24 postsPage 1 of 2
1, 2
When using SampleTank 3 I find I can only use lower latencies (e.g. for using a guitar processor plugin) if I load a separate instance of ST3 for every MIDI track, instead of one instance and using channels within ST3, when crackling occurs. I recently upgraded my PC and thought I'd get better overall performance and be able to achieve lower latencies, but this hasn't been the case. Is this because ST3 can't handle multi-processing or Studio One or both?

I found this in FAQs:
"Studio One now manages all multi-core and multi-processing related functions as part of Dropout Protection." So is S1 using all my cores or not?

Please note I am not a high-knowledge computer chap, just a musician/home recordist, so be gentle with me if the answer is something I should know! ;) :shock:

My songs

Intel i9 9940X @ 3.3GHz, 14 cores, 32GB RAM.
OS & Programs drive: 1TB SSD
Data drives: 1 x 1TB SSD drive, plus extra 1TB HDD drive
Windows 10 Pro 64 bit
Studio One 64 bit. Version: 4.6.1.55987
Cakewalk by BandLab
1 Lava Lamp
User avatar
by robertgray3 on Sat Feb 01, 2020 5:33 am
Are you using Green Z monitoring?

Mac OS X Catalina 10.15.7
Mac Pro 6.1
3 GHz 8-Core Intel Xeon E5
32 GB 1066 MHz DDR3
Dual AMD FirePro D500 3072 MB
Quantum 2
User avatar
by SkylineUK on Sat Feb 01, 2020 11:06 am
I confess I don't know what that is. :oops:
In Options, Audio Setup, Processing tab, there is a blue Z next to 'Audio Roundtrip' which is ~0.
And there is a green Z below for Instrument. Should these be at a certain setting to get over my problem with SampleTank?

My songs

Intel i9 9940X @ 3.3GHz, 14 cores, 32GB RAM.
OS & Programs drive: 1TB SSD
Data drives: 1 x 1TB SSD drive, plus extra 1TB HDD drive
Windows 10 Pro 64 bit
Studio One 64 bit. Version: 4.6.1.55987
Cakewalk by BandLab
1 Lava Lamp
User avatar
by robertgray3 on Sat Feb 01, 2020 11:27 am
Is there a green Z below your SampleTank channel in the Mixer (the one that has the little keyboard icon on it that you can click on to open the SampleTank window)?

Mac OS X Catalina 10.15.7
Mac Pro 6.1
3 GHz 8-Core Intel Xeon E5
32 GB 1066 MHz DDR3
Dual AMD FirePro D500 3072 MB
Quantum 2
User avatar
by Funkybot on Sat Feb 01, 2020 7:43 pm
SkylineUK wroteWhen using SampleTank 3 I find I can only use lower latencies (e.g. for using a guitar processor plugin) if I load a separate instance of ST3 for every MIDI track, instead of one instance and using channels within ST3, when crackling occurs.


This is a fairly complex subject. Let me try to explain some basics, as best as I understand them as a fellow layperson...

1. Some plugins can use multiple-cores themselves. U-he synths for example have a multicore option that can be enabled. But this introduces some additional CPU overhead overall. They even recommend you generally leave it up to the host, and that's exactly what most plugins do. The DAW will attempt to manage and schedule CPU usage across the available cores. Some DAWs are better at this than others.

2. It's important to note it only takes one stressed core to cause glitches, pops, etc. If one core is peaking at 100% CPU, and 7 cores are 2% CPU, your audio will still glitch and your project will struggle to play. How does that happen?

3. Well, it's HIGHLY dependent on your project and how you set it up. Generally, anything in a single audio path/chain has to run on one core/thread because everything in the chain happens in serial. Example: the EQ plugin impacts the sound of the compressor after it, which in turn impacts the sound of the delay after that. Then if they go to a bus with some additional processing, but that bus also has other instruments routed to it, then this processing must be done one after the other cannot be done in parallel. So if you have 10 tracks with processing running to 1 bus, those are all going to get parked on one core, which can easily become stressed, even if the other cores are hardly being taxed at all. If you split those 10 individual tracks up with no bus processing, the DAW could spread them across multiple cores, and suddenly your CPU usage might seem more balanced because a lot of that processing can now happen in parallel across multiple cores.

