Oh well, I just did some more testing with eXT and... It started hiccups with only 4 instances of Zebra2.
Sometimes I couldn't hear the notes of the newest instance, or the whole sequencer kinda slowed down for a second and I got nasty digital noise. The CPU usage wasn't over the top either, so this is something else. It can even totally slow down the system after playing a loop for a while.
The ASIO buffer size of 256 should be pretty generous?
EDIT:
I changed the ASIO buffer size in my driver to 512 and I can play with 6 instances of Zebra2 without hiccups (didn't try more). That makes the latency 11ms which is starting to be a bit too high for my liking.
EDIT2:
It can hiccup now and then even with 512 buffer when just listening to a loop playing those 6 instances. It seems to behave much better than with smaller buffers though.
Gotta double check which version of Zebra2 I am using but it should be the first GUI fixed SSE2 enabled version (should we with SSE2 machines use something else at this point?).
EDIT3:
I forgot that I had switched off Hyperthreading in the BIOS at some point. I switched it back on, and now 6 instances with sample buffer of 256 seems quite a bit better behaved... It's still not glitchless though.
I dunno, is it just that we shouldn't expect very low latencies with several complex synths like this, or is there some problem with the code? I am not 100% sure how the pre-release version behaved since I had it for so little time that I didn't do much multi-instance stuff. Is there any point in testing?
Zebra2 Latest Version: 2.1 - Jan 19 2007
-
- KVRist
- 41 posts since 9 Jan, 2006 from Slovakia
I have installed it and the main bugs I found yet are: Bug 1 when changing presets some of them make no sound at all but it apears to be caused by a previous preset some of the presets when changed to another are causing this, when I change two different presets then I can get some sound out of previously non audible ones. Bug 2 after an hour of not touching the PC it crashed- no sound at all.
PC AMD 3800+ X2 64
1 GB RAM
MB Asus A8N-SLI Deluxe - nForce4 SLI
M audio Delta 1010
Host: CUBASE SX 3
PC AMD 3800+ X2 64
1 GB RAM
MB Asus A8N-SLI Deluxe - nForce4 SLI
M audio Delta 1010
Host: CUBASE SX 3
-
- KVRAF
- 10815 posts since 26 Nov, 2004 from UK
-
- u-he
- Topic Starter
- 30207 posts since 8 Aug, 2002 from Berlin
Uh oh...
is that all new? Did that happen with previous versions?
My stuff should not experience any difference wheather your audio buffers are 64 samples or more.
What I suspect is that there are short cpu spikes every now and then when several oscillators need to recalculate wavetables at once. This effect may be stronger on Win than on Mac as I use different libraries for FFT and stuff there. - I already worked out some concepts to distribute the heavy osc parts more evenly across rendering, so I may try this within the next couple of days.
MoodyBat, when the PC crashed after an hour, was that with the Z2 GUI open?
Later,
Urs
is that all new? Did that happen with previous versions?
My stuff should not experience any difference wheather your audio buffers are 64 samples or more.
What I suspect is that there are short cpu spikes every now and then when several oscillators need to recalculate wavetables at once. This effect may be stronger on Win than on Mac as I use different libraries for FFT and stuff there. - I already worked out some concepts to distribute the heavy osc parts more evenly across rendering, so I may try this within the next couple of days.
MoodyBat, when the PC crashed after an hour, was that with the Z2 GUI open?
Later,
Urs-
- KVRer
- 9 posts since 10 Sep, 2005 from Stockholm, Sweden
Grid and preset name doesn't always update when changing patch after using the preset browser.
Load preset "HH Lost out there 1" by clicking on the name in the preset browser. Switch back to Synthesis view. Click next patch arrow. "HH Lost out there 2". Click next patch arrow again. Grid layout and patch name is not updated but everything else is updated to "HH Lost out there 3".
Zebra VSTWin 260706 / EnergyXT, Live, MiniHost, VST-RTAS wrapper
Load preset "HH Lost out there 1" by clicking on the name in the preset browser. Switch back to Synthesis view. Click next patch arrow. "HH Lost out there 2". Click next patch arrow again. Grid layout and patch name is not updated but everything else is updated to "HH Lost out there 3".
Zebra VSTWin 260706 / EnergyXT, Live, MiniHost, VST-RTAS wrapper
-
- KVRian
- 864 posts since 4 Apr, 2001 from Finland
Yeah, same here. Interesting... If I go one step further to the next patch and then back to "HH Lost out there 3" it shows up OK.
This quirkiness seems to _only_ happen in the Synthesis view. Not in Perform or Presets.
This quirkiness seems to _only_ happen in the Synthesis view. Not in Perform or Presets.
-
- KVRian
- 864 posts since 4 Apr, 2001 from Finland
I noticed that in the Windows Control Panel at:
Control Panel->
System->
Advanced->
Performance Settings->
Advanced->
Processor scheduling
I had processor scheduling on "Adjust for best performance of: Background services".
I switched it to "Adjust for best performance of: Programs".
Then I managed to make a loop with TEN instances of Zebra2 in energyXT which seemed to work without hitches until I overloaded the CPU. I was changing patches here and there and adding instances while it was running. And this was with 256 buffer.
The "Programs" setting should be the default, but I recall that I changed it to the "Background services" because I read in some guide that it's a good setting for audio software....
