Vienna Symphonic Library Forum
Forum Statistics

181,968 users have contributed to 42,196 threads and 254,641 posts.

In the past 24 hours, we have 5 new thread(s), 11 new post(s) and 50 new user(s).

  1. Home
  2. General & Hardware
  3. Solid State Storage - what are the implications for VSL systems? Have you tested any yet?

  • Solid State Storage - what are the implications for VSL systems? Have you tested any yet?

    D
    D

    With Apples announcement of their ultra-thin laptop with optional 64G solid stat hard drive, I guess we can soon see the the time when reasonably big drives at reasonable prices will start emerging (<3 years ?) swith all the inherent advantages of lower latency, lower power, higher bandwidth throughput etc.

    seems to me to be the ideal storage solution for samples.

    with this in mind, has VSL started looing at the implications for a typical VI DAW ?

    - with instant access, would the need for pre-loading the early part of samples into memory still be required ?

    - hence, would the limitations on number of articulations and instruments disappear, since no pre-loading were necessary ?

    - hence also would one of the needs for 64bit OS (large memory) access die ? 

    - hence can I stop worrying about memory as the bottleneck ? 

    - would we be able to get a single drive performing more instruments, perhaps doing away with the need for a separate server for strings, winds, brass etc ?

    - would the VSL software have to re-written to obviate the need for pre-laoding of memory ?

    I feel with this technology we may be at a significant turning point in VI systems ... perhaps the biggest change since disk streaming was first conceived by conexant / tascam ?

    It would be interesting to hear your thoughts.  

    Thanks

  • P
    P

    Seek times are getting better by a factor around 10 at the moment. So pre-loading won't be erased with SSD drives, just you could load around 10 times more sound into RAM on the same computer. While I expect SSD to hit the market in 2008, the capacities will be too small for serious use with the whole Symphonic Cube. Prices would have to go down also more than just a bit to really be competitive against conventional HDDs for large capacities. If they ultimately will have their breakthrough, CPU will bcome the bottleneck for streaming and effect processing, but 64bit system will already be all around by then is my guess. The rewriting (if any) for VSl software would just have to allow for lower access times. Actually I'd love to see that there could be different settings applied on the same computer, so that you can have a set of samples which you need the most and fastly loaded on SSD drives, and the less used stuff sitting idle on conventional HDDs waiting to be used (with the settings we are used to now). That would make a smooth transition possible. Don't forget that yet the lifetime of SSD drives is limited in terms of write cycles, SSDs also consume comparably as much power as conventional HDDs when being in use.

    PolarBear
  • kurtK
    kurtK
    last edited
    last edited

    @PolarBear said:

    Don't forget that yet the lifetime of SSD drives is limited in terms of write cycles, 

    PolarBear

    the lifetime looks fine, see http://blogs.guardian.co.uk/technology/2008/01/16/so_how_long_might_the_flash_drive_in_a_pricey_macbook_air_last.html
  • P
    P

    All kind of parts fail sooner or later. And some will fail sooner, SSD won't make expections to that - I've seen recent harddrives fail on people after a few weeks. But yes MTBF is usually stated with around 2 million hours for SSD drives. Consumer products do behave a bit differntly than enterprise products usually though. I wouldn't care too much for it, if it is just for streaming purposes they will durate much longer than in normal desktop use, as well as soon enough we will find ourselves wanting for the bigger capacities availible in 2 years from now. Just like conventional HDDs also did that to us when we thought, wow, 40GB, who will ever need more than 5 of these...

    640k ought to be enough for anyone...  

    PolarBear 
  • cmC
    VSL Team
    cmC

    we have a pre-series model here (64 GB) which _should_ give 150 MB/s and 0,1 ms average seektime ... my assumption was the preload buffer could be divided by 8 .... a first try to divide by 2 was ... modest .... but it was a really quick & dirty hack ...

     

    this type of drive would be great for samples, because we usually have only read access (the problem with flash is always write) so we can believe the MTBF, actually beyond ...

     

    64 GB is only usable for the special edition - for an essential performance test i would need 10 of them ... and the respective number of sATA ports .... i'll wait for a 256 GB model ... something with raided SSD drives is said to come soon.