4. The same is true for most instruments. Lets say you have 16 Sampletank patches within 1 instance, running 5% for each patch. That could be 80% of one CPU core (5%*16=80%). Add some additional effects and you could easily max out an entire core with a single instrument. Now, split that up with multiple instances of Sampletank, and each core might only be using 10% of the CPU. Why? Because separate instances can run on separate threads, versus creating a bottleneck by having everything on one.

5. But...it's not that simple. Each instance of Sampletank might have it's own overhead. So maybe instead of each core using up only 10% in the above example, maybe each core uses 13%. Or maybe the graphics get loaded to RAM and suddenly there's less free memory available. It just depends on how things are coded and how efficient they are. Kontakt, for example, apparently runs pretty efficiently when used multi-timbraly in Studio One so maybe you do want to use fewer instances of Kontakt with more instruments per each instance. Again, it just kind of depends. Get to know your plugins, and experiment. Find what works better. Maybe loading 2 instances per Sampletank instance, and using 8 instances is much more efficient than 1 instance with 16 instruments. Or maybe 16 instances with 1 patch will work best on your processor/project. Maybe try less busing.

6. There are other factors as well. Does the DAW use a dual-buffer design or anticipative FX processing? Studio One, when in Green Z mode, will make a copy of any FX chains on input monitored and run those at a lower latency at the expense of CPU. Great for latency. Terrible for CPU.

7. Also, I recently listened to an interview with Justin Frankel of Reaper, and he brought up that super-mega core machines may cause people to change their workflow and do less processing on buses and auxes and do more via inserts, because it facilitates better parallel processing across large core counts. Just an interesting, and related concept.

Hope that helps explain some of how all of this works. It's not easy to wrap your head around, and at times can be counter intuitive. The best thing is to have a combination of the fastest per-core speed you can get (to prevent per-core overloads as much as possible) and also a large number of cores (to allow lots of tracks to run on separate cores) along with projects that are setup to favor parallel processing and load balancing by the DAW.

AMD Ryzen 3950X, ASUS Creator x570 Mobo, 32GB HyperX Predator RAM (3600mhz), Radeon™ RX 5500 XT 4GB GDDR6 graphics card, RME Fireface 800, Windows 10 Pro, Studio One 5, Reaper 6, Cubase 10.5, Avid Artist Mix (EuCon please), Behringer X-Touch One, MIDI Fighter Twister, various other MIDI control surfaces and hardware instruments
User avatar
by SkylineUK on Wed Feb 05, 2020 5:26 am
robertgray3 wroteIs there a green Z below your SampleTank channel in the Mixer (the one that has the little keyboard icon on it that you can click on to open the SampleTank window)?


No, I don't see that.

My songs

Intel i9 9940X @ 3.3GHz, 14 cores, 32GB RAM.
OS & Programs drive: 1TB SSD
Data drives: 1 x 1TB SSD drive, plus extra 1TB HDD drive
Windows 10 Pro 64 bit
Studio One 64 bit. Version: 4.6.1.55987
Cakewalk by BandLab
1 Lava Lamp
User avatar
by SkylineUK on Wed Feb 05, 2020 5:31 am
Funkybot wrote
SkylineUK wroteWhen using SampleTank 3 I find I can only use lower latencies (e.g. for using a guitar processor plugin) if I load a separate instance of ST3 for every MIDI track, instead of one instance and using channels within ST3, when crackling occurs.


This is a fairly complex subject. Let me try to explain some basics, as best as I understand them as a fellow layperson...
.....


Many thanks! That explains a lot and I understand a lot more now.

It seems remiss that IK Multimedia (for example) don't see fit to cater for multiprocessing in SampleTank3; not even in their major rebuild, SampleTank 4, when most PC DAW users will have mutlicore processors. Is it too difficult to program in I wonder? I understand Kontakt caters for it but my investment is in SampleTank3 and Halion Sonic.

I believe it's a long standing request on the S1 wish list, but is it a feature that the sampler vendors should incorporate or the DAW's? Again, I think if it was relatively straight forward it would already be catered for in one, or both.

