Giga3 Polyphany Vs 2.5

[dreamuse]

Hi All,

I've just installed Gigastudio Orchestra 3.04 on my system. I brought up Gigapiano I, and ran no plugins. Though hardly scientific, I remember holding the damper pedal down and hitting lots of notes in 2.54, and while I'd peg the voices at the max (160), the CPU barely broke a sweat.

Now, in 3.04, the CPU's maxing out fairly quickly, even before I get past, say 110-120 voices.

I'm running XP, service pack I on a Pentium 4, 3.02 Ghz.

So, my big question is: is 3 more CPU-intensive in terms of straight voice playback? If not, are there things I should do to furthur optimize performance?

I'm running a Layla 24 box, and have installed their driver 6.11, w/ Gsif 2 and WDM.

Thanks in advance for any enlightenment!

Len

[rjdekker]

This won't be of much help to you...but I have observed the same thing.

Actually, in practice my harddrive speed does not appear to be the factor limiting the max. polyphony in my system, but the CPU is (P4 3.0 Ghz with Raptor RAID0). I always thought it was vice versa. Upgrading from a single drive to the RAID config only slightly increased polyphony in my case (?).

I'm no expert, but I know that just reading data from the harddrive always puts some strain on the CPU, depending on some things like the HDD interface used. On top of this GS3 has to do the mixer processing (even when no fxs are applied).

I've also tried (substantially) increasing the buffer size of my RME HDSP9632, but this has hardly an effect on the CPU use and max. polyphony. This is also a bit counter intuitive for me.

Maybe it is motherboard related? I'm using a Asus P4C800 deluxe.

Rob

[GTBannah]

Is this the same system you used for GS 2.5?

GS 3 is probably a "fatter" programme, so, with more details to check for, the CPU is under a little more stress, leaving less "horse power" available to play voices. ;)

[dreamuse]

Thanks for your replies, but the thing is this:
It's the same hardware as I was using before, and in 2.54 the cpu meter would barely register when I played a new key, thus adding 1-2 new voices. Now, the meter's jumping a lot more, and overloading easier.

Len

[JonFairhurst]

I believe that GS3 uses a new lossless compression scheme that uses a bit more CPU than GS2.5. That would account from the higher CPU usage for sample playing without effects. I don't know the inner workings, so I'm not sure of the details.

You might compare playback of a v2.5 lib, compared to a lib that has been converted to v3.

-JF

[belbin]

It seems to me that none of you have accounted for the 2 giga pulse instances that are encoded. Or was it mentioned somewhere that you're using GPiano LE? Fully modelled GigaPiano II would certainly account for greater CPU load.

Belbin

[belbin]

I should clarify: I see that you say you didn't use any plugins, but this could mean, "didn't ADD any". Just making sure that you're aware that GP II has 2 instances of Gigapulse that you need to disable/delete if you really want it plugin-free.

And no, I don't think your an idiot. I'm just a know it all. Sorry. Just trying to help, really. Hope I'm right....

Belbin

[dreamuse]

I was running this test with the Gigapiano that came w/ 2.54, so there are no plugins instantiated.

Even though I've set the max polyphony to 900, I'm pegging the processor (actually, it's claiming I'm around 150-250%!), and getting max polyphony of 238.

I guess this is better than 160 voices. :)


Thanks,

Len

[belbin]

All I can say now is that I've read (and disregarded, being a newbie) a lot of mention that certain 2.5 tasks won't run/won't run as well in v.3. Try giga piano 2. My machine is a p4 2.9 Ghz, and even before I got a second hd, I could use gigapiano II with no strain at all.

Belbin

[JonFairhurst]

It would be interesting to make a copy of the original GigaPiano and convert it to the version 3 format, and then compare its perfromance.

-JF

[skeller]

GS2.5's 160 voice limit may have been masking a midi local echo issue. Something to look out for if you play to a sequencer which echos to GS3 with the keyboard also outputting omni. If that's the problem, your voice count will cut in half when you fix it and the samples will sound better when you play them.

[dreamuse]

Nope...I checked for loopback, and everything seems to be correct.

