Full polyphony from 1 drive?

[Sam]

For some time now I\'ve been running GSa on my main machine, but I just made the jump to GSt96 and will be purchasing a dedicated machine just as the sampler box. My goal for this box is just a reliable box with full polyphony (96 voices), and other priorities are that it be modestly quiet and as cool running as possible.

Beyond that, I won\'t play games on it, and it doesn\'t need to be the latest technology rocket. My main DAW is a SCSI + IDE system, but this one will be IDE only, and I\'ll put 256 or 512M of RAM in it.

I\'m considering the drive configuration. I\'ll probably go with an IBM 60GXP or two, one 60 gigs or 1 or 2 40 gigs. If I went with one drive could I get full polyphony? Is it useful to turn off the swapfile (win98se)?

The mobo will probably be an asus cusl2, and I\'ll put the cdrom on 1 IDE controller. If I put 2 drives on the 2nd controller, am I likely to get full polyphony from a slaved drive? Is it better to get another IDE controller for 2 drives, like the Promise?

Thanks for the tips folks.
-sam

ps Since folks are likely to ask, I\'m gonna stick with the 815 chipset, since I may put VIA sensitive cards in the box in the future. I\'ve had good experinces with Asus and IBM in the past, so I\'m sticking with these. The sound card will be a Frontier or RME, I dunno yet. I have a few Frontier cards already, and like the solid drivers, midi, and audio patchbays. As for RME, I like the OS support, but it sounded like the GSIF drivers weren\'t fully baked yet and I\'d miss the midi and patchbay.

[carlgt1]

I think if you\'re going to the plunge for a separate machine you should get a separate drive for Giga. I think you can get away with one drive for digital audio recording but Giga is more sensitive in my experience. All you need is a tiny disk for your system drive & swap file (buy a second-hand 2GB ATA33 or 66 or 100 IDE drive) and get a good 30-40GB drive for Giga (I use the IBM Deskstar 40GB which is 8.5ms avg seek).

You probably could get away with a single drive (Nemesys claims the Toshiba Tecra 8200 laptop with one 4200rpm drive was getting 140-150 voices reliably); but I like to have the overhead so I\'m not pushing anything. I kind of wish I got the GS160 though, I run out of voices with big piano pieces!

[Sam]

I wonder if it\'s actually necessary though for full poly (especially with the 96 voice)? I\'m just not aware with my main DAW box that it hits the \'system\' drive for anything while the box is being used for nothing but a sampler. With 256 megs it doesn\'t swap, and the modern drives are pretty durn fast too. If I could spare myself the additional noise and heat and IRQ(and $$ though that\'s minor here) I\'d do it. Remember this box isn\'t really going to be running multiple apps.

Anybody got modern system single drive performance numbers to post?

Thanks for all replies!
-s

[Sam]

> Nemesys claims the Toshiba Tecra 8200 laptop with one
> 4200rpm drive was getting 140-150 voices reliably

Whoa, I missed that line on first read. That\'s awesome and has to bode well for me. In my experience laptop drives are way slower than desktop ones. Perhaps unlike you I don\'t need substantial extra throughput on this box, since it won\'t be streaming out other applications or recording.

thanks again Carl

[gigaDiga]

Hmmm hate to be the prophet of gloom here but I\'ve got a csul-2 and 2 ibm deskstars and can currently only get 120 max polyphony apart from two days when I could suddenly get full 160 without pops and clicks... but they whent past very quickly.

Gigasampler splurts and chokes if you overdo the polyphony. Unlike my old 64 voice sampler which I didn\'t really notice much note stealing goind on Giga just shouts and screams when you try and push it.

My current measure of happiness revolves around a composition I have made which involves two hands of 4 fingered chords on the piano with the sustain pedal held down. It is a rather beautiful and resonant piece which currently lasts all of 7 bars before GigaStudio cracks up and blurts out a tonne of crap noises and notes. One day, when I suddenly obtained 160 voices, I managed to play it through to the end of bar 8 and was really rather chuffed. I should have recorded the song that day. Unfortunately I didn\'t and two days later, and ever since, my GigaStudio was back to breaking point at bar 7. One day I\'ll get this song down.

In the meantime I\'d recommend that you forget 96 polyphony, it\'ll sound awfully like 32 polyphony when you get it going. Go for 160.
Remember 1 stereo note counts for 2 voices of polyphony. So if everything is in stereo then your actual note count will be 48.