My songs

Intel i9 9940X @ 3.3GHz, 14 cores, 32GB RAM.
OS & Programs drive: 1TB SSD
Data drives: 1 x 1TB SSD drive, plus extra 1TB HDD drive
Windows 10 Pro 64 bit
Studio One 64 bit. Version: 4.6.1.55987
Cakewalk by BandLab
1 Lava Lamp
User avatar
by Prog Rocker on Wed Feb 05, 2020 6:59 am
Funkybot wrote
So if you have 10 tracks with processing running to 1 bus, those are all going to get parked on one core, which can easily become stressed, even if the other cores are hardly being taxed at all. If you split those 10 individual tracks up with no bus processing, the DAW could spread them across multiple cores, and suddenly your CPU usage might seem more balanced because a lot of that processing can now happen in parallel across multiple cores.



Could you elaborate further on this. What do you mean by 'split those 10 individual tracks up with no bus processing' ?
Is this a situation where one has multiple tracks routed to a reverb ? And would it then be best to have a reverb on each individual track ?

Also how does Studio One in particular play into all this ?

Thank you...

Studio One 5 Artist 5.2.0
Ampire XT Metal Pack Version: 1.0.0.7
Studio One Plug-In Support Version: 1.0.0.0

Notion 6 Version: 6.4.462

Dell XPS Laptop
Microsoft Windows 10 Home 64-bit
Intel Core i5 540M @ 2.53GHz

Dell XPS 8300 Desktop
Microsoft Windows 10 Home 64-bit
Intel Core i7 2600 @ 3.4GHz
User avatar
by mikemanthei on Wed Feb 05, 2020 2:39 pm
Ummmm ...some of these assumptions are just.. not valid.

Funkybot wrote So if you have 10 tracks with processing running to 1 bus, those are all going to get parked on one core, which can easily become stressed, even if the other cores are hardly being taxed at all. If you split those 10 individual tracks up with no bus processing, the DAW could spread them across multiple cores...

Interesting you would make such claims when (a) Not True and (b) so easily tested and disproven.

First off, there is no magic to removing your tracks from a bus. Think logically here... the "Main" bus is a bus, so you already have all your tracks assigned to a single bus. So obviously Studio One can load balance tracks that are assigned to a bus. So adding your own busses doesn't break anything. And it's so easy to test for yourself. Here's a benchmark load I put on my system to test. Note that I have quite a few cores, so it's easy to tell when something is spreading across available processors or not.
I took 100 tracks and played them through the main bus, then mid-way through this test, I re-assigned them to just 3 busses. If your guess was correct, I should have had 3 processors pegged and 21 cores at idle. But the load stayed the same.

Load Balancing.png
Just in case you're wondering, this is 75 tracks of MaiTai and 25 tracks of Presence. A sort of All-Presonus-In-The-Box stress test. I don't know if Studio One will utilize all 128 threads of the new Threadripper CPU, but I know that it had no trouble with 24. Someone else will have to pay the entrance fee for that test. :)

Prog Rocker wroteCould you elaborate further on this. What do you mean by 'split those 10 individual tracks up with no bus processing' ?
Is this a situation where one has multiple tracks routed to a reverb ? And would it then be best to have a reverb on each individual track ?


I don't think so. Putting 10 reverbs in your DAW creates 10 times the load that one does, so even if there are some efficiencies to one-reverb-per-channel, I don't think it will make up for a 10x increase in processing. And then there's the tendency to "tweak" each reverb so it's perfect for the individual track, but that's just a long dark road.
Also, It has long been known that putting most of your group through a single reverb does 2 things. 1) It reduces the load on your computer by only having one instance of the plugin.. and 2) It can make the mix sound more "natural" or "cohesive". ...you know, as if the band is really playing together. Remember that your audience will never say: "Wow,that's the perfect reverb for that oboe".. but they will know if it sounds natural or disjointed. Sometimes using too many different reverbs on a mix is like... too many different fonts in a document.

I looked into the multi-threaded options in Kontakt at one point. I read where someone said to make sure that Kontakt was set up to use multi-cores... But according to Kontakt's documentation, this setting is only for those times when you're using the Kontakt player as a standalone. If you're using Kontakt as a plugin, then that functionality is passed on to the DAW.

Still, the tests has to be done. :) I tried 60 Kontakt instruments in a single Kontakt player, and then those same 60 each in their own player. The load was the same.

