Massive FPS gains in rf2 using PCI-e 3.0 x16 with higher end cards!

Discussion in 'General Discussion' started by DrR1pper, Sep 30, 2014.

  1. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    In case some of you have not been following the "Live Performance Benchmark..." thread, we had been observing some graphics card performance inconsistencies between different rf2 users. Some users graphics cards (such as my GTX 970) were under-performing in rf2 compared to others with the exact same graphics card at the exact same clock speeds, yet in all other (common) games and benchmarks they were performing normally and as expected again.

    TechAde managed to figure out what was causing this problem. Graphics cards that were under performing in rf2 than they should have been, were also found to be running their graphics card in PCI-e 2.0 x16 (or lower) mode on the motherboard. And when they switched it up to PCI-e 3.0 x16 mode, the under-performance issue was no more and the cards were now performing correctly and to their fullest again.

    This is taken from the start of the OP in the "Live Performance Benchmarking..." thread found here (worth a look as well): http://isiforums.net/f/showthread.php/21983-Live-Performance-Benchmarking-Comparison-for-rFactor-2

    The higher end your graphics card, the worse the performance loss will be of course. What i can tell as a reference is my GTX 970 was bottlenecked by PCI-e 2.0 x16 which reduced performance by 12% using the "live performance benchmarking.." thread gfx settings. If your card performs better than a GTX 970, you will most certainly not want to be running in anything less than PCI-e 3.0 x16 mode. Some cards even less than my GTX 970 will also benefit but atm we're not sure exactly where the cut off point of benefit from using PCI-e 3.0 x16 over PCI-e 3.0 x8 or PCI-e 2.0 x16 is. I hope to have the answer for you shortly on this.

    rFactor 2 is the first game/sim that i've used where i've observed a real benefit from PCI-e 3.0. I suppose it was about time. :p
     
    Last edited by a moderator: Sep 30, 2014
  2. matf1

    matf1 Registered

    Joined:
    Dec 27, 2011
    Messages:
    671
    Likes Received:
    2
    So in our case, the CPU is holding it back as I suspected. Very interesting guys, thanks for your investigations.
     
  3. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    haha, lol. It would be completely fair to view it in that way i guess. But clock speeds, cache, bus speed, etc, are not the reasons at all. In fact, if you're like myself, with a motherboard that supports PCI-e 3.0 but a cpu that does not, then yes, you can blame the cpu. But if you have a cpu and motherboard incapable of PCI-e 3.0, then it's both.

    But be honest, did you actually suspect it could have been the cpu's fault in remotely the way it has been revealed to be the problem? Or did you think more along the lines of cpu speed, cache, bus speed, etc, being the possible bottleneck?

    What's worse is that this connundrum really only occurs with the i5/i7-2000 series but it's made worse by the fact that many people still use this cpu because it was such an amazing cpu when it first came up everyone wanted to get it. The subsequent 3000 and 4000 haven't not been such big improvements to warrant upgrading from the now quite old yet still very capable 2000 series. Unfortunately, it coincides with being also the generation before PCI-e 3.0 support was released by Intel.
     
    Last edited by a moderator: Sep 30, 2014
  4. DurgeDriven

    DurgeDriven Banned

    Joined:
    Mar 20, 2012
    Messages:
    6,320
    Likes Received:
    43
    That is a big leap for single card wow.

    I cant remember 700 series anywhere near that scaling in games ( as the OP rightly pointed out )



    I was happy 3 years on I still have a 2500K@4.5Ghz and GTX670 OC easy handles rF2 @1920 for what I need.



    Was being the operative word ............. :(



    Only joking for me I would not expect to see gains, besides I have in hand 500MHz stable overclock I already don't need would give as much as my 670 could.


    Must be exciting prospect for high resolution multi GPU/monitor users contemplating 900's though
     
    Last edited by a moderator: Sep 30, 2014
  5. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    I hope that with more benchmark results poring and some people willing to test their cards with both PCI-e 3.0 x16 and PCI-e 2.0 x16, we could figure out exactly what card is the cut-off point for rf2 with the benchmark settings. The way to reduce the gpu bottlenecking if a problem with PCI-e 2.0 x16 would be to lower the graphical settings and therefore require less PCI-e bandwidth (i guess). But lower the settings if not the solution ofc, getting a new system would have to be.
     
  6. matf1

    matf1 Registered

    Joined:
    Dec 27, 2011
    Messages:
    671
    Likes Received:
    2
    Yeah motherboard is fine (P8Z68-V Pro Gen3) I knew it wasn't clock speed, but to be honest I've not kept up with technologies as much as in the past, so I couldn't say with any confidence that it was an instruction set or not. I recall though as I'm sure a lot of you do, the Celeron 300a. We all clocked it to 450 and it ran like a dream... Until we popped a Pentium in it a couple of years later and realised how bad it was at that time. Same kind of clock speed but night and day difference in performance, especially with multitasking.
     
