolution 1920x1080
    - no v-sync (obviously ;-) )
    - HDR on
    - FXAA off
    - AA Level 6
    - refresh rate 60Hz (shouldn't matter but just in case)

    3. rF 2 in-game settings:
    Everything at its maximum:
    rF2-240_i52500k_gtx770_disp_settings.jpg

    -----------------------------------------------

    The tests:
    1. Megane at Mills Outer Loop C - with 24AI:
    - 24AI (just the Megane) + you starting from last
    - sunny weather @ noon (12:00pm)
    - standing start after formation lap
    Here's the procedure:
    Switch to race session and allow AI to take control of your car. Once formation lap begin, start Fraps benchmark. Finish when ALL cars take their places on the grid after formation lap (notice left column will be behind you when you stop, so wait for them too!).

    2. Megane at Mills Outer Loop C - no AI:
    - no AI
    - sunny weather @ noon (12:00pm)
    - standing start after formation lap
    The procedure is the same as in pt.1. You just don't wait for any AI (but still, let AI do the formation lap for you).

    3. Modern F2 at Silverstone - no AI:
    - no AI
    - sunny weather @ noon (12:00pm)
    The procedure:
    Let AI take control of your car. Let it drive of the garage and benchmark one whole lap from start/finish line (can be between the old one or the new one - doesn't matter).

    Fraps benchmark reports back min., max. and average fps values but you have to check them first. Those values are what we are looking for. Thanks in advance!

    Will check something and report back with my results.
     
    Last edited by a moderator: Jul 6, 2013
  2. C3PO

    C3PO Registered

    Joined:
    Aug 31, 2012
    Messages:
    1,087
    Likes Received:
    86
  3. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,584
    Likes Received:
    101
    Thanks, I read that already but your tests were different - SLI setup, different resolution and different graphics settings.
     
  4. DurgeDriven

    DurgeDriven Banned

    Joined:
    Mar 20, 2012
    Messages:
    6,320
    Likes Received:
    43
    My specs are right in between yours 2500K@4.5GHz and GTX670
    My previous card was also a GTX570




    You really want to give those settings a miss.

    High env reflections look terrible on windscreens, blotchy mess


    Your settings ( Max shadows included ) Kills Fps, I got barely 60fps off the start line.

    I notice they fixed black FXAA ? It was worse fps with FXAA off now. !?

    I run settings I get 80 fps at Silverstone and it looks just as good.
     
  5. Natureboy

    Natureboy Registered

    Joined:
    Jan 13, 2013
    Messages:
    117
    Likes Received:
    0
    I just went from a evga 660SC to the evga 770SC. Performance gains vary largely from track to track but I see between 50 and 100% more FPS with everything on and max detail at 1440p (only 3x AA, more shows no difference). 770 gives 60-120 FPS, 660 gave 40-90 FPS, both cards being very smooth at any framerate.

    i5 3570k, not OC
    16 Gb RAM
     
  6. Natureboy

    Natureboy Registered

    Joined:
    Jan 13, 2013
    Messages:
    117
    Likes Received:
    0
    Actually I bought the 770 at the same time as purchasing the higher res monitor, expecting a big performance hit. Before at 1080, the 660 ran 50-100 FPS.
     
  7. MJP

    MJP Registered

    Joined:
    Oct 5, 2010
    Messages:
    988
    Likes Received:
    21
    Hope this thread doesn't go the usual way of people posting what they get with their settings rather than what was asked for in the OP.

    Sorry LesiU I don't have one of those high end cards you were hoping for, I'm posting purely for selfish reasons as a GTX770 is really on the limit of what I want to spend on a new card and I want to see how it compares.

