Vienna Symphonic Library Forum
Forum Statistics

180,742 users have contributed to 42,140 threads and 254,362 posts.

In the past 24 hours, we have 3 new thread(s), 18 new post(s) and 49 new user(s).

  • First Test results for Vienna Instruments on PC

    First Vienna Instruments Stress Test results on PC

    We have tested the Vienna Instruments on various computers and doublechecked on similar machines. All computers are fully equipped DAW´s that are used in everyday studio work, no additional tuning.

    More tests are in the works, with different processors and hosts. Our tests with Sonar, FX Teleport, Chainer, Plogue Bidule, Forte showed the same results, the Vienna Instruments behave the same way as in Steinberg Cubase SX or Vstack.

    The test results for Audio Units on Apple will follow as soon as possible.

    Please bear in mind, that the RAM available is also used by the hosts (around 300 MB) and Windows (another 300 MB), which is why the limit of RAM usage is 1.3-1.5 GB for the Vienna Instruments in a computer equipped with 2 GB RAM.

    The testing routine: Play a (really stressy!) stress test song with different loading options (more instruments, less RAM, and the other way around). All instruments are playing all the time, using all kinds of matrices and patches, and of course including additional MIDI data like ModWheel and different assigned sliders.

    You always see the maximum Vienna Instruments possible in the given setup without any sonic trouble.

    TEST COMPUTER 1

    Processor: Pentium 4, 3.2 GHz,
    Ram: 2 GB
    Data storage: internal SATA
    Host application: Cubase SX3
    Soundcard: RME HDSP Multiface


    Latency at 512 Samples (12ms) / 1.2 GB Ram usage
    Maximum Vienna Instrument instances: 24 (CPU 75%)
    Maximum polyphony: 200 stereo voices / (CPU 75%)

    Latency at 256 Samples (6 ms) / 1.3 GB Ram usage
    Maximum Vienna Instrument instances: 20 (CPU 70%)
    Maximum polyphony: 200 stereo voices / (CPU 75%)

    TEST COMPUTER 2

    Processor: Pentium 4, 3 GHz,
    Ram: 2 GB
    Data storage: Firewire 400
    Host application: Cubase SX2
    Soundcard: RME DIGI 9652


    Latency at 1024 Samples (24 ms) / 1.3 GB Ram usage
    Maximum Vienna Instrument instances: 24 (CPU 50%)
    Maximum polyphony: 200 stereo voices / (CPU 60%)

    Latency at 512 Samples (12 ms) / 1 GB Ram usage
    Maximum Vienna Instrument instances: 18 (CPU 50%)
    Maximum polyphony: 200 stereo voices / (CPU 60%)

    Latency at 256 Samples (6 ms) / 1.2 GB Ram usage
    Maximum Vienna Instrument instances: 14 (CPU 60%)
    Maximum polyphony: 180 stereo voices / (CPU 65%)

    TEST COMPUTER 3

    Processor: AMD 2.6 GHz,
    Ram: 2 GB
    Data storage: Firewire 400
    Host application: Cubase SX3
    Soundcard: Creamware Pulsar & Luna


    Latency at 13ms / 1.3 GB Ram usage
    Maximum Vienna Instrument instances: 20 (CPU 70%)
    Maximum polyphony: 200 stereo voices / (CPU 70%)

    Latency at 7 ms / 1.2 GB Ram usage
    Maximum Vienna Instrument instances: 16 (CPU 65%)
    Maximum polyphony: 174 stereo voices / (CPU 60%)

    TEST COMPUTER 4

    Processor: Pentium 4, 2.8 GHz,
    Ram: 2 GB
    Data storage: internal SATA
    Host application: VSTack
    Soundcard: RME DIGI9632


    Latency at 512 Samples (12ms) / 1.4 GB Ram usage
    Maximum Vienna Instrument instances: 16 (CPU 70%)
    Maximum polyphony: 180 stereo voices / (CPU 70%)

    Latency at 256 Samples (6 ms) / 1.4 GB Ram usage
    Maximum Vienna Instrument instances: 8 (CPU 55%)
    Maximum polyphony: 160 stereo voices / (CPU 70%)

    TEST COMPUTER 5

    Processor: Pentium 4, 2.4 GHz,
    Ram: 1.5 GB
    Data storage: Firewire 400
    Host application: VSTack
    Soundcard: Steinberg VSL 2020


    Latency at 512 Samples (12ms) / 1 GB Ram usage
    Maximum Vienna Instrument instances: 16 (CPU 60%)
    Maximum polyphony: 200 stereo voices / (CPU 80%)

    Latency at 256 Samples (6 ms) / 1 GB Ram usage
    Maximum Vienna Instrument instances: 14 (CPU 60%)
    Maximum polyphony: 200 stereo voices / (CPU 80%)

    TEST COMPUTER 6

    Processor: Laptop Pentium M, 2 GHz,
    Ram: 2 GB
    Data storage: Firewire 400
    Host application: Cubase SX3
    Soundcard: RME HDSP Multiface


