Vienna Symphonic Library Forum
Forum Statistics

181,986 users have contributed to 42,198 threads and 254,643 posts.

In the past 24 hours, we have 4 new thread(s), 8 new post(s) and 48 new user(s).

  • G5 with 16 Gb Ram or Server? New Setup.

     Hello Vienna (and the rest of the world of course)!

    This is my first post and question  here and it is about the beginning - in other words: the new hardware setup i'll need for my lately purchased Symphonic Cube. In fact, I made benifit of the christmas bonus, wich is as far as I know the best deal since VI release (correct me, if I'm wrong). If I had already purchased it befor I would definitely buy a second cube ; ) .... (hope you early birds have some sense of humor).

    Anyway, back to my question:
    What is the best sollution for a dedicated VI Player?
    Nowadays I use a Mac G5 2 GHz double core Processor, 6,5 Gb ram, motu 828, ... for Logic and an older Pentium IV single core with Creamware Pulsar 2 (2xAdat I/O) for Gigastudio.
    I will order an additional computer only for Vienna instruments and spend money on a new soundcard sollution for the whole setup.
    Here are the options I think about:
    1. Buy a Mac Pro as new DAW for Logic 8 and use the G5 (ram expanded to 16 Gb) for VI only. Will the processor power be sufficiant for loading the ram to the limits? I already noticed, that VE is not 64 bit on mac, yet, but is there a chance of estimating this befor the release? Also I heared you can already use more ram with multiple instances of VE. Did anyone load up to 16 Gb on a G5 in a stable way? Would you need a raid controller for the G5? In other words: Please tell me, if that wont work, as I imagine, or if it would need modifications I didn't mention.

    2. Just as in "1." , but vice versa, so keeping the G5 as DAW and using the Mac Pro as Vienna Sampler. This wouldn't make sense, though, in case option 1 works flawlessly, would it?

    3. Instead of a Mac Pro you could buy a server (Dell Poweredge, or the supermicro configuration suggested by Herb in another post. What makes this sollution interesting is the fact that servers seem to be fitter for the future in terms of upgrade (64 Gb ram, more CPUs). But I have no server experience at all. What are the flaws of this option?

    In all of the three cases I'd want to keep the old PC for Gigastudio streaming audio via adat back to daw+ MidioverLan.

    This leads me directly to the second part of my (big) question. What about audio?

    1. RME FIreface 800 as frontend and mic sollution + 2x RME HDSP(e) 9652 (DAW and VI) OR 2x M-audio Pro Fire

    2. Instead of fireface an Apogee Ensemble (I know this is not place to ask what is better - RME or Apogee - but I wonder how they work together in one DAW? And what about latency and jitter? I read that with RMEs steadyclock you wouldn't need a wordclock).
    Another point I never really got is this info from RMEs homepage:

    "The Fireface 800 is equipped with SteadyClock(TM), RME's unique sync and clock technology. With this, the device becomes a sync reference for the whole studio. SteadyClock refreshes clock signals, removes jitter, and takes permanent care of optimal conversion quality, thus guarantees a sensational sound quality, completely independent from the reference clock's quality."

    Same for the Apogees "Clocking used by Big Ben" in ADX 16 or “Intelliclock” in ensemble. Does that mean with one of these you can buy the cheapest Adat sollution possible (Pro Fire) and still get the same sound (due to one clock reference) as if you had an "RME only"-system?

    3. Now it starts getting crazy (and very expensive): What about RME Madi. I won't lay out the options in detail, cause I already bored you for such a long time (sorry!). Only that: Has anybody realized a system on madi? Something like: Daw - HDSP Madi - RME ADI 648 - HDSP 9652 - VI Server + some AD/DA unit for analog I/O? That seems to be perfectly upgradable (in theory).

    My (defenitly) last question: What about MIR? I know you guys won't tell us befor it's ready for the market. But is there at least some kind of preference how the audio stream will be acomplished? I really want to have a system where mir can be integrated, so any hint would be apreciated.

    Thanks for your patience and keep up your great work,

    Fritz von Flotow.

  • Welcome to VSL, Fritz.

    I'm not qualified to answer most of these questions, but I do have an opinion on one matter.

    "Will the processor power be sufficiant for loading the ram to the limits?"

    No. I have a slightly less powerful G5, and I bumped it up to a full 8 GB. from 6.5 GB. My performance did not improve. It was a CPU issue, plain and simple. In fact, going to 6.5 from 4.5 didn't help all that much either. Of course you can load more sounds at a time. but I'm talking overall performance. There were no more voices, no smoother operation, no fewer CPU overloads. You may *load* all the memory, but playback hits the same ceiling. I may have a couple of posts here and elsewhere claiming improvement after my memory upgrade. In hindsight, I was wrong.

    A MacPro will do most of your heavy lifting. The G5 is a great candidate for noding or respectable sample playback. I'd be cautious before putting too much money into a G5. It's older than we might realize, even if it still looks cool.

    By the way, I wouldn't make any serious MacPro decisions until MacWorld in a couple of weeks. (And NAMM is soon too.) But I'm sure you knew that already.

  • welcome fritz, a few comments ...


    i'd rather consider to setup 2) than 1) the noticeable difference for the DAW would not be as big as for the sampling machine (altiverb instances might have to be added to the math)


    there is no big difference with server boards - except you need a 64bit OS for more than 4 GB RAM and XP64 is almost the same as XP32. do not forget server RAM, especially if you want more than 16 GB will become the most expensive part of the hardware.


    from what i heard and can say the Pro Fire works better on win than on mac (and i'd assume you should put each on a seperate FW-controller ...)


    but why not looking after an apogee - best choice for a mac IMO regarding latency - in any case a solution which sits directly on the PCI bus instead of having again FW between (adds unwanted CPU load)



    and remember: only a CRAY can run an endless loop in just three seconds.
  • Thanks for sharing your ideas.

    Christian, I didn't know firewire added more CPU load, but in some way I always had a better feeling towards pci audio - now I know why! That makes the pro fire worse than RME HDSP 5628, doesn't it?

    I completely agree with you in terms of apogee. That is great indeed! The only problem is with multiple PC/Mac sollutions the costs explode when you use them to stream over many adat/AES EBU channels. Also: Apogee Ensemble works on firewire, too. Apogee is just missing a costeffective way of expanding digital I/Os (or am I missing some thing here?).

    The minimum configuration for 8 Track recording + 3 x Adat for VI Mix will be:


    APOGEE ROSETTA 800 (2-5 Out for monitoring + 8 In for analog In -3000 Euro

    + Premp (I never had a dedicated preamp but I guess it's nonsense to take a cheap one, if you spent so much in the other equipment??)

    + Wordclock to improve the ADAT signal (RME HDSP) - 1500 Euro 

    + 2 x RME HDSPE (Host and slave) - 1000 Euro

     Together: roughly 7000 Euro without the second daw! Whooo!!

    (Has somebody got a similar setup and in addition an RME HDSP 5628 (or similar) for adat? Does it work fine?)

    This price makes me thinking abut RME more carefully.

    Considering the mix: Yes, I want to use altiverb. But I thought about using it on the DAW, not on the VI Machine. That is why I want so many ADAT I/Os - to be able to mix all instruments seperatley along with composing and be able to add other elements or bounce ready instruments inside Logic.

    Both of you agree that option 1) is worse than 2). So what about 2) and 3)? Is server Ram more expensive than ram for a Mac pro? I have no idea about servers, so I would appreciate every comment from somebody, who already built such a system.


  • the most expensive RAM yu can think of is the one ordered from apple, up to 8 GB RAM some motherboards allow (cheaper) unbuffered memory, but generally i'd say ~100.- EUR per GB - the more GB you need in a single machine the more expensive is certified memory and you will notice a significant step at 16 and 32 GB


    and remember: only a CRAY can run an endless loop in just three seconds.
  •  Fritz had a really good question.... Has anyone successfully loaded anywhere close to 16 gigs of samples on a G5? If so, is it functional?


  • i doubt there are too many users out there having 16 GB in a G5 and IMO this wouldn't make too much sense for sample streaming because you're getting already close to the limits of the machine as such with 8 GB (loaded _and_ playing) ...

    however with a macPro it seems possible though you have to design your setup carefully until a native 64bit version of VI/VE will be available and then i'd assume you need to pay high attention to harddisk throughput.


    and remember: only a CRAY can run an endless loop in just three seconds.
  • Ok, now that it seems like putting 16Gb into a G5 seems like a waste of money. And since I will use the vienna slave only for that purpose and mybe in the future other sample players like Gigastudio a server seems to me like the better sollution - and a little bit less expensive I think.

    My idea of a future proof server would be one, where you can install 16 Gb in 4 Gb modules, wich would give me the option to upgrade at a later time to 32 Gb, wich should be enough. What I don't know is how the other hardware should be like in order to be as cost effective as possible but still enabling me to use the full power of the 16 Gigs and play many voices in fast speed at the same time.

