Who can play complex software synths at 64 samples?

[thesheep]

Is anyone here able to play some of the more complex software synths, for example complex atmospheric pads in Wavetable or Analog, at 64 samples or less? (assuming sensible sample rate like 44 or 48kHz). If so, what kind of hardware do you have?

I have a brand new MacBook Pro Quad Core 16GB RAM, etc and new Focusrite Thunderbolt Clarett audio interface, which gives me super low latency for audio recording and monitoring (I can go down to 32 samples and monitor live with reverb, compressor, etc and it's stable, which is awesome).

But even if I shut down all other software on my machine, I can't get below 256 samples without pops and crackles on a single MIDI track running something like the 'Sentient Shadow' preset for Wavetable (which sounds lovely, apart from the crackles).

So I'm curious, is that about as fast as most people get? I would like to get down to 64 samples but wondering what kind of hardware you might need to support that—some kind of Mac Pro maybe?

Seems like MIDI is the new bottleneck these days, it always used to be audio!

[login]

I can use a 64 buffer size using Wavetable, Diva, Serum or Bazille, in many cases more than one instance if they are not all running several notes each one, let's say 8 voices in total.

I am using an i7 4770k, 16gb RAM, SSD,(custom desktop PC) on windows, with an RME Fireface UC interface.

[kitekrazy]

Is anyone here able to play some of the more complex software synths, for example complex atmospheric pads in Wavetable or Analog, at 64 samples or less? (assuming sensible sample rate like 44 or 48kHz). If so, what kind of hardware do you have?

I have a brand new MacBook Pro Quad Core 16GB RAM, etc and new Focusrite Thunderbolt Clarett audio interface, which gives me super low latency for audio recording and monitoring (I can go down to 32 samples and monitor live with reverb, compressor, etc and it's stable, which is awesome).

But even if I shut down all other software on my machine, I can't get below 256 samples without pops and crackles on a single MIDI track running something like the 'Sentient Shadow' preset for Wavetable (which sounds lovely, apart from the crackles).

So I'm curious, is that about as fast as most people get? I would like to get down to 64 samples but wondering what kind of hardware you might need to support that—some kind of Mac Pro maybe?

Seems like MIDI is the new bottleneck these days, it always used to be audio!

A Mac desktop would do it. Laptops are bottlenecks. Also some plugins are not resource friendly. Going down to 64 samples only serves as bragging rights and liver performances.

[jestermgee]

Seems like MIDI is the new bottleneck these days, it always used to be audio!

Nope, the reason is the high number of calculations required in the audio still, as it's always been.

What works for one may not work exactly for another for a myriad of reasons.

Best thing to do is experiment with what settings work and tweak them until you find something that gives stable audio. Forget about the "lowest buffer size possible" and just set things up so you can play/record well enough without latency. 128/256 buffer is still pretty usable. I personally run a buffer of 128 @ 48Khz. My audio interface (Roland OctaCapture) has a "low latency" mode but I do find that causes the occasional audio drop every 5 minutes or so, without that option on it's fine and I notice nothing different when playing a synth.

[miyaru]

I rum my Focusrite Saffire @ 512 samples..... can go to 250 samples but keep it on the safe side.

[jestermgee]

I rum my Focusrite Saffire @ 512 samples..... can go to 250 samples but keep it on the safe side.

And do you ever have any "noticable" latency when you play/record any MIDI? My guess would be no, which is something people need to understand better that having the absolute lowest buffer isn't that necessary. Maybe a trained concert pianist would tell if an extra 12mS was added to the delay but for most of us, a 30mS audio delay is almost a non-issue.

The lower your buffer the harder the machine has to work to deliver the audio in such a short space and the easier it can be upset by the slightest things.

[thesheep]

Thanks for the responses, so it looks like many people can't get down to 64 samples either.

And do you ever have any "noticable" latency when you play/record any MIDI? My guess would be no, which is something people need to understand better that having the absolute lowest buffer isn't that necessary. Maybe a trained concert pianist would tell if an extra 12mS was added to the delay but for most of us, a 30mS audio delay is almost a non-issue.

I think it depends a lot on how you play/record and the type of part you are playing. For me, recording vocals with a 30ms delay, and monitoring live, is definitely an issue, and I expect most people would agree with that. Similar for running guitar amp modelling while you play a guitar. My approach to recording is often based around performance/improvisation and fast keyboard parts will suffer too. I guess I could try to work around it by bouncing down software synth parts to an audio track, then changing the sample size mid-project, but that would take away some of the 'live' aspect of using Live...

I guess I'm a little bit surprised that software synths seem to take so much processing compared to recording audio and applying effects, which seem to be less of an issue.

[miyaru]

I don't notice any extra latency, have recorded on 128 samples, but recording @512 samples works fine for my system. At lower sample settings my interface is prone to crackles and stuff, but when I start with midi only, I sometimes set it back to 256 samples.

It seems to be the fashion that recording at lower sample settings seems to better in a way..... I think you can do it if your interface is up to it, and my Focusrite ain't. For recording audio, I do use the direct monitor function of my interface, and I have an old Lexicon MPX1 patched in on s/pdif for having a monitor signal with some fx's if needed.

[thesheep]

It seems to be the fashion that recording at lower sample settings seems to better in a way..... I think you can do it if your interface is up to it, and my Focusrite ain't. For recording audio, I do use the direct monitor function of my interface, and I have an old Lexicon MPX1 patched in on s/pdif for having a monitor signal with some fx's if needed.

Yeah, that's the traditional way of doing it. But I bought this new Focusrite Thunderbolt interface on the promise of super low latency so I can monitor live without having to add in hardware effects for monitoring, etc. Especially when you consider being able to improvise/record electric guitar with amp modelling, etc directly in your DAW it seems like a cool idea. If I'd realised that other software synths are going to whack up the latency and stop me doing this then I probably would have got a cheaper USB audio interface!

I wonder if Logic Pro is any better at running software synths more efficiently?

[miyaru]

I don know Logic that well, last used it when they still had a Win version LOL.

[Nokatus]

The system I've used since 2012 can easily go to 64 samples at sample rates of 44-96 kHz. At 44.kHz, 32 samples is not a problem either. Going to such an extreme isn't often needed though, and I most often work at 128 or 256 simply because of the more relaxed CPU load. Still, the system can run a lot heavier projects without crackles than most similarly or somewhat higher specced computers (especially if they are general purpose ones), and it's nice to know these kinds of settings work so well when ever there's a need to perform a timing intensive MIDI part on a synth. The most typical use for the system is running project templates preloaded with Kontakt, Zebra, Diva and Reaktor instances, with track counts going to several hundreds.

I just tested Zebra 2 at 192 kHz and a buffer of 128 samples, that's a much shorter buffer than one millisecond at that sample rate, and it works without any crackles as well. No actual need to do that in any circumstances whatsoever, but it seems to work.

So, hmm. When ever I upgrade my main audio/music workstation, the system is for that use only, intended to be used for several years to come, and I go through all elements of it very carefully in order to maximise its realtime usability. This one has already been absolutely worth the effort.

Also, don't take this as bragging, it's really not a "super" system by any means these days. Just a coherent one , and well focused to its intended use. It's a Winodws 7 Pro 64 bit build, i7-3770k, Asus P8Z77-V, 32 GB RAM, with an RME HDSPe AIO interface. Unnecessary components like built-in wifi disabled, no separate graphics adapter. So the main point here is, in case you need extreme realtime performance, plan what you're getting, consider an interface with a history of the highest priority on stable low latency use (like this one), make sure the dealer where you are getting the system is willing to accept a return or modification in case the system doesn't yield low latency operation like this even after good planning, and even if it's otherwise "fully functional" (i.e. not normally a warranty issue in a general purpose system) -- and you'll get there.

[thesheep]

The system I've used since 2012 can easily go to 64 samples at sample rates of 44-96 kHz. At 44.kHz, 32 samples is not a problem either. ...
It's a Winodws 7 Pro 64 bit build, i7-3770k, Asus P8Z77-V, 32 GB EAM, with an RME HDSPe AIO interface. Unnecessary components like built-in wifi disabled, no separate graphics adapter. So the main point here is, in case you need extreme realtime performance, plan what you're getting, consider an interface with a history of the highest priority on stable low latency use (like this one), make sure the dealer where you are getting the system is willing to accept a return or modification if the system doesn't yield low latency operation like this, even if it's otherwise "fully functional" -- and you'll get there.

Interesting. I wonder if PCs do better than Macs on this. I stumbled on this thread and this thread that anecdotally seems to suggest that performance on Macs got worse in the jump from Live 8 to Live 9.

[Tarekith]

I have a 2017 MBP and just tried this using the Sentient Shadow preset with the built in soundcard at 64 samples. Plays fine, no crackles, and CPU use stays around 10-20% for most playing.

[thesheep]

I have a 2017 MBP and just tried this using the Sentient Shadow preset with the built in soundcard at 64 samples. Plays fine, no crackles, and CPU use stays around 10-20% for most playing.

Interesting, thanks for doing that. My CPU usage goes between 20 and 25% and does have some crackles (using built-in sound card) at 64 samples when I play this preset only (nothing else in the set).

I might have a slower clock speed than you as mine is 2.5GHz? If each instrument only uses 1 core then that might explain the difference I guess.

[Tarekith]

2.9gHz here, so that could be one thing.