    Anyway here's my rig and results:

    - i5 3570k @ 3.6GHz (I need to flash a dodgy beta bios in order to oc ivy 'K' chips on my mobo (Gigabyte suck))
    - 8GB RAM @ 1600MHz
    - GTX 470 1,28GB VRAM @ 650/1674, 285.62 driver (newer drivers don't seem to work as well on my old card)

    Test 1 - MILLS OUTER LOOP C, 24AI, MEGANE
    Avg: 37.570 - Min: 24 - Max: 50

    Test 2 - MILLS OUTER LOOP C, NO AI, MEGANE
    Avg: 58.300 - Min: 44 - Max: 72

    Test 3 - SILVERSTONE GP LAYOUT, NO AI, F2
    Avg: 36.363 - Min: 24 - Max: 57
     
  8. TomF74

    TomF74 Registered

    Joined:
    Jan 10, 2012
    Messages:
    5
    Likes Received:
    0
    i5 2500K@4ghz; 8gb ram; gtx680 def clocks; nv driver: 310.64 /win7 64bit
    test 3
    Frames: 6473 - Time: 112099ms - Avg: 57.744 - Min: 46 - Max: 70

    i hope it helps
    if i ll do the other tests i ll post them too
     
  9. vittorio

    vittorio Registered

    Joined:
    Jan 11, 2012
    Messages:
    1,118
    Likes Received:
    540
    Benchmarked with the exact settings asked by the OP. My GPU is a factory OC GTX670 (performance should be very close to a stock GTX680, ~2% less performance).

    My system:
    CPU: Intel Core i7-3820 @ 3.6 GHz
    Motherboard: ASUS P9X79, X79
    Memory: 16GB DDR3-1600
    OS: Windows 7 Ultimate 64bit
    Video: Gigabyte GeForce GTX 670 OC WindForce 3 2GB
    Driver: 314.22 WHQL

    Benchmarks:

    2D, @3.6GHz, 3 virtual mirrors
    Test 1 - MILLS OUTER LOOP C, 24AI, MEGANE
    Avg: 43.102 - Min: 33 - Max: 55
    Test 2 - MILLS OUTER LOOP C, NO AI, MEGANE
    Avg: 70.324 - Min: 55 - Max: 89
    Test 3 - SILVERSTONE GP LAYOUT, NO AI, F2
    Avg: 63.897 - Min: 49 - Max: 80

    2D, @3.6GHz, no virtual mirrors
    Test 1 - MILLS OUTER LOOP C, 24AI, MEGANE
    Avg: 46.311 - Min: 34 - Max: 59
    Test 2 - MILLS OUTER LOOP C, NO AI, MEGANE
    Avg: 71.344 - Min: 57 - Max: 89
    Test 3 - SILVERSTONE GP LAYOUT, NO AI, F2
    Avg: 66.058 - Min: 52 - Max: 81

    2D, CPU@4.55GHz OC, no virtual mirrors
    Test 1 - MILLS OUTER LOOP C, 24AI, MEGANE
    Avg: 46.321 - Min: 34 - Max: 60
    Test 2 - MILLS OUTER LOOP C, NO AI, MEGANE
    Avg: 71.466 - Min: 57 - Max: 89
    Test 3 - SILVERSTONE GP LAYOUT, NO AI, F2
    Avg: 77.875 - Min: 64 - Max: 90

    -> I benchmarked with 3 virtual mirrors first, but since this wasn't specified in the test I benchmarked again without virtual mirrors. FPS impact isn't much.
    -> To see whether these tests are CPU or GPU limited on my system I overclocked my CPU from 3.6 to 4.55GHz (CPU speedup 26.4%).
    Mills tests are completely GPU limited, Silverstone got an avg speedup of 17.9%.


    The following results are for personal interests since I never race in 2D. All three tests in stereoscopic 3D are GPU limited. Maybe of some interest to others.

