CPU usage and 96khz Questions about your system stability

[hearteh]

I have a couple of questions about your systems and cpu limit before running into problem.

1. How high CPU usage can you go at 96khz 64 buffer with the cpu usage simulator and the test tone in ableton audio setup before crackles?
2. What soundcard are you using?
3. Mac or Pc?

[wearemindflux]

Hey are you wanting this help?

Or asking for each persons system?

[hearteh]

Hey are you wanting this help?

Or asking for each persons system?

How do you mean wanting this help? Ok I changed the question a little hopefully its more clear

[fishmonkey]

i don't see how that is a useful question. what would that tell you (or anyone else)?

[mihai]

I don't know about 64 samples but I can run 40-60 hot channels at 24/96 with plenty of 3rd party plugins on a 6700 at 1024 with no freeze. I don't do much tracking though I imagine it wouldn't be too great based on experience. That being said I'd use something else for tracking in the case of a large channel count.

[hearteh]

This is not a question about how many channels or plugins you can use. I want to know how good your audio driver and system is since I think I might have problem with my new Apollo x6. So its the cpu usage before crackling thats interesting.

It takes a couple of seconds to do the test. There is actually a test tone in the audio setup in Ableton to test this. 128 buffer would be nice to know as well.

I can go to around 58% and have a super fast PC and Ive heard another person with old Apollo and Mac that can go full 80% with a much slower system.

[fishmonkey]

sorry, waste of time. too many variables.

what do you mean by "CPU usage" anyway? are you referring to Live's CPU load meter? do you understand what that tells you?

[hearteh]

sorry, waste of time. too many variables.

what do you mean by "CPU usage" anyway? are you referring to Live's CPU load meter? do you understand what that tells you?

Best forum ever. Learn how the test tone and cpu usage simulator works its in the manual..

[jestermgee]

I am certain fishmonkey knows how to use the CPU meter, he's been around long enough.

He is confused (as am I) what it is exactly you wish to "confirm" since CPU load and what causes your system to start to break will differ completely between systems.

For a start, if you want comparable results from anyone it might be a good idea to replace this kind of lingo:

and have a super fast PC

Ive heard another person with old Apollo and Mac that can go full 80% with a much slower system.

With actual specs of your system. How fast your computer can complete the 100m sprint dash isn't applicable without some specs. The processor, Ram, GPU, Audio Interface and things like track count, VSTs loaded etc would be much more useful details to work with.

If you have issues with crackle, simply increase your buffer from 64 to 128 for a start. It will make little noticeable difference but allow more processor time for your audio stream. I run at 256 for extra stability and I also run my sample rate at 48Khz which is something else you could do depending on what you are doing. Forget trying to be the winner of the highest sample rate and lowest buffer setting, if you have issues adjust your numbers and retest.

FYI, I can run the test tone at max (80%) running at 192Khz and 192 sample buffer (lowest my buffer goes at such a high sample rate) and it's a smooth tone without a single pop. What that tells me is I can run the test tone at 80% CPU and 192 samples without an issue. Does not translate tho to running an actual set up to 80% CPU with these settings tho. Much more business happens when running multiple plugins and FX.

For Reference:
Windows 10 64bit
Live 10
i7 4820K @ 4Ghz
16GB 2400 RAM
512G Samsung SSD
Roland OctaCapture Interface

[wearemindflux]

sorry, waste of time. too many variables.

what do you mean by "CPU usage" anyway? are you referring to Live's CPU load meter? do you understand what that tells you?

Best forum ever. Learn how the test tone and cpu usage simulator works its in the manual..

Sorry but this is a great forum, you simply have not be at all clear it what you are asking and what the purpose of the questions are. I would make sure your own house is in order before getting ratty with our people.

[hearteh]

Yes it will be completely different between system thats why Im asking. I was just curious about how high people could go since it tells me something about the quality of your audio drivers and system and how it compares to my system, no?

It seem people mistake me for a noob looking at my post count and want to set me straight. I know I can higher the buffer setting but then you get a higher latency. Im used to hardware synths connected to analog mixer so this latency is very annoying to me thats why Im trying to work at lowest buffer as possible.

