HARDWARE POLYPHONY POLL

[Danny Lux]

I am interested in upgrading my GS systems, and would like to hear what systems are acheiving the 160 voices or even close to that number.
What motherboard, processor, hard drive, and RAM configurations are you using?
Responses are appreciated.
Thanks.

[donnie]

Danny,

Believe it or not I am getting the full 160 voices on my Gateway PIII550, 256 ram, (2) 13.5 ultra dma 66 7200 rpm hard drives. The only thing different that I have done is follow the suggestions on the Win 95/98 optimizing list. I posted the link to this awhile back under software (I think).

Donnie
DS Soundware

[Danny Lux]

Donnie,
Are you getting the 160 Voices with the Giga piano??

[donnie]

Danny,

Actually I have not used the giga piano in any of my compositions. The main instruments I have been using are: Quantam Leap Brass, Advanced Orchestra, X-Sample Brass, and Ultimate Orch. Perc. (of course). I\'ll try and do an experiment with the piano when I get some time and I\'ll let you know.

Donnie

[alex kharlamov]

Danny and everyone, just in case I wanted to tell you about a Maxtor-related topic in the \'Hardware\' root-directory of this forum. There is an issue with installing Maxtor drives with other manufacturers\' drives present on the system that you guys might want to be aware of. Might save a bunch of time and maybe a couple of grey hairs :-) (for one night I thought I lost 50GB of gig and vst files.)

[alex kharlamov]

Danny, I\'ve just put a system together for Rob Kral, and also assembled one for Chris Beck last year and for myself this year. Every time it\'s a new configuration because of new (faster and cheaper) parts coming out every month. It\'s no big brainer though. I\'m researching a new setup for Rob\'s friend and can let you know what\'s best at this point. I can say that in my particular case, it\'s important that I work with a guy who only knows how to put a quality machine together out of parts I ask him to include. He does no sound card/software installation -- only raw PC stuff. The best place to get lowest prices on any PC part is Pricewatch.com. One major thing many people try to emphasize here is that you need ultra-fast harddrives, maybe even SCSI. In my personal experience, using two-three 8.5 ms ATA drives is good enough as far as the drives go. I think SCSI is a little overkill for Gigastudio for now, especially if you already have another PC/Gigastudio system to run along. I\'d suggest that if the funds allow, you only use the latest of IBM harddrives (right now, it\'s the Deskstar 75GXP), and only use and add harddrives of the same brand within the same computer. Harddrive BIOS is the major reason for that. Maxtor harddrives work fine on my system (along with IBM HD, but what a nightmare I went through in order to get all of them to work together). I\'ve also noticed that the harddisk compatibility a lot of times depends on the motherboard. Don\'t same money on the motherboard, and of course you know not to use anything except for the fastest Intel Pentium processor. On the memory side, I\'d wait until a 512MB RAM chip comes out, and install 2 of those, since most of motherboards now have not more than 3 DIMM slots. There\'s a bunch of other things to consider. If you still need help with getting your new Gigastudio machine together, please let me know via private email, alex@kharlamov.com, I\'ll be glad to help you with it.

[ursatz]

Hi Alex,

\"and of course you know not to use anything except for the fastest Intel Pentium processor. \" Why is this? I\'m going to be upgrading my processor and mobo soon, and it sure looks like an Athlon system would be a lot less expensive. I thought people were running GSt pretty well on Athlons (with maybe some problems with certain soundcards) - is that not true?

[killerbobjr]

>>>>
I am interested in upgrading my GS systems, and would like to hear what systems are acheiving the 160 voices or even close to that number. What motherboard, processor, hard drive, and RAM configurations are you using?
<<<<

Some preliminary results:

Dedicated PC
500Mhz K6-2
FIC VA-503+ Motherboard
128 MB RAM
2x 4.3G SCSI IBM 8.5ms 5400RPM Ultrastar ES
MIDIMan 2044 Audio Interface

Using Kevin Moreland\'s Kawai Piano Gig, I got up to 120 or so voices before it reached 90%+ utilization and the display would stop updating. At that point I would start getting crackle. I backed it down to 110 voices and it\'s very happy. Just for kicks, I tried GSt on my main machine (866MHz Coppermine) loading the sample over the network (100Mb ethernet). Utilization never went over 40%, voices maxed out at 138 with crackle. I guess I\'m at the limits of those hard drives. I\'ll have to try Gigapiano on the main machine to see if I can get 160.

