• Register
  • Help
Results 1 to 10 of 10

Topic: Setting polyphony straight

Share/Bookmark
  1. #1

    Setting polyphony straight

    Regarding Gigastudio polyphony we need to set some things straight. First of all, your CPU will be the main factor determining polyphony. No matter how fast your Hard Disk is, or how much memory you have available, Gigastudio will use the CPU to mix the samples together into the stereo output(s) which you finally hear from your speakers. In order to have a chance to reach the claimed \'160 voices\' mark, you will definitely need a 700-800MHz processor.

    As a reference, a measurement done on a Celeron 450MHz (overclocked, was 300MHz originally), Gigastudio reports approximately 1% CPU usage per voice. So, with 50 voices it consumes app. 50% of the CPU just to mix the voices together into a single stereo output. Assuming that an 800MHz system has exactly double the processing power, you should -in theory- be able to reach 100 voices with 50% CPU usage. This is all just theoretical since there are many other factors which come into play and may severely affect performance.

    Reccomendations: Aim at anything above 700MHz if you want to make good use of Gigastudio\'s permittable polyphony. If you plan to use Gigastudio alongside software effects/sequencers and still want high polyphony and minimum performance problems, aim at nothing less than a 900MHz system. I would recommend higher figures but I also understand that CPU prices are unreasonably higher beyond these levels. For those more technically minded, opt for fast 133MHz (PC133) RAM. Do avoid the current crop of Celeron CPUs if you can, since they work at a 66MHz Front Side Bus (FSB) speed (i.e. the speed that the CPU communicates with the motherboard). The result is decreased performance no matter how good the rest of your system is. Technologically speaking, the best performers out there are the new AMD processors, namely the Duron (low-end) and the Thunderbird (high-end). They are unbelievably cheap and happily outperform their Intel counterparts. Make sure your software/hardware has been confirmed to work flawlessly with the AMD CPUs *and* their available motherboards since these are different from the Intel ones. Alternatively, wait until the end of November. New CPUs, motherboards and memory technologies are about to be introduced very soon. Not that you need to jump on the new stuff as soon as it comes out, but do expect to see significant price drops in the existing parts.

    Aside from the CPU, real-world polyphony also greatly depends on the type of instruments used. And this is why we sometime see posts of users with less powerful systems claiming higher polyphonies than users with better systems. There are two main scenarios:

    1) The instrument has very long samples. This is the case with the BIG piano libraries (among others) out there where each sample is long & unlooped so as to capture the natural decay of the real instrument. The samples are too big to fit in RAM so they have to be streamed from the Hard Disk on demand. Polyphony is heavily dependent on the Hard Disk *sustained transfer rate*. In other words, how many Megabytes per second your Hard Disk can reliably and constantly fetch from the disk heads to the main memory. High polyphony is hardest to be achieved under this scenario.

    2)The instrument has short samples, possibly looped. Samples can fit in memory (i.e. cached) and need not be called from the HD every time you hit a new note. Assuming the CPU power is abundant, Gigastudio can quickly use the samples from memory and reach the \'160 voices\' mark with no sweat!

    I hope this somewhat explains the claimed polyphony mayhem happening in the forum. In addition to all the above, there is a huge number of technical issues and system configuration options that affect Gigastudio performance. The way a given system is set up and configured plays an important role in the \"juice\" you get out of it.

    And this brings to my mind the hot issue surrounding Nemesys and their products. Frankly, I do not understand the insistence of some users of the forum in defending Nemesys. To my eyes, Nemesys came up with a great concept but built upon it with very poor skills both in development and support. Gigastudio has been released with a trully substandard quality factor. As another user has already mentioned, it should have been still in the alpha stage (i.e. premature internal testing phase). The v2.01 patch is still labelled BETA which essentially means two things:

    * The current *official* release of Gigastudio is the one you originally bought, with NO patches applied. Yes, this is what Nemesys sell as their latest Gigastudio version. Just count the bugs.

    * The v2.01 patch being in development for so long means that Gigastudio is so full of problems that the Nemesys workforce is incapable of addressing even now. In other words, you bought an ALPHA-quality piece of software which is now in BETA phaze.

    On top of that, Nemesys are now working on Windows 2000 support! This is a complex and demanding procedure taking in mind the low-levelness of Gigastudio and the different architecture of Win2k. This means that Nemesys are dedicating resources (i.e. time + money) working for the Win2k compliant version, resources which could have been used in producing a SOLID Windows 95/98/Me release!
    Needless to say that Win2k is still far from an ideal platform for audio, considering the current software/driver support from other vendors. It will become one eventually, but give it another year at least.

    And of course, let\'s not forget Nemesys\' commitment for Gigasampler support! They took all the Gigasampler users\' ideas & comments and implemented them in a new & expensive product with a different name, leaving Gigasampler behind. I wonder when, if ever, Nemesys are going to keep their promise and find the resources needed for providing updates for Gigasampler owners.