Studio One v2, 3, and 4 Professional
Presonus 1818VSL / Focusrite 18i20 / StudioLive 32S
24-core Ryzen 9. 32 GB RAM
Tascam US-2400
Faderport 8
StudioLive 32S
User avatar
by darrenporter1 on Wed Feb 05, 2020 3:25 pm
He's not talking about sending stuff to buses.... he's talking about sending stuff to buses and then processing those buses with effects as opposed to processing each channel before it gets to the bus.

A more realistic attempt to disprove it would be to put a CPU-intensive reverb on an FX bus, like Abbey Road Plates. Put it as a SEND on each of those 100 synth instances. Then remove it and put it on each of the three buses instead and see if that makes a difference.


Studio One Professional 5.whatever, Harrison MixBus 32c v.6
i5-8400, 16GB RAM, 512 GB SSD, 2TB HD, Win10 Pro
UA Apollo QUAD, QUAD Satellite, PCIe DUO
FaderPort 8, Softube Console 1, JBL 306P Mk.II Monitors
User avatar
by SkylineUK on Wed Feb 05, 2020 4:34 pm
mikemanthei wrote:
[quote="...........snip

Still, the tests has to be done. :) I tried 60 Kontakt instruments in a single Kontakt player, and then those same 60 each in their own player. The load was the same.[/quote]


That must be because Kontakt can handle multiprocessing. Alas, not for SampleTank 3. I have a project with 14 MIDI tracks. If I put each of them through 14 instances of SampleTank 3 I can get away with setting my RME Aio to 128 samples. If I put them all through one instance on different channels I get immediate audio drop out and have to lift the samples to at least 1024 to get the project to play.

My songs

Intel i9 9940X @ 3.3GHz, 14 cores, 32GB RAM.
OS & Programs drive: 1TB SSD
Data drives: 1 x 1TB SSD drive, plus extra 1TB HDD drive
Windows 10 Pro 64 bit
Studio One 64 bit. Version: 4.6.1.55987
Cakewalk by BandLab
1 Lava Lamp
User avatar
by garyanderson5 on Wed Feb 05, 2020 6:28 pm
SkylineUK wrotemikemanthei wrote:
[quote="...........snip

Still, the tests has to be done. :) I tried 60 Kontakt instruments in a single Kontakt player, and then those same 60 each in their own player. The load was the same.


That must be because Kontakt can handle multiprocessing. Alas, not for SampleTank 3. I have a project with 14 MIDI tracks. If I put each of them through 14 instances of SampleTank 3 I can get away with setting my RME Aio to 128 samples. If I put them all through one instance on different channels I get immediate audio drop out and have to lift the samples to at least 1024 to get the project to play.[/quote]

There is a really easy way to test how any DAW handles multiprocessing.

Load a single audio track. Insert a plugin that uses say 10% cpu for example. Copy it a few times on the same track inserts till you have say 80% cpu load in the DAW. Now open task mannager in windows you will see your system cpu usage is no were near the DAW's 80%. This is because your pegging one core\thread as a single process. Most plugins do not do multiprocessing you have to micro mannage them yourself inside the DAW. Reaper may be an exception here but don't quote me on it.

Now do the same thing but put each plugin on its own individual track and it will load far more instances upto that 80% as it's multiprocessing at work. Open task mannager again and your system usage should be very close to the Daw usage 80% meaning it matches and its using all your cores.

Studio one like most DAW's can do multiprocessing it's just as soon as you process multiple tracks through a bus or fx channel via sends or routing etc with plugins on the bus\group or fx channel there sent into it becomes one process as you made it a serial process not a parallel one.There are not many multiprocessing plugins about that can help with this. Kontakt can for example hence why it can handle things well.

You could have all the cores in the world for example but if you load plugins on your main bus with everything sent to it no amount of cores in the world will save your bleep regardless of the chip type. Core speed helps for this but even a 5GB 9900K will tap out eventually.

Seperation of tracks with there own bus and fx channels, freeze & bounce are your best tools if you do intense plugin chains.Duplicate tracks complete is a great way to spread load in Studio one for example.