    3DVision interleaved, @3.6GHz, no virtual mirrors:
    Test 1 - MILLS OUTER LOOP C, 24AI, MEGANE
    Avg: 27.609 - Min: 19 - Max: 36
    Test 2 - MILLS OUTER LOOP C, NO AI, MEGANE
    Avg: 40.702 - Min: 29 - Max: 54
    Test 3 - SILVERSTONE GP LAYOUT, NO AI, F2
    Avg: 44.201 - Min: 33 - Max: 54

    3DVision interleaved, @4.55GHz OC, no virtual mirrors:
    Test 1 - MILLS OUTER LOOP C, 24AI, MEGANE
    Avg: 27.498 - Min: 19 - Max: 36
    Test 2 - MILLS OUTER LOOP C, NO AI, MEGANE
    Avg: 40.812 - Min: 29 - Max: 54
    Test 3 - SILVERSTONE GP LAYOUT, NO AI, F2
    Avg: 42.178 - Min: 31 - Max: 52

    3DVision interleaved, @4.55GHz OC, no virtual mirrors, AA Level 3 (instead of 6):
    Test 1 - MILLS OUTER LOOP C, 24AI, MEGANE
    Avg: 32.915 - Min: 22 - Max: 43
    Test 2 - MILLS OUTER LOOP C, NO AI, MEGANE
    Avg: 47.062 - Min: 34 - Max: 62
    Test 3 - SILVERSTONE GP LAYOUT, NO AI, F2
    Avg: 55.705 - Min: 43 - Max: 66

    (Since these tests are very well specified, maybe we could use these as standard tests ,with lowerd AA since AA level 6 is very high, and create a new thread to benchmark all kind of systems)
     
    Last edited by a moderator: Jul 7, 2013
  10. o0thx11380o

    o0thx11380o Registered

    Joined:
    Jun 28, 2013
    Messages:
    454
    Likes Received:
    2
    I have a gtx 680 and with all ingame settings at max I get no lower than 35 fps in any car and on any track with up to 10 AI cars.

    Im running a couple year old quad core 3.1 ghz with 12 gigs ram

    Im running antialiasing forced in NvInspector at 4x multisampling and 4x sparse grid supersampling. Vsync set to off. No FXAA. HDR is ON. Resolution set at 1920x1200

    EDIT: sorry I should have read the rest of the post.. disregard this if it doesnt help.
     
  11. DurgeDriven

    DurgeDriven Banned

    Joined:
    Mar 20, 2012
    Messages:
    6,320
    Likes Received:
    43

    THE GUY ANSWERED HAS A GTX770 YOU DONT !@!


    ANYWAY HE DOES NOT HAVE TO DO WHAT HE IS TOLD, HE CAN POST WHAT HE LIKES WITHOUT YOU WHINGING.

    WHO ARE YOU or POSTER TO SAY OTHERWISE !



    ffs
     
  12. Noel Hibbard

    Noel Hibbard Registered

    Joined:
    Oct 5, 2010
    Messages:
    2,744
    Likes Received:
    40
    GTX 770 - performance lower than expected. Need help from 680/770/780/Titan o...

    In rF1 there was a cmd line option +nosound if I recall. Has anyone tried this option in fR2 and if so did they notice any change in framerate? The sound engine in rF2 is basically the same as rF1 and in the past I recall people seeing improvements when disabling the sound.

    Not suggesting this as a solution, just trying to rule it out as a bottleneck.
     
  13. DurgeDriven

    DurgeDriven Banned

    Joined:
    Mar 20, 2012
    Messages:
    6,320
    Likes Received:
    43
    I did a fresh lite folder rFactor2 TEST and installed only test components.
    NV profiles default
    Touched no rF2 settings other then outlined



    My system:
    CPU: Intel Core i5-2500K @ 4.5 GHz
    Motherboard: MSI Z77A-GD65
    Memory: 8GB DDR3-1600
    OS: Windows 7 Home Premium 64bit
    Video: Gigabyte GeForce GTX 670 OC WindForce 3 2GB ( default clocks)
    Driver: 320.49 WHQL

    Benchmarks:


    2D, CPU@4.5GHz OC, no virtual mirrors
    Test 1 - MILLS OUTER LOOP C, 24AI, MEGANE
    Avg: 44.766 - Min: 34 - Max: 63
    Test 2 - MILLS OUTER LOOP C, NO AI, MEGANE
    Avg: 70.691 - Min: 56 - Max: 88
    Test 3 - SILVERSTONE GP LAYOUT, NO AI, F2
    Avg: 64.322 - Min: 50 - Max: 81
     
    Last edited by a moderator: Jul 8, 2013
  14. vittorio

    vittorio Registered

    Joined:
    Jan 11, 2012
    Messages:
    1,118
    Likes Received:
    540
    No one with a higher end GPU like 770/780/Titan and with some time to benchmark? I would be very interested too how these cards perform with rf2.
     
