David Kudell
Senior Member
Hopefully this video will be useful if you‘re considering a Mac, but even if you’re not, the tests comparing running samples on fast internal and Thunderbolt SSDs to standard USB SSDs were quite surprising.
Hopefully this video will be useful if you‘re considering a Mac, but even if you’re not, the tests comparing running samples on fast internal and Thunderbolt SSDs to standard USB SSDs were quite surprising.
That test was using VEPro, and the point was with 1700 tracks loaded (which does consume some CPU even when idle), how many can you still play back.Towards the end this video, I saw a reference to two different buffer sizes that make me wonder why this video wasn’t exclusively made with the kind of buffer sizes all users who want as low real time latency as reliably possible would use.
This is, according to my own experience, particularly important for those without very generous RAM amounts (most people don’t have 192 gb RAM!!) - especially when using the 6k option in Kontakt.
Thanks for your reply. Logic doesn't come with an option to select a 192 buffer, but the roundtrip latency (with RME Babyface Pro on a 2020 i7 iMac) is 7,3 ms @ 128 and 12,7 @ 256. I'm sure Apple Silicon Macs are a little better.This is likely due to the fact that there's only 4ms of latency at 192, and the audio needs to travel from VEP back to Cubase.
I’m pretty sure the latency (off the top of my head) with buffer 64 it’s in the 2ms territory, and that’s easy to achieve with hundreds of tracks not using VEP. You can even do 32 buffer.Thanks for your reply. Logic doesn't come with an option to select a 192 buffer, but the roundtrip latency (with RME Babyface Pro on a 2020 i7 iMac) is 7,3 ms @ 128 and 12,7 @ 256. I'm sure Apple Silicon Macs are a little better.
The main argument for using various PCI based Pro Tools systems back when relying on only native systems started to become usable was the low latency they had (without losing reliability). There were several discussions over at (what's today is called) Gearspace, where tolerable roundtrip latency was discussed, and around 10-15 years ago, it was possible to get a 3.3 ms roundtrip when using the 32 buffer @ 44.1 kHz. That was almost as good as the DSP based Pro Tools systems, but a lot more unstable.
Today's systems are of course a lot more powerful, but even when using the 64 buffer the best latency roundtrip I can get, even when using a 64 buffer, is 4,7 ms. Of course latency when working with virtual instruments is a different story than the audio comparisons I mentioned above, but IMO high latency is still a problem we need to deal with, which is why comparing systems using settings which pretty much cover up for what these systems can do in real life situations aren't (for me personally) that interesting. But your info is still interesting and relevant – especially what you wrote in your last post. If 192 is the lowest you can go using VEPro without getting a small pop when starting playback, that's a real bummer, but in my experience, using the 6k setting with demanding sample libraries isn't such a good idea I thought it was a while ago.
I've seen several tests performed with roundtrip latency results that are way higher than any of us would want to work with – resulting, of course, in very high track counts. Such reviews usually don't mention the elephant in the room: you'll get more tracks than you need (which isn't really interesting) if you use very high buffer settings (which isn't really interesting either).
If your powerful system, with a massive 192 gb RAM, can't give us a system that's (with the kind of workflow you describe*), my first test when I go Apple Native is to figure out the max. number of real life tracks I reliable can get by experiment with changing Logic buffer settings and those Kontakt (etc) settings, unless.... have any of you experimented with these settings on AS Macs lately?
*Personally, I don't plant to use VEPro, and won't need more than circa 100 tracks.
So – if 4ms isn't the total, real roundtrip latency... do you know what it is?As far as the buffer, (spoiler alert for the next video about VEPro on the Mac Studio) the reason for 192 is because in my testing, that's the lowest you can go using VEPro without getting a small pop when stopping playback. This is likely due to the fact that there's only 4ms of latency at 192, and the audio needs to travel from VEP back to Cubase.
Hopefully this video will be useful if you‘re considering a Mac, but even if you’re not, the tests comparing running samples on fast internal and Thunderbolt SSDs to standard USB SSDs were quite surprising.
That was my big takeaway, too. It does make me rethink whether I need to have so many libraries saved internally. Even externally, here I was trying to stick with NVMe drives instead of the cheap spinning ones!It was very interesting to see how little of a difference disk speed makes to sample and project load times.
I'm totally with you on preferring text over watching rambling videos, but this one's worth watching. It's well edited, so it's not the usual 5 minutes of info spread into 30 minutes of video. There's a lot of context to "the numbers," so I'm not sure a dry stats rundown would do the trick.What's the bottom line (for people who don't want to watch a video)?
I’m sure you meant the cheap Sata USB SSDs (and not the old school spinning drives) but just in case someone reads this and doesn’t watch the video, wanted to clarify that’s what I tested.Even externally, here I was trying to stick with NVMe drives instead of the cheap spinning ones!
You can run Logic at a 128-sample buffer and VE Pro doesn't click.As far as the buffer, (spoiler alert for the next video about VEPro on the Mac Studio) the reason for 192 is because in my testing, that's the lowest you can go using VEPro without getting a small pop when stopping playback. This is likely due to the fact that there's only 4ms of latency at 192, and the audio needs to travel from VEP back to Cubase.
I’m not sure what you mean by voice count, do you mean simultaneous playback? Other than the specific VEPro test I did in the first video, I decided it wasn’t really necessary to do any other of those types of tests…because the result always ends up being “way more than anyone could possibly, practically need.”Awesome video, extremely informative.
How about the voice count for virtual instruments running on internal SSD vs external? especially when using multi-mic instruments, such as VSL Synchron?
Right, the simultaneous playback, excuse my poor English.I’m not sure what you mean by voice count, do you mean simultaneous playback? Other than the specific VEPro test I did in the first video, I decided it wasn’t really necessary to do any other of those types of tests…because the result always ends up being “way more than anyone could possibly, practically need.”