I am by no means an expert but this aproach has worked for me ever sinse i used Cubase with the first quadcore chips and the way you micro mannage your tracks and plugins is no different in Studio one. The fact that you had to splitt your Sample tank up into 14 instances\tracks means you basicly did as i explained above.

Funkybot explains it really well to be honest.

Windows Pro 11 23H2, 13900K, Z790 Aorus Master, 64GB 32x2 G.Skill Trident Z C30 RAM, 3X 2TB Samsung 980 PRO, RX6650XT GFX Card, Corsair RM1000X PSU, Studio One Latest Version & Older versions, Baby Face Pro FS. Nektar T6,
User avatar
by Prog Rocker on Thu Feb 06, 2020 6:42 am
mikemanthei wrote
Prog Rocker wroteCould you elaborate further on this. What do you mean by 'split those 10 individual tracks up with no bus processing' ?
Is this a situation where one has multiple tracks routed to a reverb ? And would it then be best to have a reverb on each individual track ?


I don't think so. Putting 10 reverbs in your DAW creates 10 times the load that one does, so even if there are some efficiencies to one-reverb-per-channel, I don't think it will make up for a 10x increase in processing. .



Thanks for the reply Mike. Being new to digital audio I didn't quite understand what Funkybot meant by...

'So if you have 10 tracks with processing running to 1 bus, those are all going to get parked on one core, which can easily become stressed, even if the other cores are hardly being taxed at all. If you split those 10 individual tracks up with no bus processing, the DAW could spread them across multiple cores, and suddenly your CPU usage might seem more balanced because a lot of that processing can now happen in parallel across multiple cores. '

...and why that would be the preferred way to go. Maybe I just misunderstood his comment.

I recently watched a video stating that DAWs rely on serial processing more so than parallel processing. So single core performance is therefore more important than having multiple cores.

I was wondering first of all if that was true, and secondly if Funkbot's comments tied into this ? And as a Studio One user, how SO goes about this ?



phpBB [video]

Studio One 5 Artist 5.2.0
Ampire XT Metal Pack Version: 1.0.0.7
Studio One Plug-In Support Version: 1.0.0.0

Notion 6 Version: 6.4.462

Dell XPS Laptop
Microsoft Windows 10 Home 64-bit
Intel Core i5 540M @ 2.53GHz

Dell XPS 8300 Desktop
Microsoft Windows 10 Home 64-bit
Intel Core i7 2600 @ 3.4GHz
User avatar
by garyanderson5 on Thu Feb 06, 2020 8:45 am
Prog Rocker wrote
mikemanthei wrote
Prog Rocker wroteCould you elaborate further on this. What do you mean by 'split those 10 individual tracks up with no bus processing' ?
Is this a situation where one has multiple tracks routed to a reverb ? And would it then be best to have a reverb on each individual track ?


I don't think so. Putting 10 reverbs in your DAW creates 10 times the load that one does, so even if there are some efficiencies to one-reverb-per-channel, I don't think it will make up for a 10x increase in processing. .



Thanks for the reply Mike. Being new to digital audio I didn't quite understand what Funkybot meant by...

'So if you have 10 tracks with processing running to 1 bus, those are all going to get parked on one core, which can easily become stressed, even if the other cores are hardly being taxed at all. If you split those 10 individual tracks up with no bus processing, the DAW could spread them across multiple cores, and suddenly your CPU usage might seem more balanced because a lot of that processing can now happen in parallel across multiple cores. '

...and why that would be the preferred way to go. Maybe I just misunderstood his comment.

I recently watched a video stating that DAWs rely on serial processing more so than parallel processing. So single core performance is therefore more important than having multiple cores.

I was wondering first of all if that was true, and secondly if Funkbot's comments tied into this ? And as a Studio One user, how SO goes about this ?



phpBB [video]


Yes it is true and it connects to Funkybots explination. Core speed is important regardless of what people tell you if your a heavy virtual instuments and plugins user in serial.

