Here’s graphic example of the difference AG makes.
Same project , same amount of ‘Crimson’ plugins . One is Loaded directly in Cubase the other is loaded locally via the AG plugin.
Look at the difference in the performance meter.
M
Here’s graphic example of the difference AG makes.
Same project , same amount of ‘Crimson’ plugins . One is Loaded directly in Cubase the other is loaded locally via the AG plugin.
Look at the difference in the performance meter.
M
Wow, that looks impressive!
Wow. I use AG on a computer network. How does it make such a difference locally?
Thanks for sharing!
Ps: are on Mac intel, silicon or win?
Very interesting. I don’t know the reason, but maybe with multiprocessing activated the multiprocessing is not used in a balanced way and AG helps to achieve it. ¿?
Important to note that the processing of those plugs is still happening outside of Cubase and so is not shown in perf meter…the load spreading is much improved so you can definitely increase performance, but probably not as much as this comparison might suggest.
If you look at the performance meter Cubase would be in the red with a few more instances. If you look at the windows task manager it would show 20% system usage. So AG is able to use the available 80% that Cubase cant so yes it can increase performance by a huge margin.
I have a DAW bench test set up with Vin and we had channel strip plugins loaded on mix buses and added plugins until Cubase hit the red. I managed an extra 8 of these plugins on my machine. When loading the exytra plugin using AG I gave up after adding 70!!! that’s a massive difference.
M
each audio gridder instance has 3 plugins included.
Look at the Cubase meter it’snot even on half. With just an extra 8 of those plugins loaded natively in Cubase it hit the red and started dropping out.
I gave up at 70 as I still had lots of system resources left.
M
Fantastic!
I’ll dig deeper myself. Thanks for sharing!
Isn’t this effectively quite similar to sending a MIDI track from Cubase to external hardware?
So the Cubase process no longer has anything to do with generating the audio from MIDI. And therefore unsurprising that the entire load that is handled by AudioGridder disappears from the Cubase Performance Meter.
I have a related question:
Does the use of AudioGridder imply that all audio mixdowns have to be real time, and one can no longer use offline rendering?
I think there is more to it than that. The Steinberg performance meter shows what the available resources are, so if the total workload for the computer is the same we would expect the meter to show the same.
Yes…as I said the process is still being done outside of Cubase so perf meter isn’t telling the whole picture but as you can see from the screencap showing core loading, the load spreading is very even with AG so it can still give a lot of performance boost.
Does the use of AudioGridder imply that all audio mixdowns have to be real time, and one can no longer use offline rendering?
I think offline is still possible.
So why wouldn’t Steinberg rearchitect plugins to work in the same way that AG works.
It’s not magic, just a different approach?
I don’t think that’s right at all.
The Cubase performance meter is much more indicative of latency than CPU consumption. Arguably it might be somewhat indicative of the CPU load of the busiest single thread. But definitely not overall CPU consumption across all cores.
And that’s easy to confirm by comparing the Cubase performance meter with the system wide performance meter.
AudioGridder merely takes the process outside of Cubase and therefore allows more parallel threading than a single track of Cubase. This becomes a very noticeable benefit, when the Cubase project has some track(s) with very high CPU demands and others with much lower CPU demands.
If you have 100 tracks with equally high CPU demands I would expect AudioGridder to add little value.
But if you have only 1 track with high CPU demands, then I expect AudioGridder to make a significant difference in spreading out the load and therefore lowering latency.
I was hoping for a definite yes or no - but I guess I’ll have to install AudioGridder and test for myself.
For example this post seems to imply that there can be issues, depending on what is offloaded to AudioGridder:
My current assumption is, that offline rendering is possible, but can cause grief - and that it depends on the exact combination of CPU consumption and resulting latencies of the various plugins inside Cubase tracks and externalized to AudioGridder.
So offline rendering may be fine - until it isn’t. And then there will be forum posts by people blaming Cubase for their renders not being the same as when they listen realtime.
Off line processing is exactly the same using AG. It’s as reluable as using all plugins natively.
M
It’s arguably the exact opposite. If you have 1 track with high CPU demands then that equals 1 heavy thread. It can’t just be spread out over multiple cores if the processing is sequential, meaning that moving it to AG makes no difference. (or maybe you meant that spreading out the load was between AG and Cubase?)
But the problem indeed seems to be how Steinberg’s audio engine is spreading load over cores. When you wrote it is unsurprising that moving the load from Cubase to AG what I meant to say was that IF the audio engine was working better then it would not be surprising, because the load would be effectively the same but in different places, and therefore a lot of people are surprised when they see this because they assume there would be little difference.
I think that was my line of reasoning basically.
Perhaps we’re saying the same thing.
PS: I also think people consider the performance meter to show basically the same thing as the Windows Task Manager, and it doesn’t. That would be another reason people have “wrong” expectations.
Long term Cubase user here. (Some might say long suffering lol)
I’m running Cubase 13 pro on an underpowered machine- a Surface pro 4. The pro 4 has always suffered from issues with the audio subsystem which affect some software more than others. Needless to say Cubase is one of the worst performers.
It’s ok for my needs running recorded audio tracks but falls to bits under the load of 2 or three instruments.
Running those instruments in an audiogridder server seems to free up resources for the Cubase audio engine to run more smoothly.
It doesn’t seem to make a lot of sense- you have an additional server and a client running on the machine and all the information flow between these applications and Cubase but the difference for me at least is chalk and cheese- I’ll get a couple of more years out of my old machine.
here’s the screen grab of the project maxed out without using AG. You can see quite clearly cubase is at its max with only 22% system use.
When you compare it to the AG version you can see AG is using that extra 78% sat unused hence me stopping when loading 70 extra plugins with still plenty in reserve. So I’d say it DOES increase by a large margin.
M
Black magic and witchcraft - like all software!
Thanks to your excellent advice I did install Audio Gridder a couple of days ago and found it to improve performance indeed. Very glad !!
I was curious about rendering so just tested it, it only works in real-time on my setup, how do you do offline rendering? If I missed something please let me know.
Tried 2 instruments through AG instrument and one instance of Mixbox on an AG FX on the master.
Ah… you’re right. Something’s changed with Cubase or AG over the last few years then as Cubase 11 and AG 1.0 rendered as normal when I used a slave machine back then.
TBH I’ve not had the need to use AG except for testing since I’ve had my 9950x so didn’t realise things had changed.
Also i use ecternal hardware so I quite often have to render in real time anyway.
M
No worries.
I MUCH PREFER having to render in realtime but have extra performance and a stable system than the other way around.
I hope Steinberg makes improvements on performance and stability. I just had another crash disabling a track, and no dump file produced.