I think my efforts show that you can get excellent results on a well set up, low end machine.

[ursatz]

killerbobjr,

Yee-ow! I can\'t get much over 60 voices in GSt with my K6-II 500 mhz - I was thinking the K6 was to blame, but I guess not! I\'ve done all the tweaks in your post on dedicating a PC to GS; my hd is WD EIDE, 9ms, 7200rpm. Maybe the hd is the problem?

[Viktor]

I agree with Ursatz about the Athlon/Intel saga. There is new CPU on the market called Duron. It is low cost Athlon and it has great FPU performance, is very easy to overclock and is very cheap, too. www.tomshardware.com (\"http://www.tomshardware.com\")

The benchmarks showed that Duron FPU is 30-40% faster then PIII/Celeron at the same speed.

I would like to hear some athlon users here, so we can find out how it is keeping up with Intel CPUs. I am sure it will be better.
Other audio software (such as Reaktor and plugins) also works better with Athlon cos of its strong FP unit.

About HDD, I must agree with Alex K. and say that IBM are the best choice outhere, namely IBM 75GXP. It has better access time then Maxtor, and it is pretty quiet, too. www.storagereview.com (\"http://www.storagereview.com\")

take care

[killerbobjr]

Well, here\'s an update. After fooling around with GSt all night long and doing some extra tweaking, my results are thus:

K6-2 500MHz:

140 voices max for Kawai 600Mb grand without crackle @ 95%+ utilization. This is an unlooped, 4-layer, sample-for-each-key Gig file.

90 voices max for Bellsynth 9Kb Gig without crackle @ 90%+ utilization. This Gig is a single WAV mapped across the entire keyboard with a loop. That extra processing needed for interpolation and looping costs some CPU power.

Coppermine 650MHz @ 866MHz, non-tweaked machine, several background processes running, loading Gig file over 100Mb ethernet:

160 voices on Kawai @ ~40-50% utilization.

160 voices on Bellsynth @ ~60-70% utilization.

This time, I lowered master attenuation to -6dB which prevented not only crackle but also a whole bunch of overload distortion and made my Coppermine happy.

[Danny Lux]

These posts are very helpful. I am not familiar with the Kawai piano you are using, but as you have noticed different patches achieve different polyphony. For instance, I can get more polyphony from the Truan Steinway B which uses Nemesys\'s lossless compression, than the Trachtman Steinway C which does not use the compression. Not sure if this compression/results are due to the load that it puts on the hard drive? Anyways, the more posts/results people leave can help direct what systems can actually achieve the 160 voices, and the posts can also help people know if there own systems are performing up to snuff compared to similar systems that are posted here.
Thanks

[ursatz]

Alex,

Thanks for the response. Right now I\'m pretty confused about what would be the best way to upgrade my system, but I have a GSt setup that I can use at the moment, so for now I guess I\'ll hang tight and see if prices come down. I really appreciate all the info available on this board.

[alex kharlamov]

\"and of course you know not to use anything except for the fastest Intel Pentium processor. \" Why is this? I\'m going to be upgrading my processor and mobo soon, and it sure looks like an Athlon system would be a lot less expensive.

I am not familiar with Athlon processors yet. However, in my personal experience, when I tried a Celeron or AMD, they\'ve always performed worse (in a music system) than a real Pentium. The main difference in processor designs so far has been the floating point, which a lot of music software relies on. When most software is being developed, it\'s tested on the latest Pentium Intel processors. That\'s why they recommend the Pentiums. So it\'s not against the law to use other processors, and it\'s really worth it if you have time to experiment. If you don\'t have time but do have the money though, I think the latest Intel Pentium is always the most sure way to go. It may change any minute, though, and that may be the case with Anthons -- I hope, since they are cheaper. Cheaper and the same specs on paper don\'t always mean better in reality, though. A lot of it is marketing.

[Viktor]

Alex,

FPU performance of Athlons are not marketing tricks. Check out the benchmarks in 3D studio, Reaktor, DX and VST plugins. I am not talking about synthehtic benchmarks here. Athlon/Duron have more advanced FPU the PIII/Celeron. It is without the questions.

One more thing,
Athlon and Duron have the excatly same FP unit. The same thing is with PIII and Celeron. Therefore, Reaktor and some plugins I\'ve tested show the same CPU usage on both PIII and Celeron.

I am not qualified to talk about Intel/AMD performance in Gigastudio, but I strongly believe that GST loves Athlons/Durons.

But, it is true that some audio hardware manufacturers have some compatibility issues with AMD chipsets (not CPUs), but I hope new VIA KT133 chipset will be better.

[WTrachtman]

==============================
For instance, I can get more polyphony from the Truan Steinway B which uses Nemesys\'s lossless compression, than the Trachtman Steinway C which does not use the compression. Not sure if this compression/results are due to the load that it puts on the hard drive?
==============================

I finally got the Gigastudio, and I have it installed on my older, Pentium-II, 333 MHz machine. I\'m also about to purchase a new system for use with the Gigastudio, so I am very interested in hearing comparative results from diffeent hardware as well. However, I want to advise folks that the comparisons may not be correct unless a common set of test data is used.

Nemesys has sent me their lossless compression software, and I have remastered my Steinway-C piano using the compression on the waveforms.
My current system, with 128MB of RAM is choking up at about 64 voices, and I can\'t currently set the polyphony higher if I want to load in my piano without going into \"Swap-File-Hell\".
What is interesting is that on this slower machine I do not notice a significant difference in performance using the \"compressed\" version versus the non-compressed version.
As a further test, I loaded the Gigapiano and checked the method which is used by the Gigastudio program to count the polyphony.

The Nemesys Gigapiano does not have a separate waveform sample for each note. For example, the recorded waveform from the piano\'s Middle-C is stretched to cover several notes, and is used to create the sound for not only the C4 (midi note number 60), but the C#4 (61) and D4(62) notes as well. Playing these three notes will give a reported polyphony use of 6 (three notes times 2 for stereo) even though only a single stereo wavefile needs to be read from the harddisk.

So, comparisons of achievable polyphony, as reported by the Gigastudio, will be heavily skewed depending on which specific gigainstrument is being used for the test, and which specific notes are being played.
If the instrument being used to evaluate the system\'s polyphony has been programmed with multiple notes covered by an underlying sample waveform, then the efficiencies of reading fewer samples from the disk will mask the information regarding the hardware performance which we are actually trying to obtain and compare.

Unless all the polyphony checks are done using the same instrument, playing the same MIDI file, we will be \"comparing apples and oranges\".

Regards,

Warren Trachtman
http://www.wstco.com (\"http://www.wstco.com\")

[Danny Lux]

Warren wrote:
<Playing these three notes will give a reported polyphony use of 6 (three notes times 2 for stereo) even though only a single stereo wavefile needs to be read from the harddisk.>

I am a little confused...are you actually playing 3 notes simultaneously or just 1 note and GS is registering as if you were playing 3. If you are playing 3 notes, polyphony is just polyphony at that point. Even though it is just 1 stereo wave file, it still has to stream it 3 indepedant times, i.e. every voices has its own stream per voice. It seems to me, if we are all using the Gigapiano as a bench mark for arguement sake, it does not matter if the wave file is the same for multiple notes because it is still streaming the same amount of data.

[WTrachtman]

I am not by any means an expert on the details of low-level disk I/O routines. I was making the assumption that streaming the same disk file three times, for three notes, is more efficient than streaming three different files.
I have a gut feeling that this should be the case, especially because of the caching used internal to the modern hard disk drives (Nemesys is recommending use of hard disks with 512k internal cache).
I believe that reading the same disk file three times should be faster for the disk I/O routines than reading three separate files, and may mask the hardware speed comparisons if we are not careful about things. But, this is just an assumption, and I may be completely off-base about it.
I expect/hope that there are some folks here who really *do* know about this sort of thing and can give an indication as to whether there is a significant difference or not, perhaps even quantify it.

Regards,

Warren Trachtman

[aspenleaf]

Danny,
I just bought a Pentium III 733 coppermine on an Asus P3V4x MB with 128MB PC133 RAM and a Maxtor Diamondmax 7200 RPM UDMA66 Drive for the gigs, and I\'m getting 160 voices with the Truan Steinway B. I\'m using a SBLive Value soundcard and can also add reverb and EQ in Gigastudio in realtime.

[Danny Lux]

I have tried setting my vcache to 98304 since I have 384 megs ram. The polyphony does not change. Using the Truan Steinway B, I get about 110 voices of polyphony on 2 different machines, 1 is a PIII 500, and the other is a PII 400. I would think the 500 would give at least a few more voices. I have read how others are increasing their vcache and seeing significant polyphony increases. I would appreciate any suggestions anyone has why I, and probably some others, are not seeing changes when vcache is increased.