Example: If all you do is record and mix with plugins in parallel ie 10 inputs for drum mic's with 2 or 3 plugins on each insert or 10 audio stem tracks, an AMD 3900x 12 core at 4.2GHZ on all cores or simular Intel chip for example are perfect and more than enough for most things. But if you use a heavy vst instrument with 2 or 3 plugins on an eight core 9900k at 5GHZ it will allow you much more intense plugin chains in serial before it chokes as it has more headroom. The core speed tends to reduce the more cores a chip has so it's a trade off or finding a sweet spot of max core speed and cores combined. This is why the new AMD 3900x 3950x and Intels 9900k are populer for example they offer good core speed and a good amount of cores for both serial and parallel processing. The milage varies per user of course and there workflow.

Most heavy VSTI and plugin users will still bounce down or freeze at some stage as every chip has a limit. Plugins differ from chip to chip system to system also down to the way there made. Newer chips can handle more instructions per core over say a 5 year old chip for example wich helps with more plugins in serial as a single process. Your system, audio driver, latency etc all play a part in this also but thats another topic on it's own.

There is a "today we build our studio pc" thread over at Gearslutz forum that has some great uptodate info on it. Hope this helps.

Windows Pro 11 23H2, 13900K, Z790 Aorus Master, 64GB 32x2 G.Skill Trident Z C30 RAM, 3X 2TB Samsung 980 PRO, RX6650XT GFX Card, Corsair RM1000X PSU, Studio One Latest Version & Older versions, Baby Face Pro FS. Nektar T6,
User avatar
by Prog Rocker on Sat Feb 08, 2020 5:00 pm
garyanderson5 wrote
Yes it is true and it connects to Funkybots explination. Core speed is important regardless of what people tell you if your a heavy virtual instuments and plugins user in serial.
There is a "today we build our studio pc" thread over at Gearslutz forum that has some great uptodate info on it. Hope this helps.


Thanks Gary, that was very helpful.

-

Studio One 5 Artist 5.2.0
Ampire XT Metal Pack Version: 1.0.0.7
Studio One Plug-In Support Version: 1.0.0.0

Notion 6 Version: 6.4.462

Dell XPS Laptop
Microsoft Windows 10 Home 64-bit
Intel Core i5 540M @ 2.53GHz

Dell XPS 8300 Desktop
Microsoft Windows 10 Home 64-bit
Intel Core i7 2600 @ 3.4GHz
User avatar
by AnthemMakersMusic on Mon Jul 12, 2021 10:18 pm
We are all wondering if this thread ripping thing will be the next step in audio and video. It being marketed as the greatest thing since grits, eggs and cheese for breakfast. No seriously, is it the future of audio and is anyone writing code for it like they do for mac and Windows. Right now this thing seems like a different beast entirely. Help tech Jedi guide us!
User avatar
by reggie1979beatz on Tue Jul 13, 2021 8:52 am
When I did research for a new computer, I was told again and again that the larger the core the better...as opposed to many more cores.

Bye......:roll:
User avatar
by mikemanthei on Tue Jul 13, 2021 12:12 pm
reggie1979beatz wroteWhen I did research for a new computer, I was told again and again that the larger the core the better...as opposed to many more cores.


What you have been told is a simplistic answer that is only true if the application you want to run is single-threaded and/or can only use one core.

It's just not true for most modern applications.

Studio One v2, 3, and 4 Professional
Presonus 1818VSL / Focusrite 18i20 / StudioLive 32S
24-core Ryzen 9. 32 GB RAM
Tascam US-2400
Faderport 8
StudioLive 32S
User avatar
by reggie1979beatz on Tue Jul 13, 2021 12:42 pm
That's fine, but I've heard too many conflicting reports to just accept it. On my machine with the plugs I use, it's certainly not evenly distributed. No, I'm not saying it doesn't use other cores but it CLEARLY uses the first one the most. Unless task manager is just completely wrong.

Bye......:roll:
User avatar
by IanM5 on Tue Jul 13, 2021 12:49 pm
mikemanthei wrote
reggie1979beatz wroteWhen I did research for a new computer, I was told again and again that the larger the core the better...as opposed to many more cores.


What you have been told is a simplistic answer that is only true if the application you want to run is single-threaded and/or can only use one core.

It's just not true for most modern applications.


But it is true for most real-time audio applications. Most of the cores end up waiting for the heavily loaded one.

Heavy-handed moderation can strangle a forum

24 postsPage 1 of 2
1, 2

Who is online

Users browsing this forum: wolfgerb and 41 guests