New Logic Multitrack Benchmark Test

https://www.bluecataudio.com/Blog/behind-the-scenes/apple-m1-vs-intel-for-audio-a-benchmark/

 

from "horse's" mouth, sorry it's third part but i guess relevant anyway

It is a Rosetta performance comparison.

https://www.bluecataudio.com/Blog/behind-the-scenes/apple-m1-vs-intel-for-audio-a-benchmark/

 

from "horse's" mouth, sorry it's third part but i guess relevant anyway

 

It is a Rosetta performance comparison.

Native, Rosetta and Intel

 

Well I'd like to compare it especially with Pro Tools running a powerful desktop processor versus optimized M1 in Logic with the same plugins.

 

Yes this actually was a good idea to compare 2 daws on 2 different systems with same third party plugins.

This way someone could decide to go with a new Mac and logic or stay on a powerful desktop system in the PC world. Thanks for that but also agree needs its own thread.

How many tracks are you able to run with your Air after it warmed up for like 15 Minutes at this setting? I mean let's say warm it up with 90 Tracks and then see how much more you can activate?

https://music-prod.com/logic-pro-x-benchmarks/

Could you also run this benchmark? This one is with Fabfilter Plugins, which is a little more CPU intensive:

https://drive.google.com/file/d/1-X0wxZxcEv5YZGyLiUXB2UvnIycSrMyK/view?usp=sharing

 

It is a Rosetta performance comparison.

Native, Rosetta and Intel

I've got my Pro now and set up a proper benchmark:

https://drive.google.com/file/d/1-X0wxZxcEv5YZGyLiUXB2UvnIycSrMyK/view?usp=sharing

I'm getting 80 tracks with my Fabfilter setup (79 more stable for a longer time).

I just wonder now how much worse the Air is because I'm thinking about returning it for the Air, given the ventless design and I could for the same money add 1 TB to it.

I will run this tonight for you.

 

Yes this actually was a good idea to compare 2 daws on 2 different systems with same third party plugins.

This way someone could decide to go with a new Mac and logic or stay on a powerful desktop system in the PC world. Thanks for that but also agree needs its own thread.

 

How many tracks are you able to run with your Air after it warmed up for like 15 Minutes at this setting? I mean let's say warm it up with 90 Tracks and then see how much more you can activate?

 

https://music-prod.com/logic-pro-x-benchmarks/

 

m1.jpg

 

Could you also run this benchmark? This one is with Fabfilter Plugins, which is a little more CPU intensive:

 

https://drive.google.com/file/d/1-X0wxZxcEv5YZGyLiUXB2UvnIycSrMyK/view?usp=sharing

I will run this tonight for you.

 

 

 

How many tracks are you able to run with your Air after it warmed up for like 15 Minutes at this setting? I mean let's say warm it up with 90 Tracks and then see how much more you can activate?

 

https://music-prod.com/logic-pro-x-benchmarks/

 

m1.jpg

 

Could you also run this benchmark? This one is with Fabfilter Plugins, which is a little more CPU intensive:

 

https://drive.google.com/file/d/1-X0wxZxcEv5YZGyLiUXB2UvnIycSrMyK/view?usp=sharing

Can you do both? The Logic Pro X stock plugin benchmark available here and my Fabfilter one?

On the Logic Stock plugin benchmark, after having warmed up, I'm getting around 104/105 tracks with the Pro. But you need to set everything to all 8 cores, 1024 samples and large buffer size.

On my Fabfilter one, I'm at 80 / 79.

M1 Macbook Air 16 GB Ram 8 core gpu:

Logic native benchmark test(Logicbenchmarktest2.zip) : 106 tracks

Fabfilter one: 83 tracks

M1 Macbook Air 16 GB Ram 8 core gpu:

 

Logic native benchmark test(Logicbenchmarktest2.zip) : 106 tracks

Fabfilter one: 83 tracks

Impressive, that's faster than my Macbook Pro M1 :>

Did you let it get warm before you awuired these results or are they cold start results?

These results seem unrealistic on the Air - especially for long term (cycle repeat), it would throttle sooner or later

Well it could be quite possible because the air has a more potent passive cooling solution, meaning more surface area to cool the chip unless it receive more heat than the block can dissipate in time.

I wonder if he had heaten it up like I said for like 15 minutes before doing the benchmark.