    To be more specific:

    1) What about the motherboard? Herb's said his new Slave would be a supermicro. But what about Dell, or cheaper ones like pyramid? Can somebody give advice, who recently built the same thing?

    2) What about the processor(s)? Herb's system was to be quad, but christian said Quads would be wasted money. I don't get that. And how many GHz /singl enough, or double?

    3) What about HDs? Sata II, SCSI, SAS, Raptor ...? Raid, or not and if so, wich one? 

    4) Anything else, that could be important?

    I'm going to spend a lot of money and I siply don't want to spend it on a system, that turns out to give only half the power. On the other end I neither want to spend money in the slave's power that never gets used...


  • I think that the Fritz's question could be very useful.

    When I bought my first PC I followed the vendor advises and it was a nite-mare. Then  I spended a lot of time through FORUMs and my second one was very good (4 years of heavy working without a single problem). There are many good example all around about PCs working set-ups but very few news about Workstations and Servers. As they cost a lot of money more specific examples of working set-ups should be appreciate.

    I PERFECTLY KNOW that 2 computers made of the same pieces could give very different behaviours and so nobody would be guilty for suggesting!!

    For example, I don't understand why 2 core CPUs and 4 core CPUs with similar speed differ so little in price (maybe because they differ so little in performance, but if it is so, why they made them,? Just to text them on our shoulders??)

    Another idea, would it be a good idea to buy the same processor and everything you can find in a MAC Pro and instead to build a Xeon W XP64  PC saving a lot, (some 1000 euros) for buyng more Vienna Intruments?


  • fritz, it looks like many of the pyramid machines _are_ supermicro (what else), you will find supermicro systems in many OEM products ...

    i personally don't use dell because they don't provide sufficient information about the used motherboards and i don't like surprises too much.


    regarding 2) i mentioned 4-core CPUs (along with intel 5000 chipsets) don't add the awaited performance increase compared with 2-core CPUs, but clearly a 4-core system (2 proc x 2 cores) is more performant than a 2-core system (1 proc x 2 cores)


    for sampling only 2 x 2,66 GHz XEON should be already sufficient, even with 16 or more GB RAM - how much the harddisks will turn up as bottleneck with 40 or more instances has still to be found out (chipset, drivers, configuration, harddisk model, raid modus adds to the overall performance).

    SAS and SCSI (up to 15.000 rpm) is really expensive, flash not yet available with the needed capacity, so sATA II looks like the preferred choice, again raptors are more expensive than 7.200 rpm drives but add significant performance (eg. average seektime is 4,5 ms whereas normally it is about 8.5)



    and remember: only a CRAY can run an endless loop in just three seconds.
  • For what it's worth, I cannot play Chrstian Kardeis' Special Edition demo on my G5 Dual 2.7 (8GB ram) without overloading the processors at the piece's most heavily orchestrated points.  Frustrating.

    I am about to build a multi-cpu system to hopefully run a nice big template with lots of flexibility, but I have my doubts, as I never seem to get the jump in performance that I expect when upgrading.   Maybe the jump to MacPros with Penryn processors (or comparable PCs) will be different.  Well see what happens this month

  • aplanchard, this really should be irritating - the demo runs definately on much less powerful computers without problems (AFAIR actually on a 1,8 GHz iMac), i'd recommend to check your machine and/or setup ...


    and remember: only a CRAY can run an endless loop in just three seconds.
  • Since we’re all having a bit of fun talking about the new setups we’re getting I thought I’d throw my idea into the mix.  I currently have a G5 dual 2Gig. (Please note Plowman’s comments on older G5’s). I intend to use this as a dedicated reverb machine. It’ll be using a software host with multiple instances of Altiverb & Wizzoverb. There will be a Presonus Lightpipe 32-channel ADAT interface to connect with a similar audio interface on a Mac Pro 8-core with 16GB RAM. 24 channels of audio will be used to return reverb to the Mac Pro. 8 more channels (one ADAT connector) will be used to collect the 8 digital outputs from a Focusrite 428 mic pre with an A/D card.  That Mac Pro (I’ll call it Mac Pro1) will have Logic 8, a dual UltraSCSI card attached to both a RAID0 array and another individual SCSI drive – all 10K RPM – to hold String & Brass samples.  This computer will be used for sequencing and playing back of multiple standalone VE’s and standalone Kontakt instances. The standalone sample playback instances will loopback to Logic through a RME MADI interface (64 channels).  Additionally, Mac Pro1 will have an expansion chassis to hold extra PCI cards – MADI interface(s), extra FW card for the Presonus Lightpipe interface, SCSI card and TC Powercore X8 for TC reverbs & Sonnox Oxford plugins.  There will be a second RME MADI interface on Mac Pro1 that will connect with a third MADI interface on Mac Pro2. Mac Pro2 will be a 4-core with 16GB of RAM that will host standalone VE’s and instances of Kontakt for playback of woodwinds and percussion samples. Not sure which drives to use for this yet. Thinking SAS.  I’ll be using a Lucid GenX for wordclock and MIDIoverLAN for communication between the Mac Pro’s.   So: Mac Pro1 will have 64 channels of internally loopedback sample playback and 64 channels of external sample playback from Mac Pro2. All reverb and other fancy processing will be done either on the G5 or the Powercore X8 card.    UN-reality check: Waiting for new Mac Pro’s (C’mon MacWorld!), Sonnox promises that their plugs will work on Intel shortly, TC is supposed to deliver the X8 pronto. Presonus should deliver the Lightpipe any day now. Note: The Presonus is necessary because the M-Audio Lightbridge is not supported as an aggregate device. I will need to aggregate the audio interfaces on Mac Pro1 since there will be three of them.  None of the M-Audio Pro Tools-ready or Digidesign interfaces support aggregation to the best of my knowledge.