    Latency at 512 Samples (12ms) / 1.2 GB Ram usage
    Maximum Vienna Instrument instances: 14 (CPU 65%)
    Maximum polyphony: 160 stereo voices / (CPU 65%)

    Latency at 256 Samples (6 ms) / 1.1 GB Ram usage
    Maximum Vienna Instrument instances: 10 (CPU 55%)
    Maximum polyphony: 150 stereo voices / (CPU 65%)

    TEST COMPUTER 7

    Processor: Pentium 3, 1 GHz,
    Ram: 2 GB
    Data storage: Firewire 400
    Host application: VStack
    Soundcard: RME DIGI 9632


    Latency at 512 Samples (12ms) / 0.6 GB Ram usage
    Maximum Vienna Instrument instances: 5 (CPU 70%)
    Maximum polyphony: 90 stereo voices / (CPU 65%)

    Paul Kopf Product Manager VSL
  • Thanks Paul - its good to have some data at last.

    Would you mind please clarifying what a typical load of "patches" was in the tests. Obviously you wouldn't be able to load 24 Universal instrument patches onto the top specced machine - or maybe you could.....

    Regards


    Tim

  • Hi Paul

    It's good to have some data at last. However, I'm slightly disappointed that the results weren't better using FX-Teleport. In theory every instance should have a limit of 2Gb, therefore one should be able to load in excess of 3Gb on a machine that has 4Gb RAM. Did you test for this, or were your tests limited to 2Gb machines?

    DG

  • will you be posting mac results soon too [:D]

  • We're the P4s dual core?

  • Hi, thanks for posting.

    I agree with Tim - whats really important is how many samples can you load with this as opposed to GS3.

  • What looks really worrying in the first instance is the max instances of VI achieved - 24 max. I have regularly used 100 EXS 24 instruments in a score now I know the VI's with their matrices reduce the requirement for channels. But in an orchestral line-up"
    Strings 5, Wind 7, Brass 6, Harp, Piano Percussion (4) and now we've run out before we think solo's, choirs, and all the other instruments often used.

    In Logic on a mac if you run out of voices, RAM or CPU you can always freeze the most important thing you can complete the largest orchestrations on one computer.

    If there is really a score limit of 24 staves (24 instruments) this could be an issue.

    Let's hope the mac performances are a little more revealing when they are published.

    Julian

  • last edited
    last edited

    @julian said:

    Let's hope the mac performances are a little more revealing when they are published.

    Julian


    Mac will be far worse [:P]

  • Hi Paul,

    Do you think it will be possible to let this "test song" available for the end user ? Then we could post our results here.

  • last edited
    last edited

    @Peter Alexander said:

    We're the P4s dual core?

    ... and where are the dual core AMDs?

  • a basical comment on hyperthreading (= virtual dual processor), dual processors and dual core processors ....

    as you might know audio applications run in fact not multi-threaded, this means an audio-related working process cannot be split across processors (be it hyperthreaded, dual or dual core processors). the work has to be done by one processor - this goes so far, that even processes of a sequencing and sampling application have to run both on the same processor if audio (and sometimes also midi) flows between them.

    although some work can be done by a second processor: rendering and updating the grafic interface, file-access, streaming, ect.
    in this sense the Vienna Instruments are supporting multiple processors.

    christian

    and remember: only a CRAY can run an endless loop in just three seconds.
  • last edited
    last edited
    Hi all,

    this is the first test series we have made, more results will be published, when the second run is through (AMD dual core, P4 double core, I´m trying to get a hold of the fastest machines I can find).

    @Another User said:

    What looks really worrying in the first instance is the max instances of VI achieved - 24 max. I have regularly used 100 EXS 24 instruments in a score now I know the VI's with their matrices reduce the requirement for channels. But in an orchestral line-up "Strings 5, Wind 7, Brass 6, Harp, Piano Percussion (4) and now we've run out before we think solo's, choirs, and all the other instruments often used.

    In Logic on a mac if you run out of voices, RAM or CPU you can always freeze the most important thing you can complete the largest orchestrations on one computer.

    If there is really a score limit of 24 staves (24 instruments) this could be an issue.

    Let's hope the mac performances are a little more revealing when they are published.


    Julian, you can make use of RAM Optimize to stay on the MIDI level, when your RAM reaches the limits. We would like to see the perfect orchestration computer that never runs out of RAM ourselves...

    I can´t wait to have the final figures for MAC as well, I just really want to make sure that everything is doublechecked, so that you know what it really looks like....

    Thanks for your input, everybody!

    Best wishes, Paul

    Paul Kopf Product Manager VSL
  • last edited
    last edited

    @Another User said:


    Julian, you can make use of RAM Optimize to stay on the MIDI level, when your RAM reaches the limits. We would like to see the perfect orchestration computer that never runs out of RAM ourselves...

    I can´t wait to have the final figures for MAC as well, I just really want to make sure that everything is doublechecked, so that you know what it really looks like....

    Thanks for your input, everybody!

