AMD Radeon Software Crimson 16.3

Discussion in 'Videocards - AMD Radeon Drivers Section' started by d2kx, Mar 9, 2016.

  1. Rich_Guy

    Rich_Guy Ancient Guru

    Messages:
    12,681
    Likes Received:
    667
    GPU:
    MSI 2070S X-Trio
    Thought it might have helped in GTA V with the judder in Xfire, but nope, so switched it back to a single card, which runs it butter at the same settings, and turned the Cache off :p
     
  2. The Mac

    The Mac Ancient Guru

    Messages:
    4,404
    Likes Received:
    0
    GPU:
    Sapphire R9-290 Vapor-X
    exactly.

    mine is all over the place, but it has zero effect on performance.

    Thats how its supposed to work.
     
  3. Ryu5uzaku

    Ryu5uzaku Ancient Guru

    Messages:
    7,057
    Likes Received:
    273
    GPU:
    6800 XT
    my tri-x and msi gaming are at 1040 all the time. Unless somehow temps go to 94 then the msi throttles. Tho because my msi started to overheat I put both cards at 1000 no worries.

    But no power saving thingies doing things. I understand other fluctuation but the thing everyone is talking about is the amd introduced power saving feature that never has been active on my 2xx cards.

    I guess I've just ticked off something from afterburner then.
     
  4. The Mac

    The Mac Ancient Guru

    Messages:
    4,404
    Likes Received:
    0
    GPU:
    Sapphire R9-290 Vapor-X
    again, it isnt in 2xx card.

    its only on 3xx and fiji

    Mine is all over the place because i use the fram limiter.
     

  5. sammarbella

    sammarbella Ancient Guru

    Messages:
    3,929
    Likes Received:
    178
    GPU:
    290X Lightning CFX (H2O)
    How do you know it has zero performance effect if it's "all over the place"?

    Did you compared it to a straight LOCKED GPU clocks line?

    I guess not.

    The best GPU performance (and stability) occurs always when the GPU has no delay at all between the need of the highest clock and its raise from lower clock because it remained always at highest clock.

    If your GPU core clocks are always LOCKED at 1080/1250 mhz (i.e) while gaming (3D) it will not need to recover from 500/625(i.e) (2D) in cases the game requires more GPU powa.

    No delay, no performance loss.

    If you read The Mac posts AfterBurner or GPU-Z GPU clocks, FPS and frametimes monitoring must be placebo.

    :D

    Disable the frame limiter to see the truth.Compare apples to apples not to oranges.

    As you replied in other thread i don't expect you to do it: you are not going to do something to prove yourself wrong...
     
  6. Adiurd

    Adiurd Member

    Messages:
    35
    Likes Received:
    2
    GPU:
    RX 580 8gb
    PrMinisterGr, I forgot it thanks for the 7970 bios.
     
  7. SimBy

    SimBy Member Guru

    Messages:
    189
    Likes Received:
    0
    GPU:
    R9 290
    When GPU power management works as it should (pretty much on everything below 300 and Fury) there is no noticeable impact on performance. Just like there is no noticeable impact on performance with changing CPU power states. In some corner cases performance may even improve because it's less likely you will hit temp limit when you don't force your GPU running at max speed all the time.
     
  8. SimBy

    SimBy Member Guru

    Messages:
    189
    Likes Received:
    0
    GPU:
    R9 290

    What exactly will he prove by disabling a frame limiter?! Clocks will rise and so will FPS.
     
  9. TheDukeSD

    TheDukeSD Member Guru

    Messages:
    145
    Likes Received:
    11
    GPU:
    MSI GT 1030 2GH OC
    You waste your time trying to make others understand. If they don't feel they assume it's not happening. Ofc people are not the same, some are capable to feel/notice some things, some are not, it's actually normal.
    Ofc someone will come up with some tests showing that reactions time are close to 200ms, totaly ignoring that those reactions tests include the hand muscle movement (and let's face it those muscle are extremely slow) while the eye-brain stuff is considerable faster (not that easy to make a test only for the eye-brain).