Here\'s hoping tascam produce a TRULY optimised DSP-type GSIF sound card. Then all our problems will be over.

[cc]

Have a cusl2c system with IBM 40GXP drive, and quantum fireball AS.

One drive is connected (single) to a mobo IDE (the other mobo IDE goes to CDROM), the other drive goes singly thru a Promise Ultra66. The big trick is to put the Promise card on the so-called \"priority\" PCI slot, and not to use master/slave configurations at all.

I can get full 160 polyphony from either the one drive, or the other.

Gigadiga, if you are using a plug-in IDE controller, try moving it to your mobo\'s priority PCI slot, otherwise try moving your soundcard to the priority slot on this board (make sure priority PCI is turned on in the BIOS) - this should allow the soundcard to service its interrupts much faster.

[Sam]

GD - Sorry to hear you don\'t get full poly. Let me know if cc\'s trick helps. I\'m sure 96 voices will be fine for me because of the way I track. I\'ve already upgraded to 96, 160 was too much $$ for what I needed. Actually I don\'t really need GSt\'s features since I\'ve got good outboard FX and sequencers and mixer and all. I\'m mostly upgrading to get the stupid choke groups for cymbals working, and it will be nice to have faster gig loading. (Like some others here, I hope future GS development focuses more on making a better instrument, and less on the environment.)

I wonder what the deal is with note stealing? The way it works on some of my gear is to never steal the lowest note from a patch, and beyond that it seems to be a heuristic balancing oldest notes with low envelope amplitudes, which can work pretty well. My nord modular can steal voices very non-intrusively, since the oscs are free-running and the env\'s just retrigger from their old amplitude without zeroing. Can\'t really do the same thing with samples tho.

cc, thanks for the tips. I\'ll let y\'all know how it goes when the box comes together.

[carlgt1]

Have you tried running hard-drive benchmark utils such as dskbench.exe (at prorec.com) or HDTach to see what MB/sec and seek times you are getting? I did and discovered that for some reason my on-board IDE for disk 1 was very slow.

I went to BestBuy and picked up a $40 PCI Card for ATA100 (disable MB) and my hard drives scream now and polyphony is great (no problems at 96, I only have GS96). I am also using IBM Deskstar drives.

Also (as an experiment) try to turn your video acceleration off, and USB ports if you don\'t need them.

[Simon Ravn]

Many of todays drives comes with some \'acoustical\' feature to make the drive more silent. Unfortunately it also makes it slower at seeking. Some drives ships with this option set to QUIET (slowest), and when I changed this to the fastest instead, I actually noticed that notes that dropped out earlier now are playing fine. So if you have a newer Maxtor or IBM drive you might want to check that out and find a utility on the vendors homepage to be able to change the settings...

[dnortana]

Sam,
I was getting full polyphony with only a single IBM Deskstar 75GXP. But with some intermittent clicks and pops, which I thought might be hard drive related.

So, I added a second drive on a second IDE channel (the ASUS A7V I have supports 4 ATA100 devices, before activating the onboard Promise stuff). No difference in performance, that I can tell.

I think the occassional clicks and pops are related to the immature GSIF driver for my Hammerfall 9652, or perhaps my PCI slot choice (which I just changed recently). But then again, I\'m outputting 24 tracks of audio all at once....
Regards,
Trond.

[carlgt1]

>channel (the ASUS A7V I have supports 4
>ATA100 devices, before activating the
>onboard Promise stuff). No difference in
>performance, that I can tell.

I take it that you were trying to avoid the onboard Promise controller for some reason? I\'m using them and my audio stuff flies, 38MB/sec in HDTach! I\'ve heard rumors of data corruption (with VIA but not Promise or the other way around?) but everything has been fine. Plus, once my GIGS are loaded it\'s all \"read only\" anyway.

[Simon Ravn]

No IDE harddrive really reads 38MB/sec. off the disks in real world situation. That is probably the max possible throughput, which seems to never occur.... Just try doing the math here - if your drive really did 38MB/sec. (real life, including a lots of seeks back and forth all the time), you\'d be able to optain roughly 860 voices of polyphony.... http://www.northernsounds.com/ubb/NonCGI/images/icons/smile.gif

[cc]

The confusion with transfer rate sets in because people are looking at the wrong parameter.

Giga does not do sustained transfer, it does random block reads. Some programs (like sisoft sandra lite) gives a nice indication of measured access time and random block read rate.