Thanks for the first answer to my question. I thought most sound card could test at 64 buffer maybe thats another reason I don't get any answers.

Here is my system. Since I said someone with a less system and and old Apollo could go much higher with the cpu usage simulator thats why I was curious. Still I can use much more plugins than him but he can use much more of his cpu before crackles at this low buffer than I could. The cpu usage simulator tells you something right or why is it there? An actual setup at 80% would be hard to compare.

UAD apollo x6
Windows 10 64 bit
I7 8700k overclocked at 5ghz
Corsair vengeance 16gb ddr4 3200mhz
Asus prime z370-a
Asus ex3
250gb ssd 860 evo
250gb m2 970 evo
500gb m2 970 evo

I am certain fishmonkey knows how to use the CPU meter, he's been around long enough.

He is confused (as am I) what it is exactly you wish to "confirm" since CPU load and what causes your system to start to break will differ completely between systems.

For a start, if you want comparable results from anyone it might be a good idea to replace this kind of lingo:

and have a super fast PC

Ive heard another person with old Apollo and Mac that can go full 80% with a much slower system.

With actual specs of your system. How fast your computer can complete the 100m sprint dash isn't applicable without some specs. The processor, Ram, GPU, Audio Interface and things like track count, VSTs loaded etc would be much more useful details to work with.

If you have issues with crackle, simply increase your buffer from 64 to 128 for a start. It will make little noticeable difference but allow more processor time for your audio stream. I run at 256 for extra stability and I also run my sample rate at 48Khz which is something else you could do depending on what you are doing. Forget trying to be the winner of the highest sample rate and lowest buffer setting, if you have issues adjust your numbers and retest.

FYI, I can run the test tone at max (80%) running at 192Khz and 192 sample buffer (lowest my buffer goes at such a high sample rate) and it's a smooth tone without a single pop. What that tells me is I can run the test tone at 80% CPU and 192 samples without an issue. Does not translate tho to running an actual set up to 80% CPU with these settings tho. Much more business happens when running multiple plugins and FX.

For Reference:
Windows 10 64bit
Live 10
i7 4820K @ 4Ghz
16GB 2400 RAM
512G Samsung SSD
Roland OctaCapture Interface

[hearteh]

I think my question is pretty clear. Im the one that was rude, really? Maybe read the answers again to my simple question.

sorry, waste of time. too many variables.

what do you mean by "CPU usage" anyway? are you referring to Live's CPU load meter? do you understand what that tells you?

Best forum ever. Learn how the test tone and cpu usage simulator works its in the manual..

Sorry but this is a great forum, you simply have not be at all clear it what you are asking and what the purpose of the questions are. I would make sure your own house is in order before getting ratty with our people.

[jestermgee]

It seem people mistake me for a noob looking at my post count and want to set me straight.

You get mistaken for a noob not just because of your post count but because of the way in which you asked the question and did not provide details of your issue (or any system specs etc), just requesting people supply test data for a reason that was unknown and of course none of us have anything more to go by than what you have posted.

Anyway...

Im used to hardware synths connected to analog mixer so this latency is very annoying to me thats why Im trying to work at lowest buffer as possible.

You need to increase buffer or decrease sample rate, simple as that.

No one likes latency but I am not sure you are grasping what a human can actually even detect as latency. 96Khz at a buffer of 64 is ridiculously demanding and basically unachievable (especially in a windows environment that has not been painstakingly modified for low latency processing). You need to understand some maths and what your expecting from your machine. Time for a maths lesson to get you understanding what your expecting...

Lets look at a typical audio configuration:

Sample Rate: 44.1Khz
Buffer: 128
128 / 44100 = 0.0029 or 2.9mS delay

Now, no one on planet earth can detect a latency of 0.0029 seconds

And here is what your expectations are:

64 / 96000 = 0.00066 (or 0.6mS)

You want your system to crunch large amounts of realtime numbers from multiple threads in LESS than 1mS and output that to your ears without missing a beat. Not gonna happen. Our old friend physics is at play so as your sample rate climbs, there are more samples per second to process which then means having such a low buffer allows absolutely no time to do any of the work. At that samplerate you should have it at 128 minimum (256 would be more reliable)

128 / 96000 = 0.0013 (1.3mS)

That would be the equivalent of 48Khz at 64

64 / 48000 = 0.0013 (1.3mS)

There are some other elements in the system which also add latency delay so what you really experience will be more inflated but there is also other things for your system to keep running too so the lower the latency the harder it is to maintain reliability.

