Vienna Symphonic Library Forum
Forum Statistics

194,118 users have contributed to 42,911 threads and 257,916 posts.

In the past 24 hours, we have 3 new thread(s), 13 new post(s) and 81 new user(s).

  1. Home
  2. Sample Libraries
  3. Vienna Instruments Libraries
  4. 3Gb Switch sample loading tests

  • 3Gb Switch sample loading tests

    D
    D
    I have been testing on and off for the last few days and here are my findings. Reported use of RAM is always from Task Manager unless otherwise stated. I have also stated the amount with which I got the system to work in a stable manner, as well as the total loading before the system collapsed. Please note that this is in no way supposed to be a scientific test, just one for my own purposes that I choose to share!

    All tests are done with the same PC; Intel Pentium 4, 3.20GHz. Some configurations of memory were not tested due to the PC only liking matched pairs of RAM.

    I loaded 4/5 instances of Vienna VI and am working via FX-Teleport using Remote Desktop from a host PC with Nuendo as the DAW.

    I have also followed my own suggestion and removed memory tweaks from the slave PC, so memory is set at XP default.

    Test 1

    2Gb RAM (4x512Mb)

    Total loading possible 1750Mb
    System stable 1700Mb

    Test 2

    2Gb RAM (4x512)

    3Gig switch installed
    Total loading possible 1953Mb
    System stable 1933Mb

    Test 3

    3Gb RAM (2x1Gb, 2x512Mb)

    3Gig switch installed
    Server.exe pre-release Large Address Aware version
    Total loading possible 2957Mb
    System stable 2900Mb
    VI reported available memory 2781Mb
    VI reported memory in template 2663Mb
    Number of samples 42,631

    Test 4

    4Gb RAM (4x1Gb)

    3Gig switch installed
    Total loading possible 2002Mb
    System stable 2002Mb

    Test 5

    4Gb RAM (4x1Gb)

    3Gig switch installed
    Server.exe “hacked” to make it Large Address Aware
    Total loading possible 3030Mb
    System stable 3009Mb
    VI reported available memory 2860Mb
    VI reported memory in template 2717Mb
    Number of samples 43,502

    A few things to note:
    1) Using 3Gig switch increases loading even when there are only 2Gb RAM installed.
    2) FX-Teleport is not Large Address Aware without the “hack”.
    3) Each instance of VI uses about 16.5Mb, so number of samples loaded will be decreased when many instances are used.
    4) Multiple instances do not share RAM, therefore loading an identical “2nd Violin” template will double the RAM usage.

    I think that it may be possible in future versions of VI for #3 and #4 to be addressed by making VI Multitimbral.

    Loading in V-Stack

    Just for fun ( [[:|]] ) I installed V-Stack. Here are my findings.

    Test 6

    4Gb RAM (4x1Gb)

    3Gig switch installed
    Total loading possible 1940Mb
    System stable 1900Mb

    Test 7

    4Gb RAM (4x1Gb)

    3Gig switch installed
    VStack.exe “hacked” to make it Large Address Aware
    Total loading possible 2900Mb

    I hope that this is of some interest and will inspire other people to get on with their own tests.

    DG
  • M
    M
    Hey Daryl! Seriously cool. I actually pretty much did the same tests myself yesterday (too late which means I'm tired today...) and got similar results including the results with the LMA hacked server.exe!

    Let's hope it was coded for it though (although the switch wasn't engaged in the linker) or we might get random crashes...

    /Mattias
  • D
    D
    last edited
    last edited

    @Mattias Henningson said:

    Hey Daryl! Seriously cool. I actually pretty much did the same tests myself yesterday (too late which means I'm tired today...) and got similar results including the results with the LMA hacked server.exe!

    Let's hope it was coded for it though (although the switch wasn't engaged in the linker) or we might get random crashes...

    /Mattias

    Yeah, that's my next test.

    DG
  • S
    S
    Hi again DG
    Looks great 42000 samples - drooolll

    What does your "my computer->properties" report of available memory
    on startup?

    What do you mean by 'hack' - is this a real program hack or just a switch you set somewhere?

    thx
  • M
    M
    You're setting a switch on the hosts' .exe file so that the OS believes the executable is "LargeMemoryAware". This causes the loader to behave differently and allowing for a larger memory allocation. You need to run a utility (comes with some development packages) to apply the switch and that's what could be called a "hack". The problem is if the host application wasn't really written to be LMA it could lead to hard crashes.

