Thanks for all the amazing tests, Matt! And I'm really grateful that you also did the test with Studio One, since that is my DAW. And - my goodness, I'm sure glad that I stuck around to see that fun session with Commander William Ryker!
Dave, thanks so much! I use Studio One a lot at home and recording the audio for these vids, but almost exclusively PT at the studio. I'm getting better with S1. Thanks again!
Thanks Matt! I really enjoy your small experiments like this one. Keep 'em coming! Now I must run a test of my own wether there is any difference between AU and VST3. I've had a few crashes with AU plugins, not so much with VST. Could be that specific plugin but it made me switch to using VSTs instead.
Great demo Matt! I’m currently not a Studio One user but I liked the fact that latency was displayed on that GUI you had opened. Are you using any Pro Tools DSP at this point or have you gone all native? Thanks!!
Thank you! At the studio I'm still fully entrenched in DSP PT (Carbon and an HD Accel system). My studio partner, however, runs fully native using a Quantum and Studio One (and the platinum records on our walls are all his work). It took a bit of work for me to get that super stable for him, but it's working great. At home I run an Apollo rig. I also track through an Apollo rig sometimes for smaller sessions at the studio-singer/songwriter stuff.
Very interesting test even though i don't understand all the details. I am also happy because I changed from PT to S1 a year ago. Do you think the AU plugins are better than VST3? (I have only used VST-plugins so far).
@@MattHepworth Since more software is M1 compatible, I'm ready to buy a M1 Mac. I built a PC (with TB3 card) around my x6. Works perfectly, but sometimes I need a newer Mac in the building. I also have a PT10 HD-2 rig with the cheese grader (2010). It still works flawlessly, but I'm definitely missing new OSX features. lol Thanks for the testing an comparison videos. 👍🏽
Hi Matt. Thanks for your great tests. Would like to ask what Mac Pro model you use? You said it's a older model. It it i7 or i9. Just to get a feel of how it might compare to my MBP 2019 i9. 🙂
Mac Pros are Xeon based. This is an upgraded cMP. Geekbench 5 is around 5K. The M1, by comparison, is around 7400. Though the Mac Pro has a lot of total horsepower, it's not as good with low latency. I suspect you'll get better results with your i9.
Thanks very much for you answer, Matt. Yes it seems I should get better results. I haven't done these tests myself. But I get the impression that there are no issues at all, when I plunge on a load of UADx's on my i9. That is impressive. But now while I've got you on the line. I did another test short after the UADx's came out, and then I forgot about it. The infamous null test. And while they all seem to null, I found that if I gained the Neve 1073 into distortion, it did not null. The what some glithchy sound coming through. Not a lot, but it was audible. Maybe I did something wrong, but just wondering if you'd be able to notice it? I could have opened a ticket, but I'd like to know if some competent people like you are able to notice it? Or if it's just me, as then I'll apologise for being unpro. Cheers.
Hey there. So, the hard distortion is likely non-linear and cannot null. My null tests were successful on all the plugins I tested, so I suspect that's it.
Hi Matt. Yes of course it could be a non linear/emulated component noise issue. But I recall I tested it up against a similar 1073 type (UAD-2 against UAD-2) that did null. It they were both non linear they would’nt null either. And funny is that API vision nulled perfectly. I think I have to do the test again, before I start any inapropriate rumors. 😉
I think the non-linear aspect is only applicable to hard clipping, but I'm not certain it even is then. I did null the 1073 in my testing... eventually.