[hearteh]

My output latency at 128 samples and 44.1khz is 3.38ms in my setup. I can detect the latency clearly. Play a hardware minimoog directly to the speakers and then play a soft synth at that setting. Ask a concert pianist if he likes to play at that latency.

I don't need any math lesson but thanks anyway. RME can provide drivers that work at this buffer and even better so I don't see why UAD couldn't do it for a PC.

I have not even said I would run at 96khz 64 buffer. At 128 buffer I can go all the way on the cpu usage simulator thats why I asked at 64 buffer and I can still not use Serum with high voice count at this setting.

It seem people mistake me for a noob looking at my post count and want to set me straight.

You get mistaken for a noob not just because of your post count but because of the way in which you asked the question and did not provide details of your issue (or any system specs etc), just requesting people supply test data for a reason that was unknown and of course none of us have anything more to go by than what you have posted.

Anyway...

Im used to hardware synths connected to analog mixer so this latency is very annoying to me thats why Im trying to work at lowest buffer as possible.

You need to increase buffer or decrease sample rate, simple as that.

No one likes latency but I am not sure you are grasping what a human can actually even detect as latency. 96Khz at a buffer of 64 is ridiculously demanding and basically unachievable (especially in a windows environment that has not been painstakingly modified for low latency processing). You need to understand some maths and what your expecting from your machine. Time for a maths lesson to get you understanding what your expecting...

Lets look at a typical audio configuration:

Sample Rate: 44.1Khz
Buffer: 128
128 / 44100 = 0.0029 or 2.9mS delay

Now, no one on planet earth can detect a latency of 0.0029 seconds

And here is what your expectations are:

64 / 96000 = 0.00066 (or 0.6mS)

You want your system to crunch large amounts of realtime numbers from multiple threads in LESS than 1mS and output that to your ears without missing a beat. Not gonna happen. Our old friend physics is at play so as your sample rate climbs, there are more samples per second to process which then means having such a low buffer allows absolutely no time to do any of the work. At that samplerate you should have it at 128 minimum (256 would be more reliable)

128 / 96000 = 0.0013 (1.3mS)

That would be the equivalent of 48Khz at 64

64 / 48000 = 0.0013 (1.3mS)

There are some other elements in the system which also add latency delay so what you really experience will be more inflated but there is also other things for your system to keep running too so the lower the latency the harder it is to maintain reliability.

[jestermgee]

My output latency at 128 samples and 44.1khz is 3.38ms in my setup. I can detect the latency clearly.

That's about where it should be and you are saying that when you press a note, 0.003 seconds between the key press and the sound is clearly noticeable? You either have some special sensory powers or have something else going on. I do not know any musician that could not adjust to that kind of delay.

If it is that bothersome, probably need to stick to analog gear. Good luck.

RME can provide drivers that work at this buffer and even better so I don't see why UAD couldn't do it for a PC.

96/64 may "work" but how well and with how much going on? Secondly to this, RME and UAD are different companies with different drivers designed by different teams. RME is well known for the quality of their drivers and their low latency so it's not a fair comparison. I really don't know much about UAD to compare myself.

I have not even said I would run at 96khz 64 buffer.

Except in your original post, question 1:

1. How high CPU usage can you go at 96khz 64 buffer

If you are not interested to run at that, why even ask? Again, I have no idea what the purpose of all this is. It would have been a much better thread if you said "I have XYZ audio interface set to ABC123 settings and I am seeing a CPU load of X% and experiencing some dropouts. Any ideas"?

You could have other issues in your system causing you problems especially if you have not tested in detail all your components and busses with your overclocking. Drops in voltage to CPU/Mem or issues with the memory bus due to incompatible overclocking could cause some issues when you load up cores. First thing I would recommend is setting your CPU back to stock and doing your tests to make sure it isn't anything you have done there. May probably be an idea to collect your findings and contact UAD support and see what their input would be.