    Best wishes, Paul


    Hi Paul,

    The inportant number for me is instances of VI - how many will I be able to run. RAM or CPU is actually less of an issue as with freezing or RAM optomization there are ways round this. Where only 24 VI instances would be a problem would be in only having 24 arrange tracks available for Logic automation - only having the equivalent of 24 staves in a score - what happens when you have inserted the 24th
    VI and you still need 2 or 3 extra instruments. With the current VSL library this is not an issue - freeze a track or two and insert more EXS 24's but if you reach the instances limit with VI there is nowhere else to go particularly as VI's won't run as a node.

    I admit it's early days and we must give you guys time to do your tests and perhaps optomize the software but one question I would like to ask is what is the maximum number of VI's that can be inserted on a top flight PC with very low instrument content (i.e. each VI only has a minimal instrument/sample load)

    thanks

    Julian

  • last edited
    last edited

    @Another User said:

    However, I'm slightly disappointed that the results weren't better using FX-Teleport. In theory every instance should have a limit of 2Gb, therefore one should be able to load in excess of 3Gb on a machine that has 4Gb RAM. Did you test for this, or were your tests limited to 2Gb machines?


    DG, I will look into that, more to come....


    Hi Paul, I didn't mean to hassle you, it's just that there are so many options, and in this period of calm before the storm I want to prepare as much as possible before D-Day (talk about mixed metaphors [8-)] ).

    The other thing to be aware of is that Cubase SX3 (and Nuendo) are both Large Address Aware, so that with the 3gig switch (I think) one can access up to about 2.7Gb RAM. It would be interesting to see how the number of samples in this instance stacks up to the figures quoted for MAC. I think that this will eventually be helpful to users, as at the moment with no MAC figures about performance we don't really know. For example, number of instances might be more important than RAM use to some people; maximum number of samples might be of more interest to others. I think that the polyphony looks OK, but obviously it depends on how VI deals with multi-layer cross-faded instruments. The only time I get close in GS is when I have multiple keyboard instruments running, and there ain't non of them in VI [:D]

    DG

  • Vienna Instruments run all the articulations you need for an instrument in one player. Therefore I assume "instruments" really means *instruments* - i.e. violins, violas, and oboes = 3 instruments.

    So when you say 20 - 24 instruments, you're talking about the equivalent of five billion programs.

    Ja?

  • last edited
    last edited

    @julian said:

    What looks really worrying in the first instance is the max instances of VI achieved - 24 max. I have regularly used 100 EXS 24 instruments in a score now I know the VI's with their matrices reduce the requirement for channels. But in an orchestral line-up"
    Strings 5, Wind 7, Brass 6, Harp, Piano Percussion (4) and now we've run out before we think solo's, choirs, and all the other instruments often used.

    In Logic on a mac if you run out of voices, RAM or CPU you can always freeze the most important thing you can complete the largest orchestrations on one computer.

    If there is really a score limit of 24 staves (24 instruments) this could be an issue.

    Let's hope the mac performances are a little more revealing when they are published.

    Julian



    Julian, with all respect, I feel your comparison at this date is unfair. THe PC is not the Mac. Today, from my seat, with all you can do, 24 instances on a single PC machine looks impressive, especially with the simple specs needed to run VI

    I also point out, that the 64-bit systems on the PC are not yet feasible, especially in price, and probably won't be for half a year or more. WIth what I'm reading, with three smaller dedicated PCs you can set up one "whoppe"r of an orchestra!

    PA

  • I really want to thank Herb, Paul, Chris and Company for taking the time to do these tests and for taking the time to post them. This is exceptional. No other company has ever done this and for one, I really appreciate your efforts. That you went and tested a PII and a PII is also fantastic given the number of composers who have those machines because of GS2.54, and now find that they can still get some use out of those systems.

    This to me is corporate responsibility in action.

    On the P4s, it would be great to know if you were using the P4 with HT or the dual core, and the motherboard type (915, 925, etc).

    Again, many, many thanks.

    PA

  • last edited
    last edited

    @Nick Batzdorf said:

    So when you say 20 - 24 instruments, you're talking about the equivalent of five billion programs.

    Ja?


    And two bratworsts with a Diet Coke.

  • last edited
    last edited

    @Peter Alexander said:

    I really want to thank Herb, Paul, Chris and Company for taking the time to do these tests and for taking the time to post them. This is exceptional. No other company has ever done this and for one, I really appreciate your efforts. That you went and tested a PII and a PII is also fantastic given the number of composers who have those machines because of GS2.54, and now find that they can still get some use out of those systems.

    This to me is corporate responsibility in action.

    PA


    I couldn't have said it better myself.

    DG

  • Thanks, and I am in the middle of further tests...

    I´m actually not that much of a techie to answer all specific computer questions, but I will look into it and we will test all configurations within our reach.

    As far as the dual cores are concerned, Christian Marin has already posted on the first page of this thread and I don´t really have anything to add...

    To my knowledge we can adress up to 1.5 - 1.7 GB of RAM, and all suggestions here on the forum seem rather theoretical to me... If anyone has experience with adressing more than 1.7 GB of RAM with a streaming sampler on a PC, please share your infos.

    Best, Paul

    Paul Kopf Product Manager VSL