I had to try for awhile before i got to this result. At ‘cold’ try, it was about 10 tracks less in each benchmark. Of course these track counts could not sustain. These results were sustainable for close to 1 cycle. So absolutely max. Looks like the AIR is as powerful as the PRO but when it comes to sustainability the PRO will do more. We saw this in youtube tests. Especially with long video rendering. In reality in the music creation world the two machines would be very close. I choose the AIR because i record in my bedroom and there is not even a chance that a fan would turn on. The Air is powerful enough for me. Great little beast.

These results seem unrealistic on the Air - especially for long term (cycle repeat), it would throttle sooner or later

 

Well it could be quite possible because the air has a more potent passive cooling solution, meaning more surface area to cool the chip unless it receive more heat than the block can dissipate in time.

 

I wonder if he had heaten it up like I said for like 15 minutes before doing the benchmark.

I had to try for awhile before i got to this result. At ‘cold’ try, it was about 10 tracks less in each benchmark. Of course these track counts could not sustain. These results were sustainable for close to 1 cycle. So absolutely max. Looks like the AIR is as powerful as the PRO but when it comes to sustainability the PRO will do more. We saw this in youtube tests. Especially with long video rendering. In reality in the music creation world the two machines would be very close. I choose the AIR because i record in my bedroom and there is not even a chance that a fan would turn on. The Air is powerful enough for me. Great little beast.

 

 

 

 

Well it could be quite possible because the air has a more potent passive cooling solution, meaning more surface area to cool the chip unless it receive more heat than the block can dissipate in time.

 

I wonder if he had heaten it up like I said for like 15 minutes before doing the benchmark.

I appreciate your results a lot: could you check and see what would sustain a 10 minute loop?

Michael

Ok new results running tests for over 10 minutes:

Logic native plugins: 98 tracks

Fabfilter plugins: 75 tracks

Ok new results running tests for over 10 minutes:

 

Logic native plugins: 98 tracks

 

Fabfilter plugins: 75 tracks

I ran both for a few hours to get them to the hottest temperature with the following:

Logic Benchmark:

Pro 104

Air 94

Fabfilter:

Pro 79

Air 71

So the difference stays the same, 10% performance drop.

sounds reasonable

did you make a third party topic already?

Yes very reasonable and good results for the fanless AIR and the PRO.

Ideally you don’t want to use a computer for hours and hours all the time at the limit of the CPU. Both computers have great power using Logic for sure.

sounds reasonable

 

did you make a third party topic already?

I know this thread is mostly about M1 performance (and I recently tried an M1 Air and got 95 tracks with the benchmark) but my problem is with the Mac Pro 16 core.

I only get about 180 tracks with the benchmark, and I've seen reports of 235 and 260 with other 16 core users.

I'm on Catalina 10.15.7 and Logic 10.6.1

Machine is a Mac Pro 7.1 16 core 48GB RAM 1TB SSD

Buffer is set to 1024 (although I get almost identical results at 32) and I tried both Mac Pro speaker and Babyface Pro as output device. Again, almost no difference.

Mixer precision is set to 64 bit and I get a few more track at 32 bit (about 190) and buffer is set to medium. I used 32 threads but 'automatic' doesn't change the result much either,

Logic's CPU meter maxes out at 100% on all cores except the last (as expected) but iStat menus tells me the CPU is running at about 84-86%.

Quitting istat menus (and killing all the processes) doesn't help, neither does removing Dropbox, iDrive and TGPro, which are about all I have running.

Things that are also present but can't be quit are: CCC, Avast, MalwareBytes. I can uninstall them and see if it helps, but I bet it won't!

When I run the Intel power gadget, my single core turbo boost performance won't go above 4.23GHz, and when the Logic benchmark is running, iStat Menus tells me that CPU frequency is 3.9GHz, which seems reasonable for all cores going flat out.

So, any clues please? Do I just have a lazy CPU ??

I know this thread is mostly about M1 performance (and I recently tried an M1 Air and got 95 tracks with the benchmark) but my problem is with the Mac Pro 16 core.

I only get about 180 tracks with the benchmark, and I've seen reports of 235 and 260 with other 16 core users.

I'm on Catalina 10.15.7 and Logic 10.6.1

Machine is a Mac Pro 7.1 16 core 48GB RAM 1TB SSD

Buffer is set to 1024 (although I get almost identical results at 32) and I tried both Mac Pro speaker and Babyface Pro as output device. Again, almost no difference.

Mixer precision is set to 64 bit and I get a few more track at 32 bit (about 190) and buffer is set to medium. I used 32 threads but 'automatic' doesn't change the result much either,

 

Logic's CPU meter maxes out at 100% on all cores except the last (as expected) but iStat menus tells me the CPU is running at about 84-86%.