Good stuff! If I remember correctly Avid released a white paper after the conversion to AAX and release of HD IO concerning AD/DA latency where 96k was more efficient than any other sample rate by design. I believe the IO latency for 96 was 4 ms but considerably higher at lower sample rates but I can’t find that paper, which is how I stumbled onto this video. I think 48k was 32 ms or higher so it stands to reason that PT is optimized for 96k for a few reasons. One was chipset in the HD IO converters were defaulted to 96k and needed additional cycles for other sample rates, but that’s hardware. That said the software side would need to operate under the same kind of efficiency. I think this test helps to demonstrate that. Granted you are using UAD hardware and DSP plugs in record mode, still PT seems to be optimized and targeted for use at 96k for professional use. I work in games and post so much of my work is at 48k for a few reasons but I have been working and recording at 96 depending on the project and deliverable specs. Thanks!
The measured latency for HDX with HD I/O is 1.9ms at 44.1 and 0.5ms at 96kHz. Keep in mind, that's with zero plugins. The biggest downside with 96kHz is the reduced I/O.
It is possible UADX uses a AAX wrapper. & Protools is generally less efficient on Windows. I can’t remember what version Protools stopped using a Emulation layer for its Main Code Base.
@@MattHepworth yes, and there is this other layer involved- AAX vs AU. So there is Rosetta itself on Apple Silicon as one factor between the two platforms. Then there is the question if whether AAX is more resource intensive than AU inherently, or, does UA support AAX using a less efficient method? Welcome down the rabbit hole :-)
It's not. UADx aren't M1 native, so they're using Rosetta. Rosetta is actually really efficient. I did extensive testing with M1 native verses Rosetta in a couple videos and the difference was pretty small.
Thanks for all the amazing tests, Matt!
And I'm really grateful that you also did the test with Studio One, since that is my DAW.
And - my goodness, I'm sure glad that I stuck around to see that fun session with Commander William Ryker!
Dave, thanks so much! I use Studio One a lot at home and recording the audio for these vids, but almost exclusively PT at the studio. I'm getting better with S1. Thanks again!
Thanks Matt! I really enjoy your small experiments like this one. Keep 'em coming! Now I must run a test of my own wether there is any difference between AU and VST3. I've had a few crashes with AU plugins, not so much with VST. Could be that specific plugin but it made me switch to using VSTs instead.
Thank you, Thomas! UADx are actually VST2 (for now). Let me know what you find!
Hey Matt, Cool videos. I hope you keep making them when you're inspired. Gonna PM you on the forum with some comments and questions...
Just sent you a PM. Thanks!
Great demo Matt! I’m currently not a Studio One user but I liked the fact that latency was displayed on that GUI you had opened. Are you using any Pro Tools DSP at this point or have you gone all native? Thanks!!
Thank you! At the studio I'm still fully entrenched in DSP PT (Carbon and an HD Accel system). My studio partner, however, runs fully native using a Quantum and Studio One (and the platinum records on our walls are all his work). It took a bit of work for me to get that super stable for him, but it's working great. At home I run an Apollo rig. I also track through an Apollo rig sometimes for smaller sessions at the studio-singer/songwriter stuff.
Very interesting test even though i don't understand all the details. I am also happy because I changed from PT to S1 a year ago. Do you think the AU plugins are better than VST3? (I have only used VST-plugins so far).
Thanks! I think the outlier is Pro Tools and AAX. AU and VST seem to both work as expected.
Very Nice Matt, Now that the Protools M1 Beta is out, any update?
I have too many non-M1 plugs to use the beta, unfortunately.
What was the buffer set at in PT when you recorded @96K on the M1?
I believe it was 64.
@@MattHepworth Since more software is M1 compatible, I'm ready to buy a M1 Mac. I built a PC (with TB3 card) around my x6. Works perfectly, but sometimes I need a newer Mac in the building. I also have a PT10 HD-2 rig with the cheese grader (2010). It still works flawlessly, but I'm definitely missing new OSX features. lol Thanks for the testing an comparison videos. 👍🏽
Thank you! I think the only weakness on the M1 is the slow USB performance.
