Question for Chris/Helge re: low latency in Cubase

Steinberg, are you monitoring this important topic closely enough? :smiley:

Maybe they’ve heard about it? After all, they’re in the business …

Then again …

Actually the numbers are SMALL: 512, Med:1024, Large: 2048. I know in Cubase you can just change the setting if all you are doing is playback but the Logic way is much more useful.

Chris

Ah, I hadn’t realized the large setting was 2048.

I just want to clarify by the way - this thread wasn’t intended to turn into a Cubase vs Logic thread, or in some way just coming on the forum to sing the praises of Logic. In fact it’s the opposite, my trial of Cubase 6 has reminded me of so many reasons why I was a Cubase user for many years, and I am so keen to move back.

This buffer issue is really the only thing that is stopping me… on the one hand it is just “one thing” out of so many different issues and specs to weigh up when choosing a DAW, but on the other hand it feels like a deal breaker at the moment, when I try to get comparable performance to Logic in this area.

I’m looking forward to hearing from the Steinberg guys, and would be over the moon to hear of a big forthcoming advance in performance for Mac users, especially if it involved adopting dual buffers as we have been discussing.

Cheers and happy new year!
Mike

Hi Mike,
I’ve pretty much made up my mind that Cubase is the way to go. The further along I go the more and more I find it does that Logic just can’t do…getting a complete tempo map from a recorded track that wasn’t following the metronome for starters. I think if you just learn to start freezing tracks sooner then you can do a lot more. You may also want to look into getting a Mini and VEP as a slave (or a PC slave).

Chris

Only implement it if it can be turned off. Multiple buffers makes delay compensation that much more difficult…and thus placement of what I recorded…

I personally will never use Cubase or any other software to monitor (again). There are already units, both analog and digital that are real time, bufferless, and made for monitoring. Be it the digital mixer in a product like Steiny’s FW interfaces with FX or an analog mixer.

At some point, you have to pan back and ask yourself how far you’re going to go to solve a problem that’s easily solved with an inexpensive Mackie mixer and a used FX box (on the low end of price)…how serious are you about recording that an Onyx and a used Quadraverb2 is too much to spend to have ACTUAL zero latency monitoring?

That’s only useful for tracking audio but if you are using mostly VI’s as I am you have to have low latency. I don’t really have much issue even at 256 or the next setting down. I also have a dedicated Firewire card which seemed to improve things slightly (Mac Pro 2.66 quad).

Chris

The further along I go the more and more I find it does that Logic just can’t do…getting a complete tempo map from a recorded track that wasn’t following the metronome for starters.

this is unrelated…but, it’s not really that useful, because it’s not really very accurate. With either app, you want to tap the tempo in. Logic’s “reclock song” works a bit more elegantly than Cubase’s deriving tempo from a recording tapped midi track. Neither work as well as MOTU’s tap tempo functionality. All get you where you’re going.

I was looking forward to the auto detection in 6…figuring that was the next step for something I’d done for 25 years with MIDI sequencers…but, in reality, the problem is it works on transients…and just because someone hit something really loudly doesn’t mean that’s where the beat is for any purpose you would use a tempo map for.

FWIW…tap the tempo onto a midi track…extract a tempo map from the tapped track…and then turn on the warp tool to fine tune it by EAR using the built in click. You don’t necessarily want to define the beat by where the loudest transient is. You want the most steady subdivision you can get so that copy/pasted MIDI from different sections actually plays in time with other sections, regardless of where around the beat the drummer might’ve whacked the snare.

Ah…VIs…get another box for them. that’s actually what I use my Cubase PC for. :slight_smile:

Now that I have the Kronos…I only turn it on to sequence strings. Damn, VSL…

For recording audio with some basic reverb, I agree - monitoring directly through your interface is the way to go! I do that through my RME UFX for vocals etc.

I’m more concerned with situations where I’m using virtual instruments, or where we are using complex plug-ins to generate effects that are essential to the performance of the part (i.e. that can’t be added later).

For sure - if it’s vocals or other instruments that are being tracked, and only some reverb or a bit of compression is required for the monitor mix, I absolutely agree - going through Cubase (or any DAW) is unnecessary!

I had great luck with C6’s implimentation of it and tap tempo is so last year. :wink:

@ Mike:

There is no question (as I said in my earlier post) that Logic’s scheme allows for lower VSTI input latency. I have been using Cubase on a Mac since the earliest version of VST ( I think it was in 1997). The program is amazingly stable now compared to then, very powerful and I love the workflow. That said, If you’re expecting lowered Mac latency for VSTi’s to be part of Cubase’s refresh cycle, I’m sad to say I think you’ll be disappointed. Otherwise, you’d probably have heard from a moderator in this thread that it was coming.

My understanding is that things are somewhat better on the PC side. The PC users who’ve chimed in on this thread are obviously trying to be helpful with their suggestions of freezing, constraining delay compensation, etc etc. It’s all true and well and good, but if like me, you do a lot of realtime orchestral work or similarly demanding tasks where many changes need to be done before the final delivery, you will probably run out of CPU power when working at a latency under 256, sometimes under 512. For me as a player, 512 is pretty hard to work with when trying to play fast passages, and yet-that’s just the way it is. As you can see in my specs, my computer is fairly powerful, though getting a bit aged.

