Vienna Symphonic Library Forum
Forum Statistics

200,724 users have contributed to 43,209 threads and 259,123 posts.

In the past 24 hours, we have 3 new thread(s), 8 new post(s) and 56 new user(s).

  1. Home
  2. Software
  3. Vienna Instruments Pro (legacy)
  4. Which is better for you: VE Pro Over Ethernet or w/ Audio Hardware?

  • Which is better for you: VE Pro Over Ethernet or w/ Audio Hardware?

    R
    R

    There's been a discussion for a few days over at VI-control - some folks have reported better performance using VE Pro w/ audio hardware and some have reported better performance over ethernet.  By performance I mostly mean latency but there's a bandwidth element to it as well.  I'm curious what others have experienced w/ recent hardware and what VSL thinks on the matter.

    I switched over to ethernet several years ago and my overall system latency improved quite a bit - whereas my slaves occasionally had trouble running at 512 samples I got that down to 128 samples when I switched to VE Pro over ethernet.  I have a total of about 150 tracks coming into my PC DAW (Cubase) from 2 i5 PC slaves via VE Pro over ethernet.  My main DAW runs at 128 samples with 1 buffer to the instances on the slaves.  Very heavy sample streaming from the slaves puts slave CPU usage at about 50% - 60% and DAW CPU usage at 15% - 20%.  My DAW is an i7 920 w/ 24 GB RAM.

    The reason for my interest is that I did that switch several years ago and haven't revisitied the audio hardware option.  Perhaps things have changed over the past couple years that would get me better performance if I switched back to hardware...

    Anyone tested the difference recently?

    Thanks,

    rgames
  • Cyril BlancC
    Cyril BlancC

    Hello

    How can you have 150 tracks in a piece of music ?

    Best

    Cyril
  • R
    R

    OT, I guess, but since you asked: it's about 75 stereo pairs coming from the slaves.  In my template I have about 10 instances of VE Pro sending the whole orchestra back into my DAW w/ separate audio tracks for each string section (Vln I, Vln II, etc), each WW section (flute, oboe, clar, etc.), each brass section (tpt, hrn, etc.), pianos, harps, guitar, bunch of perc., etc.  Plus, I have not only VSL but other libraries, as well.  So it adds up pretty quickly.

    rgames
  • D
    D

    Richard, I had a quick skim of the thread you mention, and it seems to me that there are so many different variables that there is no correct answer.

    1. As soon as you put a Mac into the equation, you are opening another can of worms
    2. People talk about using their hardware, but in order to get 75 stereo returns back to the DAW they would need 2 MADI cards. I guess that most of these people probably have less than 20 returns to deal with.
    3. Different hosts deal with plugs and audio returns differently in regard to core loading
    4. Some hardware enthusiasts tell you that they are using a buffer of 256, when in actual fact they are forgetting to add on the slave buffer (probably also at 256) and the small extra for input latency as well.

    Personally I wouldn't want to deal with the latency of slave machines anyway, and if I had to use them, I would do it this way:

    1. Only do MIDI in the DAW (at a ridiculously low buffer), then send MIDI to the slaves via MOL. I would also have so many slaves that they would each run at a buffer of 128.
    2. Have the slave audio outputs fed directly to a mixing desk, so that there is virtually no latency between playing and hearing
    3. Have the slave audio outputs also sent to a PT machine for when I need to record the audio tracks.

    This way I would never have to worry about pops and clicks and always be able to work at low latency. However, it would actually take me longer to work this way than I currently do with my one machine set-up, so it is better for me just to upgrade my DAW machines fairly often.

    Sorry for the slight OT.

    DG
  • Cyril BlancC
    Cyril BlancC

    Hello

    I was wondering if you where not loading an articulation per track.

    Today I am only using my 12 core, playing a full, complex with very fast parts orchestral piece ; this take about 15-20 % load on each CPU

    I load in memory all the articulation availlable for Appassionata and all SE PLUS. For the wind and brass I load the big preset, I load also 2 x EW choir and a few K4

    and this represent 80 tracks.

    Best

    Cyril
  • R
    R

     DG - do you know of anyone who's actually set it up and tested it?

    The reason I'm asking is that I switched to full ethernet a few years ago and got much better performance at that time.  I haven't re-visited it since and I'm wondering if something has changed to make it worthwhile to consider going back.  Right now I can run my DAW at 128 samples w/ 1 buffer to my slaves - that works great if it's just orchestra but sometimes I have to kick it up to 256 when I add a lot of Omnisphere to a track.  Would be great if I could keep it at 128 or lower at all times.  Perhaps putting the audio cards back in would let me do that.

    The thing that confuses me is that ethernet has much higher bandwidth and much lower latency (about 10x) than the best audio hardware, so I can't quite explain why the audio hardware would be the better route.  Plus, networking is integrated at a very low level into the hardware and OS, so it can handle the transfers at much lower CPU loads than audio hardware (at least that's my theory).  That's why I'm curious if anyone has done the comparison with recent hardware - the tests I did several years ago did, in fact, show that the ethernet connection gave much better performance.

    Thanks,

    rgames
  • D
    D
    last edited
    last edited

    @rgames said:

     DG - do you know of anyone who's actually set it up and tested it?

     

    Unfortunately all the people I know stick to the method they know, be in network or soundcards. However, none of them use 75 stereo returns, so it would be a useless test for you.

    DG
  • R
    R

     Yes - that's basically what I've been doing for the past few years - sticking with what I know.  That's not always a good thing...!

    Even with only a few tracks it still seems like the ethernet connection would better.  Whether you look at bandwidth or latency, ethernet *should* offer *much* better performance:

    1.  Gigabit ethernet has bandwidth equivalent to 450+ audio tracks at 24/44.1 (theoretically it's about 473 but you'll always lose some bandwidth to overhead).

    2.  Typical ethernet LAN's have latency around 0.2 milliseconds (equivalent to a buffer of about 8 samples at 44.1 kHz)

    So at the hardware level, the ethernet solution seems far superior.  Of course, there are processing overhead issues that reduce the performance from these theoretical values but it seems like some good software design would get you in the ballpark of this performance.  Why is it not on some systems?

    rgames