Vienna Symphonic Library Forum
Forum Statistics

180,822 users have contributed to 42,142 threads and 254,366 posts.

In the past 24 hours, we have 2 new thread(s), 4 new post(s) and 76 new user(s).

  • [Solved] Help! dropouts of precisely one buffer length

    Main machine - Intel 9960X, 128gb Ram, W10 Pro high performance with all bluetooth and wifi disabled in device manager, intel x550 network card connected directly (ie no switch). 128 ASIO buffer, and 2 buffers additional in the one connected vienna instance. Running Nuendo 10, latest version. Connected to....

    New Slave machine - AMD 3950X, 128GB Ram, Corsair PCIE4 NVME system and samples drive. Vepro 7 latest version.

    I keep getting intermittent audio dropouts, about one every 30 seconds or so. If I bus the audio back to a group and to an audio track and print it, I can see that the audio literally just cuts out. Here's the funny thing though - it's always exactly 128 samples. I tried dropping down to 64 buffer just to see if it increased and the frequency of them didn't but the length oocasionaly was a multiple of 128, eg 512 samples dropout. Always a multiple of 128.

    The dropouts seem to disappear if I go up to 512, but then the latency is 20ms and unacceptable, and besides, we're talking about one patch on one slave.

    On the ethernet connections - I tried turning off interrupt moderation, which changed nothing, but even with that disabled changing the interrupt moderation rate to extreme increased the number of these dropouts. Jumbo frames currently disabled.

    I know it's not performance based because there's plenty - tons - of CPU and RAM, disk and ethernet bandwidth on both ends. The fact the dropouts are precisely one buffer long seems to be a clue. There is no switch involved but the same happens if I use a switch, incidentally...


  • Hi,

    This is indeed a strange issue. Please add following information:

    - Which additional fx plugins are in use?
    - Are you using MIR? If so, is the issue still present if you disable MIR? If MIR is in use, please attach an image with the MIR settings in VEP.

    Best, Ben


    Ben@VSL | IT & Product Specialist
  • No FX Plugins used - or even installed! There is one plugin installed, Kontakt.

    MIR - never used it. Looking into how to disable it if any component of it is enabled by default, but I've never knowingly activated it...

    Just for reference, here are two dropouts as an image attached. This is at 128 buffer, 2 vienna buffers. 24/48khz. Barely any system resources being used on either machine. The left glitch is precisely 128 samples. The right glitch is precisely 256.

    Fixed IP addresses on the 10.0.1.X and 255.255.0.0 subnet, both machines have an Intel X550 card, directly connected with CAT 6A.

    I'm starting to suspect that the Focusrite DANTE card might be a problem (it's the only thing I haven't changed between any setups experiencing these glitches), so I've got an RME Dante USB interface arriving tomorrow, to see if it's a problem with the Focusrite ASIO drivers or PCIE.

    Image


  • OK....tentatively I may have solved it. Touch wood...

    I disabled hyperthreading on the Nuendo machine, and now it's all just fine. Something about using both DANTE and Vienna and having hyperthreading on is causing Vienna or the host cpu to literally "skip" an entire audio buffer (or two, or three) sometimes. No idea why. Now I can run 64 buffer and it's not so much as glitching once....


  • Great to hear that you found a solution.


    Ben@VSL | IT & Product Specialist