We use Nuendo to digitize/restore historical recordings on behalf of customers.
As a rule, we only have one track with thousands of audio events and process everything in the direct offline process (Waves, iZotope, Voxengo, RxDevine Pro …)
Ultimately, I want to speed up the rendering speed in the DOP. My question is whether the speed improves noticeably if I switch from 8 cores to 16 or 24.
Even if you render several events at the same time in the DOP, Nuendo processes them one after the other and not in parallel. So I’m worried that multiple cores won’t increase rendering speed. Because the CPU speed is crucial instead.
To do this, you would need to know how Nuendo distributes multiple parallel rendering processes across the CPUs.
I currently have an i9 with 8 processors and 3.6 GHz. However, many new computers are offered with 16 and 24 CPUs - but only 3.0 GHz. I even fear that this will make things worse.
Hey, no experience, sounds difficult from how you are explaining it the only way to circumvent this bottleneck is to use Nuendo differently and use RX as well.
DOP will use 1 core per process/event. At least that is my experience on Mac. What does your current system do? If that uses 1 core you know the (unfortunate) answer.
I understand this is not the answer you are looking for though…
Could you think of a workflow that does batch work in the applications like RX that do use multiple cores (i can confirm this)?
RX allows you to use plugins like Revive and Voxengo (depending on what you have).
Thanks.
That’s what I feared. On the PC, Nuendo only seems to use one core at a time.
Our solution is currently several PCs. If the PC has been rendering for several hours, I simply go to the next PC and continue working there for another customer. I’ve been used to that for years. We often have batch processes carried out on Nuendo files via WaveLab on an additional computer. WaveLab processes batch processes cleanly in parallel on multiple cores.
I was just tempted to buy a new PC. But I still have an old PC with only 4 cores, also 3.6GHz - it renders exactly the same speed.
So the problem depends on the clock speed. A new PC would have a turbo clock speed that would deliver just 16% faster speed. That’s not a real win.
Well the research investment is in multicore (for several reasons: AI cores, Privacy cores, low energy cores) all in the same CPU. While this is great for the average consumer… it is not for power users who want high Speeds and enough cores. This sounds like a case of a power user with a very typical use case and that is a small market.
Though I can’t imagine the advantage is only 16% it is so on paper. The other advantage of newer CPUs is higher bandwidth connections to memory (RAM and disk). Looking at apple they have low cpu speeds but everything is lightning fast because it is integrated.
Sure, but it also depends on what even subsets of DAW users do. I’ll have to look at how my DOP behaves and if it’s the same for all plugins, perhaps it is, but I know for sure that as wel build out our projects the distribution of work on cores changes depending on what we do - i.e. long signal chain versus very parallel.
I suspect clock speeds will continue to increase as much as they can though and given heat and power constraints it seems they (AMD/intel) are not really holding anything back.
???
3.7 * 1.16 = 4.3GHz. Modern CPUs boost far higher than that. Closer to 50-60% higher boost.
Yes. I am currently using the Intel Core i9-9900K. And I was thinking about getting the currently cheap i9-13900K.
The turbo burst rate is 5.0 (9900) to 5.8 GHz (13900). That’s 16% more rendering speed.
Rendering is extremely slow, especially with dxRevive Pro and iZotope Dialog Isolate.
A new computer would make sense on a small order if rendering a dialog only took 8 or 10 minutes instead of 20 minutes.
I’ll soon have to batch-clean up 400 hours of Dialog with dxRevive Pro and iZotope Dialog Isolate.
That was my idea of buying a new PC. And put the old PC in the corner and then use it to run such batch processes.
But a reduction from 20 to about 17 minutes doesn’t justify a new computer. That doesn’t make sense from a commercial perspective.
Sorry i was just quoting, didn’t do any math myself
I think a new computer only makes sense if you can run these cpu heavy processes in batches. Have you looked into that?
Also there is ToddAO who sell similar products for batch noise reduction. Might make sense to test that!
I just went with the number I saw which was 3.7, not 5.0. That’s why I got confused.
Generally though remember that there’s more to performance than clock speed. Really what you want to look at is instructions-per-clock, which can increase by turning up the frequency but also by changing how the computer works. Things like cache performance and branch prediction and AVX512 can increase performance at the same clock speed.
So just because the clock only went up 16% doesn’t mean performance won’t increase quite a bit more.
Yes, batch operation whenever possible (WaveLab). I’ll take a look at Todd-Ao.
However, the values of the plugins must be adjusted individually for each event. These are then pushed into the Direct Offline Process. Then the next event comes with adjustments.
That’s why we have several computers. (Remote operation).
These are usually recordings of government or community meetings. People sat so far away from the microphone that the conversation was barely visible in the noise. These are lousy recordings from the 1970s. Sometimes entire syllables are swallowed.
The new version of dxRevive Pro 1.2.2 is so awesome that it restores voices and adds syllables that are no longer heard.
But when restoring tape recordings or vinyl, the processes are much faster, so a new PC for local work with at least 25% more power could make a difference.
So I guess I’ll wait until we get the larger amount of dialogue (in the next 3 months). Then a new computer will probably come along. I hope that it will be more than just the 16%.