    Nemesys obviously have limited/weak development & testing departments, and hearing all these promises, I strongly believe they\'re trying to fit too many apples in the basket. I won\'t even go on to talk about their user/software support. I believe users in this forum have given a clear view on that already.

    Closing this very long post, I want to advise you guys to gain a little more self-respect. If you shell out your hard earned money on a product of this quality from a company with this kind of support & understanding... and simply tolerate this, then you are only calling for more! And of course you would fully deserve it.

    I am only saying this because I do not like one bit the way things have headed so far, and I do not plan to pretend I like it either. I keep my head high and I will bite the hand that slaps my face, you may as well turn the other cheek

    Hope I have passed some messages through this post.

    Regards

    food4thought

  2. #2

    Re: Setting polyphony straight

    You said something about a better GUI for Gigastudio.

    I got a friend who works on GUI design stuff and he made a rather cool version of the Gigastudio GUI.

    Check it out at: www.geocities.com/amivag/gstudio.jpg

    He put it together in one afternoon. Let us know what you think.

    food4thought

  3. #3

    Re: Setting polyphony straight

    Interesting post there... I agree with you on many points. I DO have trouble with GStudio still (when I max out its performance, going over 110 voices or so with my P3 667 with UDMA Maxtor HD), it tends to screw up its streaming device or whatever it is and play wrong samples, keep notes playing etc, so I have to RESET the HARDWARE on the RESET button to get back to work. I am not so sure about the CPU being more important then the HD - I don\'t think I ever go over 50% CPU usage even when I hit 150-160 voices, but my HD\'s seem to not being able to catch up - I get clicks and some notes are delayed. Sometimes the whole thing is delayed, hanging for a moment until all samples gets cached in and ready to play.

    Speaking of their \'workeforce\' I have no idea how many people are working on GStudio - but frankly, judging by the speed of development I can\'t believe it\'s that many programmers - I\'d say two - maybe three? Regarding Gigasampler I wouldn\'t mind that they didn\'t support Gigasampler anymore, if they would just have a cheaper upgrade charge than is now.

    Anyway - if they manage to get Gigastudio really stable - as stable as I believe Gigasampler 1.61 was - I\'ll be happy Oh and then they should concentrate on a better GUI (although GStudio is much better than GSampler it is still not perfect), more stability, faster bootup! and an integrated editor and drop the pretty bad and useless NFX system. Afterall GStudio is for \'powerusers\', isn\'t it? Who would want to use cheap, not-so-good sounding effects instead of using outboard gear or software DX plugins for effects?

    Regards,
    Simon

  4. #4

    Re: Setting polyphony straight

    <BLOCKQUOTE><font size=\"1\" face=\"Verdana, Arial\">quote:</font><HR>Originally posted by food4thought:
    Regarding Gigastudio polyphony we need to set some things straight. First of all, your CPU will be the main factor determining polyphony.
    <HR></BLOCKQUOTE>


    You must be carefull with such statements. The CPU is as important as a fast HD subsystem. There is an entire chain of components at work here; weakness in any component will become a limiting factor.

    As an example: if your HD system is not configured properly with DMA, a lot of the transfer burden will be moved to the CPU. Consequently, you will not be able to do much without a *very* fast CPU. The problem is not the CPU however, it is the HD system.

    <BLOCKQUOTE><font size=\"1\" face=\"Verdana, Arial\">quote:</font><HR>
    As a reference, a measurement done on a Celeron 450MHz (overclocked, was 300MHz originally), Gigastudio reports approximately 1% CPU usage per voice. So, with 50 voices it consumes app. 50% of the CPU just to mix the voices together into a single stereo output. <HR></BLOCKQUOTE>


    It is not in general possible to make such extrapolations. There are too many variables at work.

    <BLOCKQUOTE><font size=\"1\" face=\"Verdana, Arial\">quote:</font><HR>
    For those more technically minded, opt for fast 133MHz (PC133) RAM.
    <HR></BLOCKQUOTE>


    In fairness, one should add that certain 100MHz FSB systems still outperform 133MHz FSB systems on memory throughput and latency. Currently this is still the rule rather than the exception (unfortunately)!


    <BLOCKQUOTE><font size=\"1\" face=\"Verdana, Arial\">quote:</font><HR>
    Polyphony is heavily dependent on the Hard Disk *sustained transfer rate*.
    <HR></BLOCKQUOTE>


    Untrue, and very misleading. Although there is some dependence, GSt does not do sustained transfers. Rather, the main dependence is the *random block read rate*, which is mainly determined by the HD seek/access time.

    As an example, a fast 20MB/s sustained transfer rate HD may only have a random block read rate of 8MB/s (access time around 9ms), which limits polyphony to about 100 voices (one can calculate and measure this).

    When plugging this drive into a certain plug-in IDE card, the access time will go up to 30ms (because the card adds extra latency). While sustained transfer stays at 20MB/s, random block read rate drops to about 1.5MB/s, which means polyphony drops to about 25 voices. Once again, one can calculate and measure this.