    Also some people just don't understand how electronic components work.

    It's very hard to make someone understand that it's not like him/her (often proves to be impossible).

    Regarding the clocks. If I play League of Legends and I let the driver untouch the clocks will be 300/150 most of the time and the game will feel not smooth (ab doesn't show anything wrong with frame times, fps is a nice 60 but it just doesn't feel smooth to me), if i force max clocks 800/1125 the game feels smooth for me. The difference is visible in how I feel the smoothness of the game (it doesn't matter what driver I use, it's also the same). Ofc someone can say it's placebo, I prefer to believe what I feel and not what someone else is saying. Maybe others just don't feel or notice (for example my father doesn't see the tearing and is unable to notice diferences between vsync on or off (I was at him and in under 1 min i realised he was playing lol with vsync off, and i wasn't even watching carefull what he was playing :p)).
    (To be read the game doesn't feel smooth when the gpu is at 300/150 even if those clocks stay like that for the entire game and ab show no spikes in fps/frame times and the gpu usage is low during the entire time.)
     
    Last edited: Mar 16, 2016
  10. sammarbella

    sammarbella Ancient Guru

    Messages:
    3,929
    Likes Received:
    178
    GPU:
    290X Lightning CFX (H2O)
    The performance loss caused by delay in power state changes to max in GPU and/or GPU seems to be relative in "noticiable" subjective value. :)

    AFAIK power saving features are not designed to prevent temp throttling caused by high ambient temp or lack of air flow in cases but as you said it could be a side effect.

    GPU throttling like CPU throttling are not helping in performance at all when the PC is not hitting CPU/GPU temp limits.

    I don't know your PC but mine can work 24/7 at any load without hitting temp limits.

    If he compares between power saving behavior vs LOCKED GPU clocks without frame limiter in both cases, FPS and frametimes will be performance and stability objective comparable values.

    Frame limiter in one (or both cases) make fair comparison impossible.

    Apple vs oranges.
     

  11. SimBy

    SimBy Member Guru

    Messages:
    189
    Likes Received:
    0
    GPU:
    R9 290
    I'll skip the 1st part since I'm not sure what human reaction time has anything to do with what we are talking about here.

    So you're saying LoL runs at constant 60fps and frame times are constant and yet somehow its not smooth. But when you raise clocks it still runs at 60fps and frametimes don't change but now it's somehow smooth.

    Mkaay.
     
  12. TheDukeSD

    TheDukeSD Member Guru

    Messages:
    145
    Likes Received:
    11
    GPU:
    MSI GT 1030 2GH OC
    Exactly. AB show no spikes in frame times.

    In the past I have done experiments underclocking a cpu (it was still running at max clock). From a particular voltage I started to get a similar effect, while it was stable in any test I did games were no longer smooth. And yes I was seeing nothing wrong in term of frames or frame times in ab.
    For some reasons that cpu was no longer snappy below some voltage.
     
    Last edited: Mar 16, 2016
  13. The Mac

    The Mac Ancient Guru

    Messages:
    4,404
    Likes Received:
    0
    GPU:
    Sapphire R9-290 Vapor-X
    I have sammarbella on iggy, but someone quoted him so ill respond.

    My FPS is locked at 59fps, and my clocks fluctuate all over the place.

    of course if i turn off the frame limiter my clocks raise, as does my fps.

    for the most part, clocks dont fluctuate at that point (outside of pipeline stalls caused by the game engine)

    Thats the whole purpose of the frame limiter.

    So yes, i know the fluctuations dont effect performance, or i wouldnt get a solid 59fps.

    lol
     
    Last edited: Mar 16, 2016
  14. SimBy

    SimBy Member Guru

    Messages:
    189
    Likes Received:
    0
    GPU:
    R9 290
    I'm one of those corner cases that limits max fan speed to a very specific low value. I like silent systems.