I think it'd be useful for all of us to compare how many voices we're seeing.
If you reply, please include processor type and speed. This way, we can all compare apples, etc...

Len

[dpc]

Do I understand that GS3 opens with 2 default instances of Gigapulse? How do you disable it then?

Dave Connor

[dreamuse]

I'm not seeing any instances of Gigapulse loaded on startup in any of the mixer windows, so I don't think so...

Len

[belbin]

Full giga piano has 2 instances. If you're opening the workstation by double-clicking on GP II, then you will get those. Otherwise, they shouldn't be there.

Belbin

[dreamuse]

In my case, I'm using the old GigaPiano (NemeSys_1GB), so no GigaPulse instances are happening.

Len

[digitalson]

i just built a p4 306 with 1 gb mem...single hd...and was getting a solid 500 voices...tried 600 but it crashed hard...

[belbin]

500 voice out of a single HD!! Wow.
I haven't pushed that hard yet, but I've done about 300. That was before, when I had 1 HD and 1 Gig of RAM. Current Setup is:
P4 2.93 Ghz
2Gigs Mem
2 HD's (system and Samples)

[digitalson]

thats on a clean install of xp and fresh drives

[dreamuse]

i just built a p4 306 with 1 gb mem...single hd...and was getting a solid 500 voices...tried 600 but it crashed hard...


Digitalson,

When you were running 500 voices, what was the cpu load?? Even with 200 voices, I'm seeing my meter peaking, and sometimes staying in the red.

Len

[digitalson]

cpu was at 60%...when it hit 600 it went thru the roof

[howardv]

Len:

You eating up 238 voices playing live with a keyboard or is that replaying a midi file? If it's a file, I'd be happy to run it on my system for comparison. In fact, my email machine is an old Win98 clunker with 2.5 something on it which I could use for a dual comparison.

Also, I recall trying to record once upon a time with an old unweighted Korg C-800. When I looked at the midi file, I found that almost every note played was doubled and tripled due to key bounce. Just a thought.

Howard
Ragtime (http://www.rtpress.com) Press (http://www.rtpress.com)

[dreamuse]

Len:

You eating up 238 voices playing live with a keyboard or is that replaying a midi file? If it's a file, I'd be happy to run it on my system for comparison. In fact, my email machine is an old Win98 clunker with 2.5 something on it which I could use for a dual comparison.

Also, I recall trying to record once upon a time with an old unweighted Korg C-800. When I looked at the midi file, I found that almost every note played was doubled and tripled due to key bounce. Just a thought.

Howard

Hi Howard,

This was w/ a midi file in which I duplicated piano tracks for a total of 8, complete with sustaining notes.

Thanks for the offer to run this file. I'd obviously be happy to ship it your way!

Len

[dreamuse]

I think I got to the bottom of my particular setup's problem:

I've got a Layla24 interface hooked up to my laptop via a PCMCIA card interface. I also have a Midiport USB midi adapter. Using the Midisport, I can get a ton more voices...like over double what I could through Layla Midi before I peg the CPU meter.

I'm thinking that either the PCMCIA card doesn't get the computer's attention it needs and/or is processor heavy to deal with, or the Layla driver's really inefficient.

Either way, it's nice to see/hear the laptop playing 400+ voices while the CPU meter's maxing at around 70%.

Now, if I could just get GigapianoII to stop popping while playing more than 4 notes (regardless of GSIF buffer settings), I'd be real happy.

Thanks to all for your helpful ideas.

Len

[howardv]

This was w/ a midi file in which I duplicated piano tracks for a total of 8, complete with sustaining notes.

Thanks for the offer to run this file. I'd obviously be happy to ship it your way!
Was wondering how you were encountering that kind of load. I didn't pick up on it being a benchmark thing. So are you playing the midi file from a sequencer running on a desktop and driving GS via midi cables to the laptop? Or the other way around? If you think the midi interfaces are a bottle neck, you might try dispensing with them by running GS3 and the sequencer on the same machine. I'll pm you with my e-mail address for that midi file.

Howard
Ragtime (http://www.rtpress.com) Press (http://www.rtpress.com)