  15. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,584
    Likes Received:
    101
    OK chaps, here are my scores.

    GTX580
    Test 1
    min. 36, max. 65, avg 51.459
    Test 2
    min. 53, max. 83, avg. 67.707
    Test 3
    min. 46, max. 78, avg. 64.278

    ------------------------

    GTX 770
    Test 1
    min. 36, max. 64, avg. 47.572
    Test 2
    min. 55, max. 93, avg. 71.885
    Test 3
    min. 55, max. 87, avg. 71.646


    To sum things up:
    TEST1
    GTX 580: min. 36, max. 65, avg. 51.5
    GTX 770: min. 36, max. 64, avg. 47.6
    + 0.0% - 1.5% - 8.2%

    TEST2
    GTX 580: min. 53, max. 83, avg. 67.7
    GTX 770: min. 55, max. 93, avg. 71.9
    + 3.8% +12.1% + 6.2%

    TEST3
    GTX 580: min. 46, max. 78, avg. 64.3
    GTX 770: min. 55, max. 87, avg. 71.7
    +19.6% +11.5% +11.5%

    The first test is with AI so CPU has the biggest influence on performance, so don't bother the GTX770 seems to be "worse". The difference is neglible.

    In general, the GTX770 looks to be just several % faster than GTX 580.... in other games, we see it beeing better by 40-50% so it looks rF engine still can't get full use of the newest cards. As 7xx (6xx too?) seem to have shader/geometry performance ratio closer to Ati than it was with older nVidia cards, I think that might be the case why it doesn't shine in rFactor.

    I remember, doing such comparison tests between GTX 570 and 580 in rFactor 2 Build 90. So i installed it, brought back the same cars/tracks (the same old versions of them) and did the same set of tests to see, if bad performance for GTX 770 could be just in newest builds.

    All 3 tests are basically the same apart test 3, where I use F3 Spark at Monte Carlo (instead of F2 @ Silverstone).

    GTX580
    Test 1
    min. 57, max. 86, avg. 72.420
    Test 2
    min. 71, max. 109, avg. 91.570
    Test 3
    min. 56, max. 119, avg. 82.841

    ------------------------

    GTX 770
    Test 1
    min. 61, max. 91, avg. 77.757
    Test 2
    min. 79, max. 120, avg. 101.388
    Test 3
    min. 69, max. 126, avg. 93.701

    To sum things up:
    TEST1
    GTX 580: min. 57, max. 86, avg. 72.4
    GTX 770: min. 61, max. 91, avg. 77.8
    + 7.0% + 5.8% + 7.5%

    TEST2
    GTX 580: min. 71, max. 109, avg. 91.6
    GTX 770: min. 79, max. 120, avg. 101.4
    +11.3% +10.1% +10.7%

    TEST3
    GTX 580: min. 56, max. 119, avg. 82.8
    GTX 770: min. 69, max. 126, avg. 93.7
    +23.2% + 5.9% +13.2%


    We see GTX 770 is slightly faster in older build, but still the difference is far from what I was expecting.

    -----------------------------------------------------------------------

    I saw in other threads, there are a few guys with Titans and 680s... too bad they didn't show up here :(
    None the less, thank you to all of you who did those tests for me, so we can compare our results. Thanks guys!

    Looking on those scores, for me - if someone wants to buy a new nVidia card mainly for rFactor 2, forget about it as it's not worth the money... at least not until rFactor 2 will be utilizing newer cards on a level similar to other games.


