Interesting find when comparing old nvidia driver to newest

Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by samusprime89, Oct 9, 2015.

  1. samusprime89

    samusprime89 Active Member

    Messages:
    55
    Likes Received:
    0
    GPU:
    Gigabyte GTX 970
    Hello, i wanted to share my finding with those who own an Nvidia GPU. Ive done some testing and it appears Nvidias latest drivers cause some of the highest DPC latency spikes i have ever seen, even on idle, i also want to clarify that when comparing the new driver to the older (353.62 driver) i am only trying to show how Nvidia needs to work out some of the bugs causing hitching and stuttering in games. I am using Windows 10 pro 64-bit (specs are under my name) i have some comparison screenshots of the Tests i performed.

    Tested with DPC Latency Checker V1.4.0
    IDLE Test: http://i.imgur.com/M30LUSw.png

    Load Test: http://i.imgur.com/TcvC6zy.png

    As you can see in the comparison the new Nvidia drivers cause some pretty insane latency spikes, that cause speaker pop or crackling in game and it even affects the framerate with what we call "microstutter". What is odd is that the older driver 353.62 didnt have a single latency spike above 1000Us whatsoever and games were as smooth as we should expect. but its a different story when testing on Nvidia latest driver. the highest spike i seen was 4329us causing severe frame hitching in all games i have tested.

    Games tested in order: Shadow of Mordor benchmark, Tomb Raider benchmark, Hitman Absolution benchmark. Unigine Valley 1.0

    The purpose of my testing was not to measure framerates, but the frame-times and DPC latency. But as you can see Nvidia has a lot of work to do. For the time being those of you who experience high latency or frame hitching should use the older 353.62 drivers, even though the drivers were months old i still had no issues running some of the newer game titles such as Mad Max or MGS: The Phantom Pain.
     
  2. Nastya

    Nastya Member Guru

    Messages:
    148
    Likes Received:
    35
    GPU:
    RTX 2080 Ti 11GB
    Please don't use DPC Latency Checker on Windows versions later than 7 - it's been said by the developer and other users numerous times measured values are incorrect.

    LatencyMon on the other hand is more than suitable for later OS.
     
  3. CalculuS

    CalculuS Ancient Guru

    Messages:
    3,062
    Likes Received:
    287
    GPU:
    GTX 1660Ti
    ^ Correcto mundo.

    http://www.thesycon.de/deu/latency_check.shtml
     
  4. snight01

    snight01 Master Guru

    Messages:
    333
    Likes Received:
    28
    GPU:
    RTX 2080 Ti FE
    I'm not sure about the inaccuracy of dpc latency. If inaccurate it would show mostly the exact results when he's testing different drivers. Also I test latency mon and compare and they remain closely consistent.

    You're right about the findings because I noticed it also.
     

  5. siriq

    siriq Master Guru

    Messages:
    790
    Likes Received:
    14
    GPU:
    Evga GTX 570 Classified
    I have the same latency with 358.50 as you have with 353.62 driver. I run my cpu at 800 will be higher. For me only the power saving feature causing the higher latency. Since 3.7 against 800 mhz, i would say it is all fine for me.
     
    Last edited: Oct 9, 2015
  6. tsunami231

    tsunami231 Ancient Guru

    Messages:
    12,120
    Likes Received:
    929
    GPU:
    EVGA 1070Ti Black
    latency time in nvidia drivers have been the same for years providing everything is working correct.

    in game nvidia driver latency is 200-500us is the normal range if you getting high spikes like 1000us + you have other issues

    and use
    http://www.resplendence.com/latencymon

    it is accurate compared to DPC Latency Checker V1.4.0 which isnt even updated for newer OSes

    Power saving functions cause spikes being CPU features or GPU features.
     
    Last edited: Oct 9, 2015

Share This Page