Djsubject, can you check your setting? I'll continue to test this theory a bit more...
EDIT: Yeah, that "Background services" setting is recommended for example here:
http://www.homestudioschool.com/homerec ... uction.htm
Control Panel->
System->
Advanced->
Performance Settings->
Advanced->
Processor scheduling
I had processor scheduling on "Adjust for best performance of: Background services".
I switched it to "Adjust for best performance of: Programs".
Then I managed to make a loop with TEN instances of Zebra2 in energyXT which seemed to work without hitches until I overloaded the CPU. I was changing patches here and there and adding instances while it was running. And this was with 256 buffer.
The "Programs" setting should be the default, but I recall that I changed it to the "Background services" because I read in some guide that it's a good setting for audio software....
Djsubject, can you check your setting? I'll continue to test this theory a bit more...
EDIT: Yeah, that "Background services" setting is recommended for example here:
http://www.homestudioschool.com/homerec ... uction.htm
-
- KVRAF
- 10815 posts since 26 Nov, 2004 from UK
Background services enabled here aswell
just checking with it set to programs & all SEEMS fine
good work spritex
Subz
just checking with it set to programs & all SEEMS fine
good work spritex
Subz
-
- KVRian
- 864 posts since 4 Apr, 2001 from Finland
Still, I am not 100% sure of it. I just made a 3 instance thing with Background ENABLED, and it started glitching. I saved the project, opened it again, and it was OK.
But maybe it's less likely to happen if it's set to Programs? Maybe we should just like torture it now with the Programs setting on, and see if it can take a reasonable amount of abuse?
But maybe it's less likely to happen if it's set to Programs? Maybe we should just like torture it now with the Programs setting on, and see if it can take a reasonable amount of abuse?
-
- u-he
- Topic Starter
- 30207 posts since 8 Aug, 2002 from Berlin
Wow... now that is something. I guess it depends on the host software and drivers, i.e. how the audio threads are set up.
Anyhow... it would be cool if you could try that with patches that do not use oscillators (i.e. only FMOs) vs. patches that make huge use of oscs... if the oscs are the cause, I probably have a quick fix for that.
Cheers,
Urs
Anyhow... it would be cool if you could try that with patches that do not use oscillators (i.e. only FMOs) vs. patches that make huge use of oscs... if the oscs are the cause, I probably have a quick fix for that.
Cheers,
Urs-
- u-he
- Topic Starter
- 30207 posts since 8 Aug, 2002 from Berlin
Okay. I thought I had that one fixed long time ago. It may have come back when I merged the Mac sources with the Win stuff... will investigate, but it should be possible to fix quickly as well...R206 wrote:Grid and preset name doesn't always update when changing patch after using the preset browser.
Later,
Urs-
- KVRAF
- 10815 posts since 26 Nov, 2004 from UK
Urs wrote:Wow... now that is something. I guess it depends on the host software and drivers, i.e. how the audio threads are set up.
Anyhow... it would be cool if you could try that with patches that do not use oscillators (i.e. only FMOs) vs. patches that make huge use of oscs... if the oscs are the cause, I probably have a quick fix for that.
Cheers,
yep it dose seem to be patches with more than one osc that trigger this also closing & reopening eXT solves the problem
intill i add another zebra2
hth
Subz
-
- KVRian
- 864 posts since 4 Apr, 2001 from Finland
There can be a huge variation on the energyXT CPU meter when playing the 10-instance loop. It can go from like 40% to 100%. BUT, with the "Programs" processor scheduling on, I haven't been yet able to get into the "digital noise" state where it hangs - glitches of course when the CPU meter is at the top, though.
I'll try some intensive non-OSC tests as you asked.
I'll try some intensive non-OSC tests as you asked.
-
- KVRian
- 864 posts since 4 Apr, 2001 from Finland
Hmm. I was just starting to test again. I made a blank loop and started adding instances looking for non-OSC patches. On the fourth instance while browsing for patches while the sequence played, I got into the state where only some of the patches played in the newest instance and I got just digital crackle or nothing at all in others. Then the sequecer got a slowdown. CPU usage was not very much before the fourth instance.
There is definitely _something_ wrong here. Of course some of the patches I browsed for had OSCs in them. Perhaps I should test so that I don't even browse for patches that have OSCs?
There is definitely _something_ wrong here. Of course some of the patches I browsed for had OSCs in them. Perhaps I should test so that I don't even browse for patches that have OSCs?
-
- KVRian
- 743 posts since 14 Apr, 2004
Urs, what about over the thing on restarting sound on sequencer loop mode?
I will try to be more clear:
In FLStudio when I make a "midi" piano roll loop using Zebra2 the sound restart when the loop return to the first note....this is more audible when you use a delay...
This thing is very annoying when I use a sequencer in loop mode to make some "audio sketches"...on FLStudio I have this problem only with Zebra.
I've asked this thing some time ago but I never saw this on the buglist..
Do you think that this thing will be fixed on the final release??
I will try to be more clear:
In FLStudio when I make a "midi" piano roll loop using Zebra2 the sound restart when the loop return to the first note....this is more audible when you use a delay...
This thing is very annoying when I use a sequencer in loop mode to make some "audio sketches"...on FLStudio I have this problem only with Zebra.
I've asked this thing some time ago but I never saw this on the buglist..
Do you think that this thing will be fixed on the final release??