    <BLOCKQUOTE><font size=\"1\" face=\"Verdana, Arial\">quote:</font><HR>
    And this brings to my mind the hot issue surrounding Nemesys and their products.
    <HR></BLOCKQUOTE>


    Interesting your comments about alpha/beta test software being released to the public. However, everybody is doing this. You will be hard pressed to identify any company producing high-performance software who do not do this. As a matter of fact, you will be horrified to learn of the myriad of bugs present in the software (firmware) on board common consumer electronics such as mobile phones and CD players (not to mention in the telephone exchanges). Some of these are life-threatening.

    What you are describing is indeed a problem, but not only with Nemesys. It is with the industry as a whole; and with the time-to-market pressure that causes it. If people built bridges and buildings the way software is built, we would have had serious problems.

    Also, a lot is dependent on the operating system - any software pushing the computer *this* hard will show up the cracks in the underlying OS (and let us not go into the buggy stuff that Microsoft passes off as software).


    <BLOCKQUOTE><font size=\"1\" face=\"Verdana, Arial\">quote:</font><HR> On top of that, Nemesys are now working on Windows 2000 support! This is a complex and demanding procedure taking in mind the low-levelness of Gigastudio and the different architecture of Win2k. This means that Nemesys are dedicating resources (i.e. time + money) working for the Win2k compliant version, resources which could have been used in producing a SOLID Windows 95/98/Me release!
    <HR></BLOCKQUOTE>


    I do not honestly believe that it is possible to build *anything* solid on top of 95/98. Once again, any system is only as solid as its weakest link, and Windows is not a very hot foundation.

    By comparison, many people have been clamouring for Win2k support, mainly for the exploitation of dual processors.

    <BLOCKQUOTE><font size=\"1\" face=\"Verdana, Arial\">quote:</font><HR>Closing this very long post, I want to advise you guys to gain a little more self-respect. If you shell out your hard earned money on a product of this quality from a company with this kind of support & understanding... and simply tolerate this, then you are only calling for more! And of course you would fully deserve it.
    <HR></BLOCKQUOTE>


    Once again, true. But you have to apply this consistently, which means that you also have to take a stand against the substandard performance of Windows, and most other software. Until there is major pressure on the *entire* software industry, you will *always* have problems like these.




    [This message has been edited by cc (edited 09-24-2000).]

  5. #5

    Re: Setting polyphony straight

    I think everyone who adds their $.02 is just trying to help. We all need it a little.

    It would be nice if Nemesys or someone would post those little tricks people have done to their systems to help out with performance using GS. Then, users who are having trouble could go through that list of items and try different things out. I\'m sure newbees have no where to start, and it might eliminate the repeated questions here.

    But who will do this????

    Ron

  6. #6

    Re: Setting polyphony straight

    Donnie,

    I for sure can\'t get 160 voices and everything running smoothly. Usually when I get above 110 or something I get a few breakups in sounds and reaching for 160 results in delays in the whole playback sometimes. I\'d say 80-100 without any trouble. I have a P3-667, 133mhz RAM, Vcache setting right, 320MB ram, a fast Maxtor 7200 RPM UDMA 66 drive running on an internal UDMA 66 controller, and an IBM UW 10.000 RPM SCSI disk.

    Regards
    Simon

  7. #7

    Re: Setting polyphony straight

    ditto. There are just too many variables that makes a good giga system. my old machine was Athlon 550 with MSI motherboard (the original one), can get 80 poly max... more than that will get occasional breakups. When I upgraded it to Thunderbird 900, and replaced motherboard to ASUS A7V, I get 160 full, with all kind of samples including the gigapiano, steinway c... All other peripherals unchanged.

    And yes, I\'ve tried so many different VCACHE settings on the older PC and the best one got my the 80 poly.

  8. #8

    Re: Setting polyphony straight

    Simon,

    What is your vcache set to? If thats not the problem then there are a couple of other things that we can check next...

    Donnie

  9. #9

    Re: Setting polyphony straight

    Actually no...you do not need a PIII 700-800 to come close to reaching the max polyphony. I have a PIII 550 with 256 megs of ram and get 160 everyday. All you need to do is set the vcache min. and max to around 48000. I thought we had been through this before???
    It always upsets me when someone comes on and makes statements like the one above when they don\'t really know how to maximize their own system much less anyone else\'s.

    Just my two cents...

    Donnie

  10. #10

    Re: Setting polyphony straight

    Could it be that you have to many NFX effects on to many tracks? I\'ve run into problems before with that and I too have a 667 MHz P3, only 128 megs ram. I don\'t know if that\'s a polyphony problem though. My CPU gets to around 86 percent if I have about 13 tracks with reverb effects on every track and it simply will not record anymore and I get \'dropouts\', but if I take some reverb off some tracks it lowers my CPU big time. I just reroute the same midi channel to a different track and it lets me use the same reverb only if I\'m using the same sound though. I don\'t know if that will help or not, but I just thought I\'d throw that in.

Go Back to forum

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •