ASIO Guard and Vienna Ensemble

[paaltio]

@JMCecil

Of course it’s a convenience thing, but you can change the buffers on the VEP instance.

I could also render any session on a Motorola 68000 as long as I don’t mind waiting, but I’m pretty big on convenience.

[keyman_sam]

@noiseboyuk

@keyman_sam

Does this problem exist in 7.5 as well?

i.e. with regards to VE Pro, which is better - 8 or 7.5? I’m trying to decide which one to pick.

What trailerman said. It might be true that it was also problematic with ASIOguard 1 in C7, but let’s be honest no-one used it anyway. The issue is that general CPU use increased in C8, AG2 has compensated for most people but not for VE Pro users for these much discussed reasons. So if VE Pro efficiency is your only consideration, then C7.5 is the better bet.

If Steinberg / VSL really have given up on AG2 and VE Pro ever working (and yes, the stony-faced silence isn’t encouraging, is it?) then one might conclude that the only hope is to make Cubase more efficient in general, regardless of AG2. Since they seem to be ignoring all the public clamour for such things these days, I wouldn’t bet the farm on that happening either.

Thanks.

So, if I disable AG2 altogether in C8 is it comparable to C7.5 or is it much more heavier on the CPU that I’m better off with C7.5 without AG?

Also disabling AG2 for VE Pro tracks → won’t that mess with the latency compensation?

The only thing I REALLY need from C8 is bounce in place. Everything else I’m covered in C7.5. Or if there’s a workaround in C7.5 I’ll stick with that (specifically bouncing midi parts to audio and import it in a track in one shot).

[misohoza]

Hi keyman_sam.

I haven’t noticed any sync problems when disabling asio guard for certain plugins only.

Why you don’t get a trial(s) and see yourself which scenario works best for you?

[keyman_sam]

@misohoza

Hi keyman_sam.

I haven’t noticed any sync problems when disabling asio guard for certain plugins only.

Why you don’t get a trial(s) and see yourself which scenario works best for you?

Thanks - I already have 7.5 and 8 full versions. I wanna get as much feedback as possible before I dive in to one or the other. I’ve used C7.5 briefly and I love it. But as expected, performance wise it is behind Reaper. So, 8.x is supposed to be better in terms of performance but only with AG2.

BTW do you have/use VE Pro in your setup? Or are you talking about all plugins in general excluding VE Pro?

[Trailerman]

Best thing would be to test both yourself. Different plugin combinations will probably yield different results.

For me personally, C7 was better than C8, because I use a LOT of VE Pro instances, and after I’d disabled AG2 on VE Pro, C8 was less efficient than C7. Hence I still use C7/Nuendo 6.5 and will continue doing so until the AG2 audio muting issue is resolved.

Jules

[SteveHelstrip]

Just to clarify, I have ASIO Guard turned on for all plug-ins except Ensemble. I have no problems whatsoever. There is more latency when playing Ensemble instruments (since ASIO guard is not enabled), but this is by no means a showstopper in any sense.

[keyman_sam]

@Steve Helstrip

Just to clarify, I have ASIO Guard turned on for all plug-ins except Ensemble. I have no problems whatsoever. There is more latency when playing Ensemble instruments (since ASIO guard is not enabled), but this is by no means a showstopper in any sense.

THIS is what I’d love to hear more of. The more latency part → is it noticeable? compared to 7.5?

[paaltio]

@keyman_sam

@Steve Helstrip

Just to clarify, I have ASIO Guard turned on for all plug-ins except Ensemble. I have no problems whatsoever. There is more latency when playing Ensemble instruments (since ASIO guard is not enabled), but this is by no means a showstopper in any sense.

THIS is what I’d love to hear more of. The more latency part → is it noticeable? compared to 7.5?

What some people have observed is that you do need a higher buffer without ASIO Guard in C8 to run a maxed out C7 session, and the same thing C6 vs. C7. That seemed to be the result in my benchmarking as well. That’s the extent of the difference.

It seems fine to go towards more robust prerendering rather than trying to optimize every last ounce out of realtime processing for basically non-realtime tracks, so I don’t fault Steinberg in that. But the prerendering has to work if that’s the emphasis, and with VE Pro being a massive part of today’s composer workflows, having to turn off ASIO Guard means it doesn’t. The end result is a performance regression through and through.

Really hoping Steinberg and VSL sort this out soon. There’s no reason we should be choosing between multi-computer setups or four digit buffer sizes in 2016. Not with the amount of CPU and RAM headroom available in modern systems.

[SteveHelstrip]

@paaltio

@keyman_sam

@Steve Helstrip

Just to clarify, I have ASIO Guard turned on for all plug-ins except Ensemble. I have no problems whatsoever. There is more latency when playing Ensemble instruments (since ASIO guard is not enabled), but this is by no means a showstopper in any sense.

THIS is what I’d love to hear more of. The more latency part → is it noticeable? compared to 7.5?

