Still getiing pops and crackles with Diva

[rod_zero]

Hi, hope someone can help me.

So I upgraded my computer to an i7 4790k, from an i5, to avoid cracking and pops with Diva.

While the CPU actually came down and I can play DIva on divine without cuts I still get some Pops from time to time, even with CPU load way below 50%. Multicore option just makes it worse.

All this on a empty session, just using one instance of Diva.

I am on windows 7 and Ableton live 9.6 and using an RME Fireface UC, settings: 441.Khz sample rate and the pops are present at 64, 128, 256 and even 512 samples in buffer size.

Any idea what could be the problem?

[ghettosynth]

Likely that some other driver or process is raising havoc with DPC latency.

http://www.thesycon.de/deu/latency_check.shtml

It's not all that easy to debug, it can be any number of things. However, you want to run the DPC latency checker and see where you are.

Turning on multicore shouldn't make it worse, but I'll defer to the experts at Uhe.

Also check your power settings. Power saving modes can raise havoc with your CPU, which might also explain why it's worse on mulitcore.

[rod_zero]

Power settings are at max, all saving options dissable.

I will check with DPC latency but I don't have many process running behind.

[ghettosynth]

Power settings are at max, all saving options dissable.

I will check with DPC latency but I don't have many process running behind.

Could be other drivers.

[rod_zero]

I used the DPC lattency checker and everything is OK, latency well under the acceptable limit.

Just know I am playing and no pops and crackles. I wonder if maybe chaging the buffer size once could introduce the pops and crackels.

[ghettosynth]

I used the DPC lattency checker and everything is OK, latency well under the acceptable limit.

Just know I am playing and no pops and crackles. I wonder if maybe chaging the buffer size once could introduce the pops and crackels.

I would put the blame on something else running in the background.

[#rob]

Disable Diva's Multicore option when using it in Live.

Where did you get the CPU % value from? If you used Live's performance meter, don't.
Instead, use your operating system's tools like Windows Task Manager or OSX Activity Monitor.

If your host and interface let you, then try raising the buffer size to 1024 or even 2048.
Not for productive use, but only to see if the crackling stops.

Please also consider that the crackling might not come from the system overloading after all, but possibly from the patches themselves.

- Loud patches can clip if too many voices stack up over each other, so make sure that Diva's output doesn't clip your host.
- Envelopes with very fast attack/release phases or fast modulation may also play a role.
- Especially if patches with many stacked voices and long release times are played fast or with polyphony, voice stealing might (have to) occur which may be perceived as pops or crackles when happening a lot or at inconvenient times.

So please confirm that the patches themselves are not causing the issues.

[rod_zero]

Hi Rob,

- I checked that patches weren't clipping.
- No clicking from the envelopes, I got this with all kind of patches, specially pads.
- Voice stealin, could be, but I was trying just holding a chord and getting the pops, before playing new notes.

Yeah, I cheked Live CPU meter, will check the system one next time I get this behaviour. Multicore option increases this, so I turned it off.

I didn't raised the buffer so high, in part because I thought a single instance on an i7 shouldn't have pops at 256 samples. I will do it next time just to see how it performs.

[exmatproton]

Hi Rob,

- I checked that patches weren't clipping.
- No clicking from the envelopes, I got this with all kind of patches, specially pads.
- Voice stealin, could be, but I was trying just holding a chord and getting the pops, before playing new notes.

Yeah, I cheked Live CPU meter, will check the system one next time I get this behaviour. Multicore option increases this, so I turned it off.

I didn't raised the buffer so high, in part because I thought a single instance on an i7 shouldn't have pops at 256 samples. I will do it next time just to see how it performs.

I am sure it is driver and/or usb related. Try to detach all usb devices, set your latency @ 512 with asio4all and attach your usb devices one by one. Sidenote: try this without an internet connection, e.g., by disabling your broadband/wifi adapters.

[~Pd~]

This may be relevant, since you went from an i5 (no hyperthreading) to an i7 (has hyperthreading) - I recently bought a Macbook Pro with an i7 processor, and was getting strange behaviour from U-he synths - occasional pops and dropouts - until I disabled hyperthreading.

I don't think hyperthreading is useful for realtime applications like audio - for one thing it seems to limit each virtual "core" to half the clock speed of the physical core it's running on. After disabling hyperthreading, I saw a noticeable performance boost in other audio apps too - for example Reaktor reports less CPU usage and the CPU load meter fluctuates less.

[exmatproton]

This may be relevant, since you went from an i5 (no hyperthreading) to an i7 (has hyperthreading) - I recently bought a Macbook Pro with an i7 processor, and was getting strange behaviour from U-he synths - occasional pops and dropouts - until I disabled hyperthreading.

I don't think hyperthreading is useful for realtime applications like audio - for one thing it seems to limit each virtual "core" to half the clock speed of the physical core it's running on. After disabling hyperthreading, I saw a noticeable performance boost in other audio apps too - for example Reaktor reports less CPU usage and the CPU load meter fluctuates less.

No. That should not be the problem. I am on a core-i7 4970K@4,6 GHz and i came from a i5 as well (4670K). The 4 "extra" pumped my 'realtime' music making really good. Using Diva at the highest settings (@3 ms latency) with aroun 46 voices (multiple instances) without any hassle.

[ghettosynth]

This may be relevant, since you went from an i5 (no hyperthreading) to an i7 (has hyperthreading) - I recently bought a Macbook Pro with an i7 processor, and was getting strange behaviour from U-he synths - occasional pops and dropouts - until I disabled hyperthreading.

I don't think hyperthreading is useful for realtime applications like audio - for one thing it seems to limit each virtual "core" to half the clock speed of the physical core it's running on. After disabling hyperthreading, I saw a noticeable performance boost in other audio apps too - for example Reaktor reports less CPU usage and the CPU load meter fluctuates less.

Hyperthreading has caused problems for me in the past as well. I am not skilled at all in multicore programming so I may say something that's nonsense here, but, the problem seems to come from demanding applications that treat logical cores like physical cores.

https://bitsum.com/pl_when_hyperthreading_hurts.php

I've noticed this largely with academic code where there is really no application level understanding of the difference, e.g., using any of the multicore frameworks for R. The application thinks that it has X real cores and if every core is doing CPU intensive work, there is little advantage and often a disadvantage to using hyperthreading.

Now, there could be a lot of other issues related to scheduling and OS implementation here that I'm not at all qualified to address. But I think that it's a good point, perhaps Uhe's guys will have some better insight.

[~Pd~]

https://bitsum.com/pl_when_hyperthreading_hurts.php

That's a terrific article, thanks! Have to think about this and do some testing...