  7. matf1

    matf1 Registered

    Joined:
    Dec 27, 2011
    Messages:
    671
    Likes Received:
    2
    Same Durge, I'm finding i'm on the cusp of good performance, most things I want on full and constant 70+ frames. The issue starts to creep in for me with Shadowplay. I love recording live sessions for editing later and you can tell it's giving everything it's got and just getting by. A few more frames (10-15%) and i'd be golden.
     
  8. DurgeDriven

    DurgeDriven Banned

    Joined:
    Mar 20, 2012
    Messages:
    6,320
    Likes Received:
    43
    Pity after all these years all the major review sites can give us is benches in Shift, F1 2013 and Grid and Dirt. ;)

    I would like to see a huge test for all sims rather then 22 games with 21 that I never heard of......... hehehe


    Also sucks a high percentage of sim users opt for multi card/monitor setups yet manufactures snub simulations in drivers updates and scaling.


    I blame all your crappy shooter and wizard games for that. lopl
     
  9. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    You need to ask Spinelli. He could do it for you.

    http://isiforums.net/f/showthread.p...-for-rFactor-2?p=310363&viewfull=1#post310363
     
  10. rogue22

    rogue22 Registered

    Joined:
    Jan 14, 2012
    Messages:
    261
    Likes Received:
    18
    So explain this, I just ran GPUz and it reports I have a PCI E 3.0x card running at 3.0x.
     
  11. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    Did you mean PCI-e 3.0 x3? Are you using the correctly advised PCI-e slot for you graphics card that's recommended in your motherboard hand book? which motherboard is it and which lane are you using?
     
  12. rogue22

    rogue22 Registered

    Joined:
    Jan 14, 2012
    Messages:
    261
    Likes Received:
    18
    How do I take a screen shot of my desktop? I'll show you what I'm looking at.
     
  13. MrPix

    MrPix Registered

    Joined:
    Jan 10, 2012
    Messages:
    249
    Likes Received:
    2
    When I read the title I thought ... yay... more FPS... I checked and I had not applied the force PCI 3 thingy since reapplying the drivers after upgrading from the 780... so applied that... rebooted, checked I was at PCI 3.0x16.... reran my benchmark ... (not the one you have but against the one I ran before) and no difference :(... not even 1fps :( so it may not be true of every system .
     
  14. rogue22

    rogue22 Registered

    Joined:
    Jan 14, 2012
    Messages:
    261
    Likes Received:
    18
    [​IMG]
     
  15. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    run the live benchmark with the pci-e fix on and off. You might be right but you won't know for sure otherwise.
     
  16. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    That's fine, it's just repeating the 3.0 after PCI-e. If you look at the gpu-z screenshot in the op you'll see the same thing but it's at 2.0 instead.

    Edit: If you hover the mouse over that field in GPU-Z it tells you what it is. Basically the second part is telling you what it's at currently. So for mine, it says "PCI-E 2.0 x16 @ x16 1.1" when my gpu is idle and when i do anything that puts the gpu into load it changes to "PCI-E 2.0 x16 @ x16 2.2". It's a power saving technique much like throttling down the cpu and gpu when idle.
     
    Last edited by a moderator: Oct 1, 2014
  17. DrR1pper

    DrR1pper Registered

    Joined:
    Apr 29, 2012
    Messages:
    3,294
    Likes Received:
    36
    I would ask that everyone posting there scores from here on out also include what PCI-e mode their graphics card is running under when in use please.
     
  18. rogue22

    rogue22 Registered

    Joined:
    Jan 14, 2012
    Messages:
    261
    Likes Received:
    18
    When I move my mouse over it its showing PCI-E 3.0. Also if you hit the question mark it tells you at the bottom during the benchmark test what its running at.

    Ripper, click on the question mark, run the little test and tell me what it says at the bottom of your screen.
     
  19. Coanda

    Coanda Registered

    Joined:
    Jun 9, 2013
    Messages:
    689
    Likes Received:
    3
    I have a 780ti and I know that X79 MB's have trouble reading PCI 3.0 therefore a patch called "force-enable-gen3.exe" is needed to solve this issue.
     
  20. MrPix

    MrPix Registered

    Joined:
    Jan 10, 2012
    Messages:
    249
    Likes Received:
    2
    I do know for sure... with the settings I ran previous when it was on 2.0... it is frame for frame the same as when I force enable 3.0. It's a live bench, just not the same one as you suggest, it#s based around my in game settings... the only one that matter to me... anything other is synthetic and not relevant to me. Not being awkward, but synthetic shows a synthetic result... e.g. if I were to change my in game settings to those you provide, I may see an improvement... but I run at different settings.... http://isiforums.net/f/showthread.p...elp-is-needed!?p=308314&viewfull=1#post308314
    so I repeated the test when I got the 980 and got an improvement, but I was on PCI 2.0 as I had over installed the driver (which clears out the reg fix).. ran again last night after applying and no change.
     

Share This Page