Doing the calculation from that last, gives a pretty good correspondence with polyphony in Giga.

BTW I steer clear of dskbench after it destroyed one of my drives http://www.northernsounds.com/ubb/NonCGI/images/icons/smile.gif

[carlgt1]

>did 38MB/sec. (real life, including a lots
>of seeks back and forth all the time),
>you\'d be able to optain roughly 860 voices
>of polyphony.... http://www.northernsounds.com/ubb/NonCGI/images/icons/smile.gif

860 voices, I\'ll take that! dskbench does give me about 1/3rd of my Hdtach 38MB/s (10mb/s at my highest blocksize). Actually dskbench does a read/write at 33MB/s but in the \"track size\" part it quick drops to 10MB/s and ultimately 1.7MB/s at 4k blocks.

Is there really a rule of thumb for x mb/sec = y voices in GS?

One thing that annoys me is I bought GS96 thinking 96 voices was plenty. Then later I found out that a monophonic stereo counted as 2 voices, and using the sustain pedal for a second bumps up to 100 voices, etc! I wonder if Nemesys will allow me to do that cheap upgrade they had recently!

[Simon Ravn]

<BLOCKQUOTE><font size=\"1\" face=\"Verdana, Arial\">quote:</font><HR>Originally posted by carlgt1:


860 voices, I\'ll take that! dskbench does give me about 1/3rd of my Hdtach 38MB/s (10mb/s at my highest blocksize). Actually dskbench does a read/write at 33MB/s but in the \"track size\" part it quick drops to 10MB/s and ultimately 1.7MB/s at 4k blocks.

Is there really a rule of thumb for x mb/sec = y voices in GS?

<HR></BLOCKQUOTE>


There is no \'rule\'. It\'s maths. 38MB=roughly 38.000.000 bytes. The sampled data uses 44.100 bytes pr. second, so 38.000.000/44.100 is the number of voices you\'d be able to get if everything was ideal - which of course, it never is.

[Simon Ravn]

<BLOCKQUOTE><font size=\"1\" face=\"Verdana, Arial\">quote:</font><HR>Originally posted by Simon Ravn:


Actually theres an error in my calculation:

There is no \'rule\'. It\'s maths. 38MB=roughly 38.000.000 bytes. The sampled data uses 44.100 bytes pr. second, so 38.000.000/44.100 is the number of voices you\'d be able to get if everything was ideal - which of course, it never is.
But I forgot that we\'re talking 16-bit data. So we have to divide that by 2, so actually it\'s only 430 voices.


<HR></BLOCKQUOTE>

[dnortana]

carlgt 1 wrote:
\"I take it that you were trying to avoid the onboard Promise controller for some reason? I\'m using them and my audio stuff flies, 38MB/sec in HDTach! I\'ve heard rumors of data corruption (with VIA but not Promise or the other way around?) but everything has been fine. Plus, once my GIGS are loaded it\'s all \"read only\" anyway.\"

Carl,
I\'ve read and read my mobo manual, and it never was clear to me that the Promise controller could be used in a non-RAID configuration. I do not want to use RAID, since I believe it to be inappropriate for an application like Gigastudio. But it would be nice to have access to the extra two IDE channels, if that\'s possible via Promise.
Maybe we could take this off-list?
I\'m at dnortana@bellsouth.net
Thanks,
Trond.

[carlgt1]

>never was clear to me that the Promise >controller could be used in a non-RAID >configuration. I do not want to use RAID,

well there\'s usually a BIOS option to run the RAID as a real RAID0 or whatever or just a nice ATA100 port. That helps in keeping drives from sharing a single port (e.g. I have my boot drive on the \"RAID\" running at ATA100, the Giga drive on the regular ATA100, and the CDROM on another separate IDE port). Just look in the BIOS for this \"RAID\" or \"ATA\" option for the onboard RAID controller.

[seclusion]

I run a P3 1 gig 512 puter dedicated to gigastudio.. It has a 4 hub raid ide controller in it... The puter see\'s 2 seagate 45 gig drives as 1 90 gig drive..
So the gig files are distributed between two fast drives..
Absolutly no click and pops.. Going strong for over 3 months..
I had this system made for me for audio..
They say that 1 of the biggest problems people do when building a heavy system is install a ****ty C:/ drive..
So mine has 3 ata100 rates seagate drives..
My only complaint is I should\'ve got 160 instead of 96..
Oh well soon..
Brian