Audio break-up for some patches

#1

I’m getting audio break-up on some patches, more complex ones I think. This is not affected by the choice of audio driver, sample rate or buffer size. It is affected by the amount of modulation, amount of effects and in particular, the number of voices (unison or keyboard). As an example, playing a single note with voices set to 5 is OK but setting the voices higher (say 16) breaks up.

Other synths will suffer audio break-up at some point but always the CPU load is at 100%, this Vital break up is unique in that the CPU load is relatively low (say 65%). Also the break-up is kind of rythmic, a slow pulsing noise which continues with no keys pressed.

This is my first post, so hello to all.

This is an AMD A9-9410 CPU.

#2

Can you give us some more information on your system:

  1. What is you OperatingSystem
  2. You need to watch if specific Cores are getting overloaded - not the full CPU with a Multicore CPU.

Looking up the Specs of your CPU - that looks like a entry level Notebook CPU with 15W TDP. At that point in time (2016) as they got released and as of today, those have very limited CPU power and depending on workload may run into bottlenecks, even if CPU Utilization does not show 100%. That is because of the architecture of the chip.

Not the best option IMO

#3

Some users are reporting high CPU usage (though most users see a reasonable amount) and I haven’t tracked down all the reasons.

Would help if I got a little more information like what your OS and version is.

#4

Thanks for the prompt replies, this is a Windows 10 home system, version 1909. I agree that this is a rather basic cpu but can only say that this problem has not happened elsewhere with only moderate cpu usage. Sorry, don’t know how to see individual cpu core usage.

#5

Hears an odd recording, playing “keystation” patch, first with original long release, and then with short release. Notice how the break up persists until the end of the release.

#6

For individual core CPU usage, open the task manager - switch to performance and then click on the bottom to Resource Monitor. Here you have it … all cores seen individually.

#7

Thanks for that, both cpu0 and cpu1 show approx’ 70%.

#8

Yeah there’s a combination of things at play, I think the integrated graphics card causes some extra heat from the animations which slow down the cpu process. Then my FFT library is a little bit slower on windows machines with AMD CPUs.

I don’t really have an immediate fix or solution but doing some graphics optimizations and a possible faster FFT library for AMD+Windows systems are things I’m looking into improving in the long run.

#9

Similar here, decent system, linux on i7-4790k, 32GB RAM, RX570 gfx.

#10

The geekbench score on the 9410 multicore is about 3500 - I find that you need at least 5000 to handle heavier leads and multi-track instances. Best (these days imho) you’d aim for 7000+.

#11

wow … my i7 is from 2016 or something at that time and a single core has 5000 + the whole thing >32k

I was not aware how poor the AMD A9-9410 was.

@ewen.bates
Can you please give use more info?
What Linux are you running on?
What WindowManager
What Driver (OpenSource or closed source)
What OpenGL layer?

Linux is a complete different set of problems than Windows. - I recommend you open a new thread in support for it, as each Linux issue is more or less an individual one.

From CPU Side of things (actually anything in your PC) I cannot see any issue. Hardware is great, Software will be an issue. There are some threads out there for Linux, that point back to suboptimal OpenGL Chain.

#12

Operating System: Linux Mint 19.1 with Xfce
Kernel: Linux 4.15.0-135-lowlatency
Architecture: x86-64

I host Vital in VCV (no issues since upgrading from integrated gfx to RX570) & route everything through Harrison Mixbus.

#13

You may want to increase the buffers in your settings, maybe that works. It sounds that the CPU cannot deliver fast enough data to the buffer, and the buffer runs out of data.

#14

Did you use the .deb file to install? or the zip?

#15

.deb

And while I’m replying, and to make up the minimum post length, thanks for an awesome synth Matt, nicely done.

#16

Can you clarify your CPU benchmark scores? 5000 for a single core on geekbench seems unlikely. Take a look at the current top geekbench scores:

https://browser.geekbench.com/v5/cpu/singlecore

Only one processor is better than 5K, and the rest are all under 3K.

I’ve struggled since moving from a MBP to Windows – Vital and Ableton are having a hard time. I’m trying to figure out the most cost-effective way to improve this. I’m almost certain I need to go back to a Mac.

Even when Ableton+Vital aren’t breaking up on this Windows machine, I have to constantly change the buffer size to listen to audio in Chrome, and then back to a different size for Ableton --it’s maddening; I have tried about 4 or 5 ASIO drivers for the built-in audio, and I added a Focurite solo. I’m just realizing that some windows machines are not built for this kind of workload.

But before I take the leap into that super expensive Mac hardware, I want to make sure I’ve looked at other Windows options. I just don’t know how you can be sure that Ableton and other audio software will run well on a given Windows machine. It seems like it’s more than just CPU speed. Things like audio hardware (rarely advertised in the specs, from what I can tell), the drivers for that hardware, even the graphics adapter all play a role. Not to mention the shovelware that the OEM throws on there for “enhancing your audio”, or options that they take away from the control panels. What a freaking mess!

And you can’t really know any of this stuff until you test drive it. Maybe I just have to buy one, install the Ableton trial, and try it out for a few days and return it if it sucks.

#17

Maybe the confusion comes from geekbench version and I did not pay attention…

Relations:
I have a core i7 5930k … which was at the time it has been released a beefy CPU.

Geekbench4 is:
SC MC
5000-5700 25k-30k

Geekbench5 is:
SC MC
~1000 ~6000

So it is really a difference at what you look at, so my numbers was from Geekbench4 and I was not even ware that Version5 cut the same numbers.

In comparison Geekbench 5 numbers for AMD 9410 are:
SC MC
~350 ~550

So I assume (as GB5 means that 5k for MC is pretty beefy), that 5k SC was meant for GB4 results.

My specific numbers (on my computer) are
1004 and 5734

Looking at AMD … I have another CPU in my ESX lab Server … 3900x
GB4: maxes @6.7k and 60k
GB5: 1300, 12k

So depends on what you are looking to. definatly no issue at all with GB4 5k MC … and absolutely nothing within the range of a i7 5930k … (30k). and that is a pretty old CPU. modern CPUs will deliver even more (AMD 5900x: 7.3k and 60k)

Notebooks general suck in CPU performance… it is not at all value (CPU performance) for money.
And just comparing a Windows Notebook to a MacBookPro does not help at all. Manufacturers just put in all kinds of Marketing to let you thinks that performance is great. But simplified a i7 in a Notebook today has nothing to do with a i7 in my PC (self build btw). The notebook CPU has a maximum heat dissipation based on the air volume available - that of cause means that a CPU in a notebook cannot dissipate more than 15w constantly … where I have a water cooled system which can transport >150w constantly. So obviously that makes a difference.

And tbh Windows was never and never will be optimized… Generally talking (IMO) I do not think that any notebook is a great machine for a Audio workstation…