  • this sounds like we have more than a bit of fun ;-) ...


    SCSI host adapter in a breakout box - i wouldn't do that, because those breakout boxes have limited throughput (usually 1 PCIe 1x spreads to 5 PCI slots), i'd rather consider to use a *real* PCIe or PCI-X SCSI host adapter to stick it directly in the macPro. i have also seen a IOI adapter with SCSI + FW800 on a single expansion card.


    additionally: any new penryn motherboard should be able to access SAS disks (which is the TLA for Serial Attached SCSI) which would be the more modern and performant (and internal) solution to hold sample data (and finally less expensive than a SCSI host adapter + SCSI disks + enclosures + a lot of cables)



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

    Yes, I also had to freeze some percussion tracks while creating this piece. I did this on my G5 Dual 2,5 (6 GB Ram) and without the frozen tracks also my machine cannot playback this piece.

    The irritating thing is that an Intel iMac or an Intel macmini with 2 GB Ram plays back the demo without all the hiccups and overload messages known from the G5 :-(



  • Hi Christian, 

     Thanks for the info.  Perhaps it is not unreasonable to expect a big performance boost when I upgrade my setup.  Fingers crossed.

  • CM, Thanks for the comments about my proposed new rig. Yes, I know that there will be some ‘adventure’ in getting this setup working. I do plan to get one Mac Pro first to figure out the issues then get the 2nd Mac Pro once the first is stable.   

    The reason for the SCSI is that I already own them. All my VSL samples are on a RAID with 300GB 10K SCSI drives with my current G5 setup.  Putting the SCSI adapter card in one of the PCI slots of the Mac Pro (and not the expansion chassis) is a good idea. It will be dependent upon how happy the RME MADI interfaces for Mac Pro1 are in an expansion chassis. I should look into the possibility of repurposing these drives using SAS. Perhaps you might know this.

      I look forward to hearing any other thoughts about the workability of this setup before I begin purchasing the components.  

  • did you mean repurpose the SCSI disks using SAS? not possible ... SCSI = 68pin parallel, SAS is the same (serial) connector as sATA, so you had to keep them in an external enclosure connected to a host adapter.


    and remember: only a CRAY can run an endless loop in just three seconds.
  • Yes, having never seen a SAS card before, I now realize that that wouldn't work.

    The new Mac Pro came out today & does have SAS - but at a pretty hefty Apple-type price. I might put SAS on one machine and leave the SCSI on another just because of cost.

    The other issue is that I'm waiting for the PCIe expansion chassis to become a viable product. However I could just start with the single Mac Pro for the time being.

    CM, do you see any VSL issue with how I'm vizualizing the loopedback playback on the Mac Pros?

  • yes, SAS disks are in a similar price range than formerly SCSI - they are typically used in high performance servers.


    btw: the RME MADI is also available as PCI express ... and maybe you should ask someone who tried this setup, but possibly the internal routing (VE out - MADI - logic in) can be actually done in the internal RME mixer, so leaving the 64 channels (external in) free for routing audio from macPro-2 to macPro-1


    look at the specs ... Three open full-length PCI Express expansion slots One PCI Express 2.0 x16 slot Two PCI Express x4 slots

    looks like you could even insert 1 SCSI hostadapter + 2 x PCIe MADI (in case the brackets/wordclock modules allow)



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