     

    street price is ~ 600 EUR for 64 gig currently ... not a bargain ... but i'd assume the price will drop quickly

    christian

     

     

    and remember: only a CRAY can run an endless loop in just three seconds.
  • D
    D

     Thanks for your replies everyone.

    Yes I agree at the moment the size of drives are only big enough for parts of the VSL library, and they are not cheap .... but they are not exhorbitant either.  I think if we apply Moore's law, we could have 256 Gb disks in two more years, for the same price a 64Gb disk costs now.

    2 x 256Gb might just be big enough for most of my VSL library, though 3 x 256Gbwould be great.

    The limited write cycles of these drives is not a problem, since we are talking READing samples, not writing / recording to the drives. And perhaps two years down the line the MTBF will have improved. I could see the MTDF a READ only SSD well exceeding a conventional hard drive.

    Polar, I have read that power consumption is much less, but you sound pretty informed on this ......

    Christian, can you convert the improvements in bandwidth and seek time into "instrument efficiencies" ?

     e.g.

    -  if the preload buffer could be reduced by 8, then eight times more samples could be loaded into a given size of memory, right?

            (or to put it another, you wouldn't need 8 times as much memory - a 2 Gig allocation for samples would be the sames as a 16 Gig system)

    - but could the bandwidth of the new drives be able to keep up with this ? 

    Are you also saying the software would have to be tweaked ?

    Has anyone else got view the timelines involved? Is Moore's Law the correct assumption about improvements for memory storage? 

    Thanks andrew 
  • cmC
    VSL Team
    cmC
    last edited
    last edited

    @drew buchan said:

    if the preload buffer could be reduced by 8, then eight times more samples could be loaded into a given size of memory, right?
    right ... but everything in detail has to be found out ... no timeline, no info on type of tweaking or similar available.

    christian

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

    @drew buchan said:

    Polar, I have read that power consumption is much less, but you sound pretty informed on this ......

    To put things in the right perspective as I misquoted it above... You'd need around the same power for 500GB of capacity with an array of 0.5W SSD drives compared to a single 10W  desktop drive at the moment. With capacities going up this will change. So yes, you're surely right that a single drive will consume considerably less power.

    All the best, 

    PolarBear 
  • S
    S
    Now that Intel has 80GB SSD drives for $400 and 160GB SSD drives for a little over $800, this seems like a worthwhile topic to bring back up. Is the pre-load size user-adjustable in Vienna the way it is in some other engines? I'm tempted to get an 80GB drive to try with VSL SE (Plus+Extended). But if I can't reduce the pre-load size, the win is not as great as I'd like, as I'm hitting RAM limits in my template. Thanks!
  • cmC
    VSL Team
    cmC

     i'm still not considering this to be a reasonable solution ... 160 GB is something for say the special edition and a few additions ...

    it would be cheaper to add loads of RAM rather than purchasing such a drive.

    still waiting for a reliable report about latency dropouts (resp. having confirmed there are none) - so far i have 2 unusable samsung SSDs here and refuse to spend more money on pieces just for the hardware museum ...

    christian

    and remember: only a CRAY can run an endless loop in just three seconds.
  • S
    S
    Yes, I could buy more RAM, but as far as I've been able to use VE2, things crap out when I get a little past 2GB on my Mac (Leopard... 13GB RAM). Am I just doing something wrong?

    But a $370 Intel X25M (80GB) with smaller prebuffers might make the RAM I already have much more efficient.

    (What's the deal with the Samsung SSDs?)

    Thanks :)
  • cmC
    VSL Team
    cmC

     on a 13 GB mac (assumably a MacPro - but a somehow uneven value for memory ...) you should at least reach 3 GB for VE without problems ...

    samsung SSDs and all clones / re-branded models have latency spikes every 1 - 4 seconds which makes it unusable for sample streaming ... from the intel SSDs it is reported they are fragmenting when used (should not apply to read access though)

    and remember: only a CRAY can run an endless loop in just three seconds.
  • S
    S
    Interesting.

    Well, I'll wait then! I did manage to run more than one instance of VE2 now, using IAC and Jack Audio. Starting to max out RAM feels good!

    Thanks!
  • F
    F
    last edited
    last edited

    Speaking of Samsung SSD's, here's a promotional video of someone attaching 24 of them to one box, and treating them as one huge 2GB-per-second array.

    Samsung: for when ludicrously expensive just isn't ridiculously expensive enough.