    If these tests are positive however, we could possibly bug the host developers to set the switch officially and thus support the 3GB switch. I've discussed this already with one developer and they had tested this before but with mixed results as they managed to get the system onto its knees during hard stress tests... Ups and downs.

    /Mattias
  • D
    D
    last edited
    last edited

    @Another User said:

    What do you mean by 'hack' - is this a real program hack or just a switch you set somewhere?

    thx

    Mattias' answer explains this better than I could. I call it a hack because you have to get your hands "dirty" opening Command Prompt etc. It's not exactly elegant.

    DG
  • M
    M
    last edited
    last edited

    @DG said:

    Mattias' answer explains this better than I could.

    For the record..what the humble Daryl tries to say here is: "Mattias explains this better in English than I could do in Swedish" [:D]

    /Mattias
  • D
    D
    last edited
    last edited

    @DG said:

    Mattias' answer explains this better than I could.

    For the record..what the humble Daryl tries to say here is: "Mattias explains this better in English than I could do in Swedish" [:D]

    /Mattias
    [:O]ops:

    DG
  • N
    N
    Dear Daryl and Matthias,

    Could you tell us the name of the utility?

    Kind Regards,

    Nick.
  • D
    D
    last edited
    last edited

    @nicks@aubergine.co.uk said:

    Dear Daryl and Matthias,

    Could you tell us the name of the utility?

    Kind Regards,

    Nick.

    I have sent a detailed report to Paul and he is currently testing it. I would rather not post the method I used until VSL has looked at it in detail, as it could cause your system to become horribly unstable. However, if you really, really, really want to know, PM me.

    DG
  • D
    D
    OK, just a small update to say that in spite of trying my hardest I can only get FXT to crash when I load more than 2.85Gb samples and it seems to be stable even with the memory this full.

    I am about to test Gigastudio and VI on the same machine and later will try K2 and VI, just to see if they like to play. I already know that GS and K2 are not really fond of each other [:'(]

    I will update the tests to include 3Gb memory when I have time to take the top off the PC and mess around again...!

    DG
  • R
    R
    Hey DG for us 'low-tech hacks' - could you tell me how to get my 4 slaves set up for this 3 gig switch. What is it exactly (soft or hardware). The .exe you mentioned.

    My 4 machines all have 2 gig ram but range in age form 2 months to 2 years old (2.7 to 3.2 ghz).

    Thanks for any help you can be on this.


    Rob
  • C
    C
    Looks very promising - still awaiting the 3GB machine test. Anyway this looks like a releatively cheap way of getting more milage out of each computer.
  • D
    D
    last edited
    last edited

    @Rob Elliott said:

    Hey DG for us 'low-tech hacks' - could you tell me how to get my 4 slaves set up for this 3 gig switch. What is it exactly (soft or hardware). The .exe you mentioned.

    My 4 machines all have 2 gig ram but range in age form 2 months to 2 years old (2.7 to 3.2 ghz).

    Thanks for any help you can be on this.


    Rob

    It is all software based.

    The switch is an alteration to the boot.ini file in order that Windows allows you to use up to 3GB for applications and 1GB for Windows as opposed to 2GB each. Even with only 2GB RAM there can be an advantage in using the switch (for me it was around 200MB of samples), but it can make the system unstable. I have sent all details to VSL and they are testing at the moment.

    In the meantime, let me know what host you are intending to use on your VI machines.

    DG
  • M
    M
    I added a small comment on the issue in the VI and memory configuration thread as well... [[;)]]

    /Mattias
  • D
    D
    last edited
    last edited

    @Mattias Henningson said:

    I added a small comment on the issue in the VI and memory configuration thread as well... [[;)]]

    /Mattias

    I was hoping you would, to save my typing fingers [:)]

    DG
  • M
    M
    Daryl, do what you do best and get on with that link I forwarded instead... [[;)]]

    /Mattias
  • D
    D
    last edited
    last edited

    @Mattias Henningson said:

    Daryl, do what you do best and get on with that link I forwarded instead... [[;)]]

    /Mattias

    I'm on the case already..................!

    DG
  • M
    M
    Haha, I knew it! You have way too much spare time... [:D]

    /Mattias
  • D
    D
    last edited
    last edited

    @Mattias Henningson said:

    Haha, I knew it! You have way too much spare time... [:D]

    /Mattias

    No, I just avoid work wherever possible and with the builders in, you can't expect an artiste of my calibre to work with noise going on [[:|]] Well actually, I haven't seen them for over a week, but they might come in today and I would have to start some work only to have to stop because I've got a headache.

    In any case I've still got a few DVDs that I haven't watched yet, so it must be time to stop work for the day.

    DG