What some people have observed is that you do need a higher buffer without ASIO Guard in C8 to run a maxed out C7 session, and the same thing C6 vs. C7. That seemed to be the result in my benchmarking as well. That’s the extent of the difference.

It seems fine to go towards more robust prerendering rather than trying to optimize every last ounce out of realtime processing for basically non-realtime tracks, so I don’t fault Steinberg in that. But the prerendering has to work if that’s the emphasis, and with VE Pro being a massive part of today’s composer workflows, having to turn off ASIO Guard means it doesn’t. The end result is a performance regression through and through.

Really hoping Steinberg and VSL sort this out soon. There’s no reason we should be choosing between multi-computer setups or four digit buffer sizes in 2016. Not with the amount of CPU and RAM headroom available in modern systems.

I’m running at 128 samples buffer here, mostly. If the session gets really heavy, I notch up to 256.

[paaltio]

@Steve Helstrip

@paaltio

What some people have observed is that you do need a higher buffer without ASIO Guard in C8 to run a maxed out C7 session, and the same thing C6 vs. C7. That seemed to be the result in my benchmarking as well. That’s the extent of the difference.

It seems fine to go towards more robust prerendering rather than trying to optimize every last ounce out of realtime processing for basically non-realtime tracks, so I don’t fault Steinberg in that. But the prerendering has to work if that’s the emphasis, and with VE Pro being a massive part of today’s composer workflows, having to turn off ASIO Guard means it doesn’t. The end result is a performance regression through and through.

Really hoping Steinberg and VSL sort this out soon. There’s no reason we should be choosing between multi-computer setups or four digit buffer sizes in 2016. Not with the amount of CPU and RAM headroom available in modern systems.

I’m running at 128 samples buffer here, mostly. If the session gets really heavy, I notch up to 256.

VE Pro buffers have to be counted as well.

Not that it matters though, completely pointless to compare arbitrary buffer sizes without a standardized session like DAWbench etc.

[aptmusic9]

Hopefully this isn’t hijacking the thread, but has anyone tried using VEP on a slave computer and transferring audio using a physical sound card output (instead of ethernet) to try and reduce CPU load and run at lower latencies? Perhaps a single stereo pair via SPDIF or AES?

[Trailerman]

Hopefully this isn’t hijacking the thread, but has anyone tried using VEP on a slave computer and transferring audio using a physical sound card output (instead of ethernet) to try and reduce CPU load and run at lower latencies? Perhaps a single stereo pair via SPDIF or AES?

I understand why you’re asking this, but that would render the idea of VE Pro on a slave server somewhat pointless. You may as well use VStack/System LInk, or Plogue Bidule, or a second instance of your DAW on the slave computer, and use Midi Over Lan (or a hardware interface) and ditch VE Pro altogether.

Much of the value of VE Pro on slave machines is it’s ability to transfer large track-counts at low latency (relatively) back to the host via ethernet, without having to mix down on the slave system or use additional hardware. I’ve worked with hardware audiolinks between systems in the past and this would be a massive step backwards IMHO.

[aptmusic9]

@Trailerman

Hopefully this isn’t hijacking the thread, but has anyone tried using VEP on a slave computer and transferring audio using a physical sound card output (instead of ethernet) to try and reduce CPU load and run at lower latencies? Perhaps a single stereo pair via SPDIF or AES?

I understand why you’re asking this, but that would render the idea of VE Pro on a slave server somewhat pointless. You may as well use VStack/System LInk, or Plogue Bidule, or a second instance of your DAW on the slave computer, and use Midi Over Lan (or a hardware interface) and ditch VE Pro altogether.

Much of the value of VE Pro on slave machines is it’s ability to transfer large track-counts at low latency (relatively) back to the host via ethernet, without having to mix down on the slave system or use additional hardware. I’ve worked with hardware audiolinks between systems in the past and this would be a massive step backwards IMHO.

Good point trailerman. I’m just curious how the CPU and latency performance would compare on a 64GB slave between VEP via ethernet or physical outputs. That would be a great thing for DAW bench to test.

[SteveHelstrip]

@aptmusic9

@Trailerman

Hopefully this isn’t hijacking the thread, but has anyone tried using VEP on a slave computer and transferring audio using a physical sound card output (instead of ethernet) to try and reduce CPU load and run at lower latencies? Perhaps a single stereo pair via SPDIF or AES?

I understand why you’re asking this, but that would render the idea of VE Pro on a slave server somewhat pointless. You may as well use VStack/System LInk, or Plogue Bidule, or a second instance of your DAW on the slave computer, and use Midi Over Lan (or a hardware interface) and ditch VE Pro altogether.

Much of the value of VE Pro on slave machines is it’s ability to transfer large track-counts at low latency (relatively) back to the host via ethernet, without having to mix down on the slave system or use additional hardware. I’ve worked with hardware audiolinks between systems in the past and this would be a massive step backwards IMHO.

Good point trailerman. I’m just curious how the CPU and latency performance would compare on a 64GB slave between VEP via ethernet or physical outputs. That would be a great thing for DAW bench to test.

It wouldn’t compare!!!