Dual Xeon Optimization in Cubase?

I have read on the Steinberg site that Nuendo is optimized for dual processor workstation/servers.
Does Cubase 8 have this optimization as well? (discussion of it is not included in the features listed for Cubase)
I have a dual xeon e5-2697 64 gb workstation that im running Cubase on currently and even running asio guard I dont get the level of performance that I would expect for the level of processing power I have. On an average (not large) session mix with a reasonable number of plugins (but no instruments) i can easily hit 40% on the meter. Additionally I will get momentary peaks every once in a while when i reach this about 40% level.
I should point out that the machine is (of course) strictly a studio machine. It has been optimized for audio in the OS and Bios. It has no anti virus or other TSRs running on it and has no other things consuming resources.
This is why I’m interested in whether or not the multi processor optimization is specific to Nuendo.
Is anyone else seeing performance challenges on a high power workstation?
Any insights/information are greatly appreciated.
Thanks
Cheers
Ian

You are not alone my friend. I too had problems with my update to 8.0. I lost a lot of vst power. I mean a lot.

Now…before I get beat up too bad by the believers (I am one too…or I would not be here), I’ll list my specs.

This is a brand new build, with a fresh win 8 Install, straight to 8.05.

-latest generation Haswell 5960 (watercooled and clocking at 3.7)
-32 gigs of crucial ballistix ddr4 2400 ram
-all ssd drives
-$380 MSI motherboard
-2 Lynx Aurora 16s with pcie connection.

If I run at 128 buffer, I can hit 40% with 1 instance of Steinberg’s own Dark Planet. At 256, its more stable…but starts getting hard to play and sequence because of latency.

Now…here is the second part of the equation…just to go ahead and stop these replies before they start…I also have Sonar X3 on this computer (my old go to for audio work). It literally stomps THE GUTS OUTTA CUBASE WITH VST USE. I can run 10 instances of Dark Planet at 64 buffer and it does not even look like it affecting my CPU in any way. The CPU performance meters are not even at 5%. This is a Cubase problem…not a computer problem.

I am not here to bash Steinberg. I LOVE Cubase. Really…I used Sonar for 13 years before converting 2 years ago. However…something is definitely wrong with the code here. No way should I be getting the mediocre performance that I am. Come on Steiny…I’m rooting for you guys. Get the CPU/Asio problems sorted. They are real.

I also feel that I need to point out that this is a VST problem. When I record audio and track mics and live gear directly in, the performance is superb.

Yes, Sonar will toast Cubase in the VSTi use-case, but in another use-case Cubase will beat Sonar: live effects on the stereo buss, if you use a lot.

With Sonar, all live buss effects (and any “serial” chains of audio) must “fit” into a single core. The audio scheduling engine won’t break up a serial signal path across cores.

http://forum.cakewalk.com/FindPost/2656141

Cubase will balance the load, in this use-case.

So for VSTis, Sonar performs better. For lots of busses and sub-groups, fx channels, as well as the master buss (and with effects on all those), Cubase performs better.

And of course Reaper performs better than anything.

Watching, at last a window on the real world thing

I really don’t know what to say to this. I have just set up a new machine - see below. I have been amazed at how much more efficiently it works. I can track at 64 samples which has been great as a guitar player. In fact the performance is as good as it was at 256 samples on my previous machine.

The only weird thing is that u-he’s Diva can still max out the cpu - I can only guess that it does not have a multi-core operation.

Since I have to eventually bounce or freeze the VSTis, anyway, Sonar’s main advantage goes away. It then becomes all about the live buss effects (and Cubase wins).

It’s too bad that Sonar has tied one hand behind its back like this, as it’s a much higher performing audio engine, otherwise.

If Sonar had Cubase’s ability to schedule audio, for serial circuits, across cores and had an equivalent to ASIO Guard 2, it would probably even beat Reaper and be the king of the audio engine world.

A 64 sample buffer is a bit too low for some of the really CPU demanding soft synths. Even with your new CPU (especially if it’s not overclocked).

Actually, that CPU with 8 cores may perform worse than a Devil’s Canyon 4 core that’s overclocked for things like VSTis, if it can’t fit into a single core (even for Cubase).

Diva, at 64 buffer, probably wants to split between two cores (and can’t). Remember, those 8 cores are running pretty underclocked, per core, compared to other processors.

In that way, you may have taken a downgrade (per core) from your last CPU (especially if it was overclocked).

But, you’ll probably be able to run more real-time effects as long as each effect can fit on a single core.

That said, overclocking gives non-linear benefits for floating point (and therefore plugins). Since the best overclockers are currently the 4790k Devil’s Canyon, I think it might even beat your 8 core (and is almost 1/4 the cost). … Nah, yours will probably win. Those 4 extra cores.

Actually I tend to avoid overclocked cpu’s as I often keep my computer working for long periods of time and am wary of the heat generated by some such machines. Having said that I am more than happy with the performance so far. Running Diva in draft is no hardship as it still sounds better than any other Moog-alike.

Yeah, Diva sounds amazing.

Yup, you’d need watercooling, for sure. These new CPUs take heat a lot better than previous generations though.

I managed to get my 4970K stable at 4.6ghz, even under maxed loads, but it took a week of tweaking voltages.

At least the option to OC is always there if you need to stave off another upgrade, years from now. :slight_smile:

These new Intel CPUs are very DAW friendly. Exciting times.

http://www.hpcmusic.com/

A bit early for April 1st? :nerd: :open_mouth: :unamused:

:laughing: Wait, is this real? :open_mouth:

Wow. It’s either real or an extremely elaborate hoax, complete with youtube screencasts.

Yes, appears to be real.

https://www.youtube.com/watch?v=ztgvSt6VeXk

Super-high sample rates for super-low latency, which only works if there is a massive amount of CPU/GPU cycles to handle the workload fast enough, hence the need for HPC.

Couple that with hyper-accurate emulations, which will use those massive amounts of cycles.

The presentation was excessively stretched out, especially as it didn’t really show how cloud-based HPC could actually deliver on the latency, nor give any indication of how they farmed out the workload efficiently to all the available cores.

10 instances of DarkPlanet 256 buffers all busy playing - My CPU is nowhere near 40% as you can see…
DarkPlanetTest.jpg



For the sake of comparison with bigboi’s Sonar setup, what do you get for CPU use when using 64 sample buffers?

Okay just tried it - processor use looked a bit more bumpy with spikes into the 30s.
DarkPlanetTest64S.jpg

For your viewing pleasure…

http://youtu.be/OHXqPoayo6k

Also…Peppa…show us your f12 vst performance monitor? Maybe a video showing it on 64 buffer then playing with the vst performance meter showing? And use the very first patch that is loaded by default.

I just don’t understand how your system is running this at 64 buffer…I have a BEASTLY build and it has trouble. God I want this sorted out. I am not alone here either…others really are having issues as well. Read the YouTube video comments…one person right there.

dark plannet test.jpg
here’s 12 dark planets playing at 64 sample buffer with the sound you suggested.


see my sig for DAW, this is with my DAW NOT overclocked as my sig but in quiet mode at standard spec.


MC

Obviously Steinberg is having trouble system to system. Not sure why so many of us are having trouble while others are running great. What motherboard and video card are you running?

Also…put them in record ready and send me a pic. If you watch my video a couple of posts back you will get a better idea of where my problem is.

OK, I’ve watched your video and see the issue.

It would seem in record mode that record armed tracks all drop to a single thread for lowest latency, have you looked at the task manager to see if that the case?

What is the actual problem though when recording if the meter is in the red? I mean does it matter if there are a few clicks and pops if on playback everything is smooth?

I’m away from the studio now but will check tis out again tomorrow for you in record :slight_smile:


MC

Hi guys. Part of the problem talked about here is probably about the audio interface differences. (Of course this doesn’t explain Cubase vs Sonar performance that bigboi mentioned above)

I also switched to 5960x with shockingly very little difference. Using my Lynx AES-16e, I get dropouts, cpu spikes etc, in large projects. I switch the ASIO to my other card (which is a cheap USB 2.0 card) and it has no problems and still I can set it buffers even less without problems. (But it has one stereo out compared to 16 outs of the AES-16e)

I am using AES-16e with Aurora-16VT, but I will probably get an RME card, which are said to have optimized drivers and don’t bottleneck the CPU. (And I will have to find a specific cable to connect RME to the Aurora, because of the Yamaha/Tascam pin difference)

Still there is the weird case of Cubase getting heavier with every new version. If I were Yamaha/Steinberg, I would hire those Reaper guys and have them re-write the exact Cubase from scratch and optimize it as far as they could.