    PS As for mirrors - for all my tests I was using just the real ones at default seat position.
     
  16. Marek Lesniak

    Marek Lesniak Car Team Staff Member

    Joined:
    Oct 6, 2010
    Messages:
    1,584
    Likes Received:
    101
    To let you know, I overclocked my GTX 580 by 20% on core/shaders and 10% on memory (from 772/2000 to 926/2200) and in general, I get a performance boost of typically 15-16% compared to stock clocks. After OC, it's faster than mentioned GTX 770.

    TEST1
    GTX 580: min. 36, max. 65, avg. 51.5

    580 OC : min. 42, max. 72, avg. 58.2
    GTX 770: min. 36, max. 64, avg. 47.6


    TEST2
    GTX 580: min. 53, max. 83, avg. 67.7

    580 OC : min. 61, max. 96, avg. 78.2
    GTX 770: min. 55, max. 93, avg. 71.9


    TEST3
    GTX 580: min. 46, max. 78, avg. 64.3

    580 OC :min. 48, max. 89, avg. 75.5
    GTX 770: min. 55, max. 87, avg. 71.7

    That also shows, questionable performance of GTX770 is not because of something is holding it down like too slow CPU or whatever... otherwise, I won't be seeing as much gain after OCing the GTX580.
     
  17. DUFFBAG

    DUFFBAG Registered

    Joined:
    Apr 7, 2013
    Messages:
    20
    Likes Received:
    0
    Looking forward to RF2 working like everything else I have installed...and to working on my driving... not my control panel & ISI Forums. 2xGTX660... :( ... Fail.
     
  18. Quo-Fan

    Quo-Fan Registered

    Joined:
    Oct 16, 2010
    Messages:
    55
    Likes Received:
    1
    Nearly the same problems with my GTX670. Picture quality with everything on full/high is bader than RF1 with the same settings.
    Even with tools like Nvidiainspector it looks better, but FPS is going down then.
    Normally, if you choose in Nvidia system settings "Best Quality" the game should look very fine, but this isnĀ“t at this stage.
     
  19. Bart S

    Bart S Member

    Joined:
    Oct 5, 2010
    Messages:
    851
    Likes Received:
    104
    For a start running AA in rFactor 2 is garbage kills performance start overiding in nV inspector turn all AA off in rFactor, turn off FXAA leave HDR on. Special effects and Shadows medium and optimal cos you cant bearly tell any difference to Max. Sun occlusion is an fps eater also I have it off, you wont miss it. Road + Enviro reflections medium, you wont miss high here either. With just these adjustmans you should really notice some fps unlocked.

    As Noel said above I can kind of confirm what he is saying regards sound engine however I run 24 samples max instead of default and I think less samples helps quit alot.

    Why not read my thread about Graphics, its got a bit of Durge babble in there but try to ignore it, this thread has helped many with alot of thankyous http://isiforums.net/f/showthread.php/12633-4x-Better-Graphics-and-way-more-FPS-Guide

    My last post in the silverstone thread will give you an idea of what FPS im getting now after my recent upgrade.
     
  20. Bart S

    Bart S Member

    Joined:
    Oct 5, 2010
    Messages:
    851
    Likes Received:
    104
    LesiU reading your post above seems like you may have a driver problem with your GFX the following guide may help resolve your problem : http://www.overclock.net/t/1150443/how-to-remove-your-nvidia-gpu-drivers Then do a fresh install of latest nVidia Drivers.

    If that still doesnt help I would look into your bios do you have option for PCI-E 3 x16 enabled also do you have your GTX770 in the ffirst slot as your mobo may only have x16 on the first slot.
    GTX770 is pretty much a GTX680 maybe overclocks better and has new TXAA support.

    Also check with your GTX770 manufacturer if bios is of latest you may have to reflash due to added support for older mobos.

    Also check your mobo manufacturer for latest bios.
     

Share This Page