C8 performs much worse than C7.5 with VE Pro

I have a project in C7.5 using 4 instances of VE Pro, 16 stereo outs each, 9 group channels, 5 FX channels and 10 active audio tracks. RME 9652 HDSP at 256 samples.

That’s it. Fairly light-moderate sized project. Nothing heavy. Cubase 7.5 avg. at 60% and real-time higher than that. Almost no crackles.

C8 flat out peaks at 100% (both avg and real time) at many places and almost doesn’t even play. I thought maybe I’d move to C8 for render in place, but the initial tests are disappointing.

What’s going on? Why the huge discrepancy? :frowning:

On some investigation as to where the CPU power is going:

  • Removed all extra outputs in every VE Pro, making each instance a stereo only -> Doesn’t make any big difference. Maybe 2% for each instance.
  • Removed 8 group channels - not much difference. Again, 2%.
  • Removed ALL VE Pro tracks - all 64 of them. Hardly any difference at all. I can’t even tell. This is good.
    - Removed Ozone 7 - WHOPPING 25% drop!! What wizardry is this plugin doing that it sucks 25% of my CPU?
  • Removed other Izotope plugins - and no surprise - each one of them was contributing 5-7% to the CPU usage.

Turns out the Maximizer IRC 4 setting with True Peak Limiting, is quite a kick-ass limiter in the sense it kicks your CPU’s ass.

Putting the individual plugins vs. having them all in the Ozone 7 plugin seems to make no difference. It’s just that each individual Ozone 7 plugin consumes CPU that all adds up to ~25%.

That’s why I master my stuff in an extra DAW session or sometimes in the standalone version of Ozone, which is also capable of loading other plugins. This stuff is just too heavy un cpu to use it in the mix session.

Yes I’ve learned to use the constrain delay compensation feature to kill ozone until rendering stage.


Now the question is - why the performance gap between v7.5 and v8? Shouldn’t it atleast be marginally better in v8?

Are you using ASIO guard? Your numbers seem really high in all cases.

Also, are you sure you have multi processing disabled on Kontakt and enabled on VEPro?

Yes it’s in normal mode in C7 and the default setting in C8.

I didn’t know you were supposed to do that for kontakt. Let me try that too - thanks for the tip!

I didn’t know about this either! Do you mind sharing where you got this information? Either way I’m going to try it on my current session and see what happens.

izotope RX brings my cubase to its knees…

Something in their code gives me loads of cpu peaks and performance issues…

I also talked to izotope about it , but well, i dont think their supporters had great insight into the code…

Either way, izotope really needs to do some code optimization…
(Example->RX decrackler A and B settings work with huge cpu useage… Settings C and D are impossible to run… -> there is enough cpu power, but nonstop realtime peaks make the latter settings (c&d) impossible to use with my system)

(To give them the benefit of the doubt, i haven’t tested it in other DAWs (RX) , but izotope is a real cubase killer on my system)

I didnt upgrade my izotope package because izotope told me that there wasn’t any code optimization done in the new updates…

Soooo… Idk… Hoping that someone, whoever… will fix their code…

This is off topic, but is there a reason why you dont run RX in stand alone to process your audio, then bring it back into Cubase? I’d personally much rather have something sound amazing and use heavy CPU (which I render out) than have something optimized for real time use but sound bad.

As a test I opened up my current Cubase 8 project, which has moments of AISO meter peaking but still plays back ok in C8, in C7.5 and it wouldn’t even playback. I am running a latency of 1024 and in Cubase 8 I have AISO Guard2 on high (C7 also has AISO guard but no options) and on this project, so perhaps this is the difference. But basically I’m actually seeing better CPU performance in Cubase 8 vs Cubase 7.5. The only way to get C7.5 to playback the song was to crank the latency to 2048, and even then it had peaks.

You mean with the new connect system?

The new connect system is very nice actually, because with the older versions of RX, and using the offline processing in cubase, cubase didnt compensate the latency visually… (Sound and gfx didnt match up) - this is a cubase issue that one can also observe even with cubase stock plugins…

RX connect fixed this issue by externalizing the process…

Even when stopped and no processing done:
decrackle:
low setting = no cpu useage visible
Medium= massive peaks - not useable
High= yeah, right…


While playing:
Denoiser:
A&B settings -> no probs… Same cpu as maybe a halion instance
C&D settings -> super peaks and totally unuseable

-> this has nothing to do with cpu power but with code optimization and how the latter settings utilize certain cpus…
Like i said, there is enough cpu for these tasks… If the code or cubase or whatever wouldnt bug out…

Yes, I have to use RX in connect mode now with latter settings(basically standalone) … But… There are limitations in not being able to process many snippets of audio over time in real time. Using it in offline Connect mode is nice if you only have a limited number of files to process or one big file… Realtime use with the “better” settings should be very achievable if the izotope or cubase code gets tweaked to work better with eachother…

I am considering buying iZotope Ozone 7 for mastering use in wavelab E9, but after reading these posts, I am confused about it. Does anyone use it without problems?