Hi Matt. Thanks for your great tests. Would like to ask what Mac Pro model you use? You said it's a older model. It it i7 or i9. Just to get a feel of how it might compare to my MBP 2019 i9. 🙂
Mac Pros are Xeon based. This is an upgraded cMP. Geekbench 5 is around 5K. The M1, by comparison, is around 7400. Though the Mac Pro has a lot of total horsepower, it's not as good with low latency. I suspect you'll get better results with your i9.
Thanks very much for you answer, Matt. Yes it seems I should get better results. I haven't done these tests myself. But I get the impression that there are no issues at all, when I plunge on a load of UADx's on my i9. That is impressive.
But now while I've got you on the line. I did another test short after the UADx's came out, and then I forgot about it. The infamous null test. And while they all seem to null, I found that if I gained the Neve 1073 into distortion, it did not null. The what some glithchy sound coming through. Not a lot, but it was audible. Maybe I did something wrong, but just wondering if you'd be able to notice it? I could have opened a ticket, but I'd like to know if some competent people like you are able to notice it? Or if it's just me, as then I'll apologise for being unpro. Cheers.
Hey there. So, the hard distortion is likely non-linear and cannot null. My null tests were successful on all the plugins I tested, so I suspect that's it.
Hi Matt. Yes of course it could be a non linear/emulated component noise issue. But I recall I tested it up against a similar 1073 type (UAD-2 against UAD-2) that did null. It they were both non linear they would’nt null either. And funny is that API vision nulled perfectly.
I think I have to do the test again, before I start any inapropriate rumors. 😉
I think the non-linear aspect is only applicable to hard clipping, but I'm not certain it even is then. I did null the 1073 in my testing... eventually.
Great job man.
Thank you!
Good stuff! If I remember correctly Avid released a white paper after the conversion to AAX and release of HD IO concerning AD/DA latency where 96k was more efficient than any other sample rate by design. I believe the IO latency for 96 was 4 ms but considerably higher at lower sample rates but I can’t find that paper, which is how I stumbled onto this video. I think 48k was 32 ms or higher so it stands to reason that PT is optimized for 96k for a few reasons. One was chipset in the HD IO converters were defaulted to 96k and needed additional cycles for other sample rates, but that’s hardware. That said the software side would need to operate under the same kind of efficiency. I think this test helps to demonstrate that. Granted you are using UAD hardware and DSP plugs in record mode, still PT seems to be optimized and targeted for use at 96k for professional use. I work in games and post so much of my work is at 48k for a few reasons but I have been working and recording at 96 depending on the project and deliverable specs. Thanks!
The measured latency for HDX with HD I/O is 1.9ms at 44.1 and 0.5ms at 96kHz. Keep in mind, that's with zero plugins. The biggest downside with 96kHz is the reduced I/O.
It is possible UADX uses a AAX wrapper. & Protools is generally less efficient on Windows. I can’t remember what version Protools stopped using a Emulation layer for its Main Code Base.
Interesting. I don't remember hearing about that. I'll check it out!
Pretty weird, but cool that you narrowed this down to PT and AAX
What was shocking to me is how much more efficient Studio One was at both samplerates. Over twice the performance.
@@MattHepworth yes, and there is this other layer involved- AAX vs AU. So there is Rosetta itself on Apple Silicon as one factor between the two platforms. Then there is the question if whether AAX is more resource intensive than AU inherently, or, does UA support AAX using a less efficient method?
Welcome down the rabbit hole :-)
How long can AVID cut leeks?
Anybody know what Processors are in those computers? Are they laptops?
Which ones?
@@MattHepworth the ones he's using.
The ones I'm using in the vid?
@@MattHepworth yes.
M1 in a Mac Mini. The Intel is a dual CPU XEON cMP. Geekbench 5 of >7400 and
Isn’t this because studio 1 is M1 native and pro tools is not?
It's not. UADx aren't M1 native, so they're using Rosetta. Rosetta is actually really efficient. I did extensive testing with M1 native verses Rosetta in a couple videos and the difference was pretty small.
@@MattHepworth thanks for the reply!