User Tag List

Results 1 to 12 of 12

Thread: The world explodes with frame latency testing

  1. #1

    Default The world explodes with frame latency testing

    Its a bit disconcerting to wake up one morning and find that overnight the entire review scene for both graphics cards and processors has changed dramatically. Together with several websites and writers in the technology field, Nvidia released their in-house frame latency testing tools to the public. Frame latency testing first took off with Tech Report's Scott Wasson showing the world that frame latencies were to blame with jitters and micro-stutters observed in games, and that FPS averages were being used to mask the otherwise obvious issues to give graphics cards more favourable scores. Over time many sites have begun to include frame latency data using FRAPS, but Nvidia's solution, designed with the help of PC Perspective, takes things to a different level entirely.

    Linky

  2. #2

    Default Re: The world explodes with frame latency testing

    I think it's a good idea that this whole frame latency thing has come about, and hat's off to The Tech Report for kicking it all off. While raw fps numbers are fine, I've discovered that it doesn't always mean that the gaming experience will be smooth. Finally also, we have cold, hard data that conclusively proves that microstutter does exist, especially in dual GPU setups. So it's good to know that microstutter is not an old wive's tale, but fact.

    The only negative of frame latency testing is that it's harder work to make sense of it all when reading reviews rather than reading a simple number or score.

  3. #3

    Default Re: The world explodes with frame latency testing

    I agree, some people may not be able to follow what's going on. With PC Perspective offering a more visual indication of what and where the problem actually is, people will be able to see how and why stuttering occurs. The part that most surprised me is the runt frames in a Crossfire setup. AMD's either using those to boost their scores, or the run frames are inserted by the drivers in the "T_present" section to allow both cards to match up the frames properly. AMD used to use AFR rendering to make Crossfire work and I wish they'd stuck to it, but split rendering seems to be the way to go these days.

  4. #4

    Default Re: The world explodes with frame latency testing

    I completely get the validity of this when you're using crossfire, but as far as I'm aware single GPUs are "immune" to microstutter, as they don't need to sync frames with anything besides the monitor when Vsync is enabled. So, essentially, this doesn't change anything for single gpu users, which is basically 90% of the enthusiast segment?

    I do agree though that some games are noticeably smoother at a specific fps than others (Crysis needs a special mention, 30fps feels very smooth where bioshock at 30fps feels hitchy) I haven't noticed a difference between different cards as such, and I've had quite a few in the last couple of years.

  5. #5

    Default Re: The world explodes with frame latency testing

    Quote Originally Posted by McDangerous View Post
    as they don't need to sync frames with anything besides the monitor when Vsync is enabled
    I would have thought so too, but how would that be factored in to a port from a console to the PC? With the console you have set hardware and a set rendering pipeline to work with. What if that now gets ported over to the PC and the game still expects the same reliable frame delivery through optimisation, but doesn't cut it because things don't run at the same speeds?

  6. #6

    Default Re: The world explodes with frame latency testing

    Quote Originally Posted by Wesley View Post
    I agree, some people may not be able to follow what's going on. With PC Perspective offering a more visual indication of what and where the problem actually is, people will be able to see how and why stuttering occurs. The part that most surprised me is the runt frames in a Crossfire setup. AMD's either using those to boost their scores, or the run frames are inserted by the drivers in the "T_present" section to allow both cards to match up the frames properly. AMD used to use AFR rendering to make Crossfire work and I wish they'd stuck to it, but split rendering seems to be the way to go these days.
    I read that PC Perspective article as well. It was pretty heavy going I must say. :) It made a lot of great observations though, and I'm impressed that they took that much time and effort to highlight what's actually happening in the whole rendering process.

    But yeah, it highlights just how bad Crossfire is, and it's put me well off getting another card to crossfire with. Hopefully, this will force AMD to sort out the issue. (Not that Nvidia are blameless, just much less so it seems.)

    @McDangerous

    I've experienced frame latency issues with a single card before, so they're not totally immune when it comes to that.
    Last edited by .exe; 02-04-2013 at 12:02 PM.

  7. #7

    Default Re: The world explodes with frame latency testing

    Where did you experience this @.exe, out of curiosity?

    Wouldn't that be game engine dependent though @Wesley? There is obviously synchronization happening to some degree between the different pieces of hardware that form the pipeline, but I can't see how hardware could be at fault for frame latency issues. Drivers and software, yes, but not hardware.

  8. #8

    Default Re: The world explodes with frame latency testing

    Quote Originally Posted by McDangerous View Post
    Where did you experience this @.exe, out of curiosity?
    I experienced it with both Arma III and Crysis 3, using a 7950. It had to do with AMD's Boost function, and once I had managed to set the overclock at 1100 mhz and keep it there, the problem seemed to go away.
    I've seen some users over at overclock.net complain about latency issues with single cards as well, although it is very rare to be fair.

    But funnily enough, it can be dependent on the software too. Nearly everyone who played IL-2 Sturmovik Cliffs of Dover had microstutter, even single card users, regardless of card type or driver set. The game engine was the problem, rather than the cards.
    Last edited by .exe; 02-04-2013 at 01:14 PM.

  9. #9

    Default Re: The world explodes with frame latency testing

    That I can think though. I used to have that issue with my 5850 and secondary monitor, where the memory clock changing would cause the frames to go out of sync with the monitor. As I said though, it's a driver issue or a software issue, not a hardware issue.

  10. #10

    Default Re: The world explodes with frame latency testing

    Quote Originally Posted by McDangerous View Post
    but I can't see how hardware could be at fault for frame latency issues. Drivers and software, yes, but not hardware.
    Well, Tech Report's testing with FRAPS uses the "T_present" frame timestamps for their analysis, which is just after the hardware is done with the raw code. Hardware could indeed be a factor in frame latency. I wish they could show us results using older cards like the HD6970 or the GTX580. That would help to see if the previous generation had the same issues.

    Quote Originally Posted by .exe View Post
    It had to do with AMD's Boost function, and once I had managed to set the overclock at 1100 mhz and keep it there
    I know that AMD's Boost tech doesn't work as elegantly as it should. Some of the tests I've seen that showed GPU frequency over time also seems to indicate that Boost was responsible for a lot of the stuttering. Then again, this is still relatively new territory for everyone so who knows what the final cause will be.

  11. #11

    Default Re: The world explodes with frame latency testing

    Further to this discussion, check this out:

    These are frame latency times for Bioshock Infinite using single cards. (The first three panels anyway)

    Note how on occasion, AMD cards do spike above 50ms, which will be perceptible to the eye as stutter. So again, single cards - AMD cards specifically - are not above microstutter. It's slightly worrying being an AMD user myself, and I really hope that this exposing of the issue will force AMD to do something about it.
    As said in the article, it may be hard to see with the naked eye for some, but it is still there.

    http://www.guru3d.com/articles_pages...nchmark,7.html
    Last edited by .exe; 04-04-2013 at 01:40 PM.

  12. #12

    Default Re: The world explodes with frame latency testing

    Note that unlike TechReport's graphs, their benchmarks show that there's a stutter that happens frequently and predictably. I'm assuming that's AMD's memory manager breaking things up, because they've been trying to make GCN work perfectly with regards to memory management for ages.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •