V18.01 possibe bug - modifiers very slooooowww at only 89% CPU?

[Dpoirier]

I haven't been able to nail down the exact scenario under which it can be reproduced reliably (will try to do that tomorrow), but a new preset I'm working on is rather complex, and has a lot of modifiers. I was baffled by the fact that the modifiers (controlled by three expression pedals) were extremely slow - almost like I had a very long damp value (which in fact I did not). So of course, I checked CPU level - barely 89%, lots of headroom. I then assumed a corrupt preset, so I rebuilt it from scratch (and a wobly memory). But to my surprise, as soon as I had all the blocks in there, boom, sluggish modifiers, and still well under 90% cpu.

So I remove a compressor block, and bingo!, the modifiers came back to life. I restored the saved preset (with the compressor) just to see if removing anything else also "fixed" it, and it did.

So either the displayed CPU level is off somehow, or modifiers attached to an expression pedal become very sluggish even at well under 90% cpu.

Details: Axe-FX II XL, fw 18.01; preset in question had two amp blocks, two cab blocks, two pitch blocks, one drive block, delay, reverb, one filter, one EQ, and 3 or 4 mixer / volume blocks. Oh, and the amp blocks both had a little bit of out compression (which does use up more CPU). In fact, looking at this laundry list, I'm leaning towards v18.01 displaying the incorrect cpu usage, 'cause that's a lot of stuff for less than 90%. This would explain the sluggish modifiers, though I didn't get any audio crackles (which are normally the first sign of excess cpu?)

I'll work out more details tomorrow, but my current guess is that this can be reproduced as follows: build any preset with about 90% cpu, and watch the slow behavior of a modifier attached to, say, a volume block when you control it with an expression pedal.

If nobody can reproduce it, and after I explore more, I can post a preset.

 

[mbrown3]

I haven't been able to nail down the exact scenario under which it can be reproduced reliably (will try to do that tomorrow), but a new preset I'm working on is rather complex, and has a lot of modifiers. I was baffled by the fact that the modifiers (controlled by three expression pedals) were extremely slow - almost like I had a very long damp value (which in fact I did not). So of course, I checked CPU level - barely 89%, lots of headroom. I then assumed a corrupt preset, so I rebuilt it from scratch (and a wobly memory). But to my surprise, as soon as I had all the blocks in there, boom, sluggish modifiers, and still well under 90% cpu.

So I remove a compressor block, and bingo!, the modifiers came back to life. I restored the saved preset (with the compressor) just to see if removing anything else also "fixed" it, and it did.

So either the displayed CPU level is off somehow, or modifiers attached to an expression pedal become very sluggish even at well under 90% cpu.

Details: Axe-FX II XL, fw 18.01; preset in question had two amp blocks, two cab blocks, two pitch blocks, one drive block, delay, reverb, one filter, one EQ, and 3 or 4 mixer / volume blocks. Oh, and the amp blocks both had a little bit of out compression (which does use up more CPU). In fact, looking at this laundry list, I'm leaning towards v18.01 displaying the incorrect cpu usage, 'cause that's a lot of stuff for less than 90%. This would explain the sluggish modifiers, though I didn't get any audio crackles (which are normally the first sign of excess cpu?)

I'll work out more details tomorrow, but my current guess is that this can be reproduced as follows: build any preset with about 90% cpu, and watch the slow behavior of a modifier attached to, say, a volume block when you control it with an expression pedal.

If nobody can reproduce it, and after I explore more, I can post a preset.

89% is not "well below" 90%...and 89% is still pretty high. Would 89% cause issues? I wouldn't think 89% would be any different than 90%, which is where others have said they generally start having problems.

 

[chris]

I'd just post the preset

 

[GM Arts]

I found the same thing with a patch with around 94% CPU which is pushing my luck, so just switched to low-res cab and problem solved.

 

[Klausi]

Same for me. Hires Stereo and reverb high quality, 2 pedals. Can't remember all details. But i think CPU is now 1 or 2 % higher than in beta 7. I don't use the new rotary and amps are running in the 2nd CPU.

 

[Patzag]

I pushed a preset to 92% just to check and can't reproduce the Modifier sluggishness.

 

[k2green]

I can attest to similar CPU "issues" or changes with new firmware. Everything comes with a cost. I have a few presets that I had to tweak things. I don't see it as a bug....more of reality when working with computers. Hardware is developed - it is static. Firmware is fluid.....improvements are made until eventually the cost is too high and new hardware is needed to keep up. In x years, the firmware will go so far that axe 3 is needed, etc. If anything, the new firmware will force me to be smarter and more efficient as I develop new presets.

 

[Dpoirier]

I pushed a preset to 92% just to check and can't reproduce the Modifier sluggishness.

Hmm, that's interesting. Well, in fact, I would expect any preset to behave properly up to 93% plus (and at 94%, you're on the edge and start getting weird behaviour). At least, in the past, that was how it worked.

That's why I said 89% was well below 90% - a 1% delta makes a world of difference in the Axe when you're close to the limit... and 5% below the 94% usage commonly known as the threshold "should" (normally) be totally safe.

I understand that the new algorithms use more cpu, and that's normal (and embraced). But I have a feeling that either the new f/w is displaying cpu usage that is optimistic / under evaluated, or that the modifiers react (become sluggish) well before the usual 94% threshold. That is what I'm wondering about. Has the threshold now been lowered with v18.01?

Now, given that someone was able to run at 92% with modifiers behaving properly, my theory / suspicion just took a nose dive. I have to explore some more and post a preset (not now, I'm at work).

Thanks for looking in and responding with comments, y'all!

 

[Zwiebelchen]

Remember that it's not only about the raw cpu usage, but also the routing complexity.

I always recommend to stay below 85% cpu usage. Even if you don't have noticable audio artifacts until like 92%-94%, everything becomes slower and less responsive. This is because the UI and externals get processed after the sound. I like my UI to be fast and responsive and I don't even hear the difference between a low res and a hires cab anyway.

Plus, it's good to have some CPU headroom for the USB overhead.

 

[Dpoirier]

Remember that it's not only about the raw cpu usage, but also the routing complexity.

I always recommend to stay below 85% cpu usage. Even if you don't have noticable audio artifacts until like 92%-94%, everything becomes slower and less responsive. This is because the UI and externals get processed after the sound. I like my UI to be fast and responsive and I don't even hear the difference between a low res and a hires cab anyway.

Plus, it's good to have some CPU headroom for the USB overhead.

Thanks for that point of view but (a) my top 5 presets are all running at 90%+ cpu levels, and (b) I've never seen this issue before, not unless I was at 94% plus. Like I've already said, I used to always be able to rely on proper performance up to 93% cpu (beyond that, things can get dicey). But never seen an issue at 93%.

In addition, I'm not talking about a moderate sluggishness here, with this preset at 89%, the modifiers are unusable - I get like 1 to 2.5 seconds of lag!

I don't think your comments apply to the behavior I'm seeing. I need to explore more and upload the preset.

 

[Rock Chalk]

Yep, I have the same thing going on here. I have a lot of patches that hover around the 89 to 91 percent mark, and am getting intermittent issues such as volume pedal lag. Yes, I know I could get rid of it by choosing normal quality reverb, pre, or cabs, but I'm a glutton for the high quality stuff! Plus it only happens now and then and I'm not playing out so no biggie.

I will say this though.....I would step up to a Axe III in a heartbeat if it had the power to do all of this without hiccup, even if it sounded the same.....I think 18.1 is that good!

 

[Mike Snider]

I lost bidirectional communication with the mfc on certain presets after the update- cpu creep, drove me bonkers for a few days until i figured it out. Good point someone mentioned above, high res on the reverbs and cabinets can chew up a ton of cpu and can be a good place to reserve cpu without modding a preset too far.
1% can be the difference between stability and unusual behavior....without the usual symptoms of clipping or loss of signal.
My layout was fairly complex too, I got wobbly around 91%. I dont track cpu usage much because it has never been an issue for me in the past but clearly the gain in clarity has it's cost (no complaints from me).

lag of 1-2.5 does seem extreme given 89%

 

[Patzag]

Hmm, that's interesting. Well, in fact, I would expect any preset to behave properly up to 93% plus (and at 94%, you're on the edge and start getting weird behaviour). At least, in the past, that was how it worked.

That's why I said 89% was well below 90% - a 1% delta makes a world of difference in the Axe when you're close to the limit... and 5% below the 94% usage commonly known as the threshold "should" (normally) be totally safe.

I understand that the new algorithms use more cpu, and that's normal (and embraced). But I have a feeling that either the new f/w is displaying cpu usage that is optimistic / under evaluated, or that the modifiers react (become sluggish) well before the usual 94% threshold. That is what I'm wondering about. Has the threshold now been lowered with v18.01?

Now, given that someone was able to run at 92% with modifiers behaving properly, my theory / suspicion just took a nose dive. I have to explore some more and post a preset (not now, I'm at work).

Thanks for looking in and responding with comments, y'all!

Sorry I don't have time to dive further into this but take a look at whether you're using Axe Edit AND a DAW at the same time. This sometimes strains the CPU a bit more.
But I've never seen a 2.5 second lag on anything in the Axe. I would suggest that your preset is corrupted. You might want to rebuild it and try again.

 

[AlbertA]

It also depends on what you are controlling with the modifier.

 

[Dpoirier]

Sorry I don't have time to dive further into this but take a look at whether you're using Axe Edit AND a DAW at the same time. This sometimes strains the CPU a bit more.
But I've never seen a 2.5 second lag on anything in the Axe. I would suggest that your preset is corrupted. You might want to rebuild it and try again.

That was my first guess as well, and I did rebuild it from scratch (as mentioned in the very first post in this thread). But as soon as I hit the 89% mark on the cpu, the lag on modifiers returned. I don't think it's a corrupt preset anymore.

Oh, and if I typed 1 to 2.5 second lag, that was a typo. I meant 1 to 1.5 second (and I kid you not). That is more lag than I've seen on anything ever.

By the way (because someone asked), the modifiers that were very slow were on two volume blocks (and I replaced one with a mixer block, but same issue). Another modifier (on delay in level) did not seem to be affected (in the same preset. at 89% cpu).

 

[Patzag]

That was my first guess as well, and I did rebuild it from scratch (as mentioned in the very first post in this thread). But as soon as I hit the 89% mark on the cpu, the lag on modifiers returned. I don't think it's a corrupt preset anymore.

Oh, and if I typed 1 to 2.5 second lag, that was a typo. I meant 1 to 1.5 second (and I kid you not). That is more lag than I've seen on anything ever.

By the way (because someone asked), the modifiers that were very slow were on two volume blocks (and I replaced one with a mixer block, but same issue). Another modifier (on delay in level) did not seem to be affected (in the same preset. at 89% cpu).

Can you load the preset. I'll load it into my machine and see what happens.
- P

 

[FractalAudio]

It's simply a question of resources. Audio is always the top priority. Expression pedals are a lower priority and if CPU usage is high they may not get serviced right away. When the office is busy you have to take a number and wait.

 

[Dpoirier]

It's simply a question of resources. Audio is always the top priority. Expression pedals are a lower priority and if CPU usage is high they may not get serviced right away. When the office is busy you have to take a number and wait.

Thanks for the clarification, Cliff. Good t6o know it's not a bug.

I guess the "lesson learned" here is that the old 94% cpu threshold I used to watch out for is no longer a valid stake in the sand, and things may get a little slippery well before reaching that level. I need to go on a cpu diet! V18.01 is well worth the sacrifice.

 

[vinnyburns]

It's simply a question of resources. Audio is always the top priority. Expression pedals are a lower priority and if CPU usage is high they may not get serviced right away. When the office is busy you have to take a number and wait.

Could this lead to problems when quickly scrolling through presets? Maybe an expression pedal controlling a VOL block can be ignored while scrolling and the preset comes up silent? Had this last year a lot of times.
I have since scrapped everything and started again from a clean slate and have not had the problem since. The presets are a little simpler now too so I would imagine they are not taxing the system as much. Just want to understand what could have caused this last year so I can avoid it in the future.
Happy camper now though.