So-if you love Cubase’s workflow as much as I do, you have choices. Accept things as they are, or use the suggestions others have made, or most effectively, get VE Pro and a fast slave PC with SSD’s for sample streaming and CPU horsepower. Well, I guess one last option is…to keep hoping for re-coding from Steinberg. That seems the least practical : )

Cheers and happy New Year!

I use both Cubase and Logic - Mac and PC (Cubase only for PC of course)

And it’s true that Logic uses a separate buffer to recording and playback BUT it has too.

In Cubase you can change the buffer setting virtually on the fly! Instantly switching between 64, 128, 512 etc

Try that in Logic … every single time you change the buffer setting in Logic, the program reloads every single VI and EXS24, Kontakt, SD2 intrument - it’s a royal PITA.


best
tmy

That solution doesn’t work for MIDI recording of a VST instrument, or say recording direct guitar with a VST amp sim plug, both of which where the desired sound will affect the performance. Gotta have low latency. Your solution works when you’re recording voice, acoustic guitar, or some source where lack of final processed sound won’t be detrimental to the performance.

+1 for different buffers for active and inactive tracks.
I was about to jump from Logic to Cubase when buying a new Mac Mini Server. I was shocked when I started to record in Cubase. The performance was worse than on my retired computer that, on the paper, was inferior to the 2011 MMS. First I thought there was a problem with the new MMS but I installed Logic on it and that revealed that the MMS was much more powerful that the old computer. The problem was not the MMS but Cubase.
I compose and arrange music exclusively with VSTi and ITB effects and need low latency (max 128) while doing this. Now I´m back to Logic and am kind of disappointed. I prefer Cubase over Logic but this just kills it.
While the CPU´s gets faster the effect/VSTi developers starts using more demanding alogarithms etc that gives us better sound quality but also taxes the CPU´s more. Logics solution to use larger buffers for tracks that is not input monitored is the best there is right now. One can keep on tracking guitars with AMP sims or powerful VSTi´s on large projects with low latency without having to think about freezing tracks etc.
Please do something about this Steinberg because buying a new computer is not the solution, it is in the code! :wink:

  • Vesa

I am running Cubase 6.0.5 on Windows 7 in a live event environment: recording, monitoring and room amplification all run through Cubase, live. So from the very beginning, low latencies were the key concern when I started building the system. All I/O channels need to run at a roundtrip latency 5ms or lower, and all tracks have several performance expensive plug-ins active during live performance.

I experimented with several audio interfaces and pcs (Windows only, I have no Mac experience). When I did my experiments, I was unable to find some clear information from other users about what is possible, so want to share some numbers from the system we have in place right now:

The system is:

  • a high performance Dell system with 2 Xeon cpus, 6 cores per cpu running at 3.4 GHz
  • 3 rme Multiface II audio interfaces, attached to the PC with 3 rme PCIe HDSP cards
  • the operating system is Windows 7 64bit

The system load is:

  • 25 input channels
  • 36 output channels
  • about 100 tracks in Cubase

The most performance hungry plug-ins in the system are:

  • 27 instances of Izotope Alloy running in tracking mode
  • 10 instances of Izotope Nectar running in tracking mode
  • 6 instances of SIR running (impulse response reverb plugin, running in high quality mode)
  • 1 Pianoteq instance running (VST instrument, takes quite some cpu but great sound)

The system runs with 64 samples buffer size at 44.1 kHz sample rate, the electronically measured roundtrip latency (analog in to analog out) is 5.40ms, and the Cubase VST performance meter runs at around 75%. The system is almost maxed out, there is some headroom left, and the system is absolutely stable and clean.

Regards, Mikael

the problem with logics hybrid buffer engine is very bad multi-threading/core loading/scalling. OSX is generally not good in this matter either compared to windows hence logic’s use of the hybrid buffer to overcome this OS limitation.


OSX also doesn’t have ASIO direct monitoring so you’re hampered there as well. These issues are the main reason i’ve been a windows user for a long time.

i did a large orchestral album last year all at 96k and never changed my 1.5 ms RME buffer size for the whole project.


MC

If you have the MR816 series you have Direct Monitoring in OSX…it may be the only interface that allows it.

Chris

+1

I find I can stay at very low latencies as well with very large projects also with an RME interface, RME do seem to have very efficient low latency drivers.

I have found Cubase to be much more efficient with VI’s than Logic, maybe because Steinberg invented the technology?

tmy

The MR816 allows direct monitoring with direct control through Cubase (and is unique in this), however most other interfaces also allow direct monitoring - either through their hardware or their virtual mixer.

RME do this, and when I am recording an audio source that doesn’t need specific plug-ins during the tracking, I monitor directly through the RME interface.

When using VIs or when I want to use a specific effect, and need to hear it while tracking the take, the only option is to monitor directly through the DAW…

Wow - that’s pretty amazing, I would love to know how you managed to stay at such a low buffer when doing such a large project? Were you using VIs, or only working with actual miked up instruments?

Also, were you on Mac, or on Windows (where most users seem to report amazing low-latency performance even on huge projects…)?