    That means high temps. In some cases even above 90. Hawaii GPUs are fine running up to 94 C. If I forced my GPU to run at max clocks and voltage all the time, I would almost always hit 94 C in newer demanding games.

    And regarding FPS cap. If he removes the limit his GPU will ramp to max speed and stay there. As does mine. Because power management works as it should. Well unless the GPU is not fed and stalls. XCOM2 is a great example of that. Lot's of background loading causing FPS to drop and GPU clocks follow.
     
  15. SimBy

    SimBy Member Guru

    Messages:
    189
    Likes Received:
    0
    GPU:
    R9 290
    Well I don't know whats going on in your case but frame times and FPS are the only objective measures of 'smoothness' that I know of.

    Maybe you should try using FCAT.
     
    Last edited: Mar 16, 2016

  16. sammarbella

    sammarbella Ancient Guru

    Messages:
    3,929
    Likes Received:
    178
    GPU:
    290X Lightning CFX (H2O)
    You can't know that the AMD clock roller coaster doesn't affect performance because you limit your FPS at 60 FPS limiting the max performance in the first place.

    How you compare by the max 60 FPS?

    A lot of games can be "locked" at 60 fps even at 2D clocks.

    No need of dedicated GPU to play tetris.

    LOL
     
  17. sammarbella

    sammarbella Ancient Guru

    Messages:
    3,929
    Likes Received:
    178
    GPU:
    290X Lightning CFX (H2O)
    I agree.

    Both values comparision with power saving feature vs GPU clocks locked without limiting factors like frame limiter, it's a no brainer.
     
  18. TheDukeSD

    TheDukeSD Member Guru

    Messages:
    145
    Likes Received:
    11
    GPU:
    MSI GT 1030 2GH OC
    Yep I admit I was lazy. Should had used something batter than AB. :)

    Either my particular chip just doesn't like to run below a particular voltage or the voltage reported by the gpu sensors are wrong (the sensor reports a value but in reality it's lower).
     
    Last edited: Mar 16, 2016
  19. janos666

    janos666 Maha Guru

    Messages:
    1,038
    Likes Received:
    171
    GPU:
    MSI RTX3080 10Gb
    Interesting. Do you mean by "globally enabled" that you applied some shader cache related tweaks like that of the managed registry editor, RadeonMod?

    I didn't fully understand the implications of using the hidden "cache everything" value, so I left it at default. I though the size is hard-capped to 32Mb (based on previous experience), so caching literally everything (including web browsers, media players, random small applications, etc) would just make it a constantly rotating scratch-board which rarely if every has anything in store which is actually needed, thus it's better to cache only the most important things.
    But I guess I should try that undocumented value. :)

    I have never seen a file bigger than 32Mb and/or more than a single file with >1Mb size (I occasionally have some small files but it's hard to track due to cache wipe upon driver update).
    I am sure the file I checked for this comparison was for DAI because I explicitly enabled the cache in DAI's RS profile and the file showed up after I launched the game (this is the only gave I own which has it's own Mantle shader cache to make some rough comparison).
    I loaded some different saves in the game after wiping both caches (this driver cache and the mantle cache) and the mantle cache was significantly bigger after that (32 vs. 64Mb) even though I have previously seen even bigger mantle cache size (128Mb after many hours of actual playthough, visiting various areas on the map, including those only accessible during specific plot events).
     
  20. rogue221979

    rogue221979 Master Guru

    Messages:
    643
    Likes Received:
    7
    GPU:
    Gigabyte 3090 GTX
    Are the crossfire issues fixed that plagued 16.1 and .2?

    I'm still on 15.12 as its the only way I can get crossfire to work in Elite dangerous and a few other games I run that have no profile. Using the "enable crossfire for apps without a profile" option.

    It vanished in 16.1 so I had to reinstall 15.11 then run 15.12 over top to get it back.
    I don't dare install these and go thru all that again if I don't have too.

    I suppose, I could download them for now.
     

Share This Page