Quitting istat menus (and killing all the processes) doesn't help, neither does removing Dropbox, iDrive and TGPro, which are about all I have running.

Things that are also present but can't be quit are: CCC, Avast, MalwareBytes. I can uninstall them and see if it helps, but I bet it won't!

When I run the Intel power gadget, my single core turbo boost performance won't go above 4.23GHz, and when the Logic benchmark is running, iStat Menus tells me that CPU frequency is 3.9GHz, which seems reasonable for all cores going flat out.

So, any clues please? Do I just have a lazy CPU ??

I dunno, 180 tracks sound pretty good for 16 cores of that Intel Type processor, given my 9900k manages around 130 to 140, tracks, which is akin to a 12 Core Pro 2019.

My 9900k is OCd to run at 5.1 GHz on all cores during AVX and it is the same speed as the 12 Core Mac Pro from 2019, doing 140 tracks. Now, since the 16Core is 20% faster than your 12, according to geek bench, it is 130 x 1.2, which is , 168, which means 180 is actually very realistic and good.

The numbers posted on music-prod are wrong.

For example, the Air says 110 there, but in reality, it is 94/95 tracks. Oddly enough the Pro is correct, with 105 tracks.

But 180 tracks with the 16 core xeon is pretty good. Xeons are made for stability not performance. Using a 10 Core 11 series processor at 5GHz on every core will be very close to your 16 Core Xeon.

lol, i can't believe were at a point we're comparing an entry level laptop with a single fan that barely comes on, to a friggin 16-core workstation

i really hope they scale M1 performance to workstation level.

sounds reasonable

 

did you make a third party topic already?

Tried to but I'd need copyright free music or maybe use that Vi track from the original Benchmark because all the stuff I got is label bound, hard to out something up which doesn't have a copyright issue.

Can't put it up with Ocean Eyes, it's gonna cause issues.

 

I dunno, 180 tracks sound pretty good for 16 cores of that Intel Type processor, given my 9900k manages around 130 to 140, tracks, which is akin to a 12 Core Pro 2019.

 

My 9900k is OCd to run at 5.1 GHz on all cores during AVX and it is the same speed as the 12 Core Mac Pro from 2019, doing 140 tracks. Now, since the 16Core is 20% faster than your 12, according to geek bench, it is 130 x 1.2, which is , 168, which means 180 is actually very realistic and good.

 

The numbers posted on music-prod are wrong.

 

For example, the Air says 110 there, but in reality, it is 94/95 tracks. Oddly enough the Pro is correct, with 105 tracks.

 

But 180 tracks with the 16 core xeon is pretty good. Xeons are made for stability not performance. Using a 10 Core 11 series processor at 5GHz on every core will be very close to your 16 Core Xeon.

Thanks, that makes sense to me.

I found another video of someone managing 309 tracks on their 16-core MP 2019. I wonder if they're just adding audio/midi tracks without duplicating all the plug-ins? Who knows....

Just makes me think of 4 track Studers. Who, other than film/orchestral composers needs that many tracks? I don't. I don't think I've ever used more than 27, tops, VIs and audio.

Things move on. Music now doesn't sound like it did in the 1960's. Technology is one of the things that changes - it's musicians as well as engineers who push the boundaries of technology, in order to find new, novel sounds. Even the Beatles linked two 4-tracks together.

Lots of people who do film / orchestral composition need 70-80 tracks for each 'articulation' - a whole orchestra (or more normally just a string section) playing legato, or pizzicato, or tremolo. So not all tracks are sounding at the same time, but the VIs are loaded and ready to play on multiple articulations. In the 1960's, you needed an orchestra, and a hall. But probably only two mics.

For pop/rock/country, stuff gets shipped around as stems - 5-10 takes per instrument. Easy to get high track counts for that stuff too, although not as high as film.

And in the middle there are fusion bands and world music with choirs and tens of percussion players or multi-miked drum kits. It's really easy to swallow up the tracks.

Just makes me think of 4 track Studers. Who, other than film/orchestral composers needs that many tracks? I don't. I don't think I've ever used more than 27, tops, VIs and audio.

I am currently working on a rock song with a friend, collaborating overseas. I have like 10 alternatives now in the session and trying out different guitar sounds so I'm hiding tracks etc. I am already around 70 tracks. Simple rock song. This includes Logic's Drummer tracks (in a summing track) etc. There is compression EQ delays reverbs virtual amps etc. all over the place. Flextime, and some flex pitch used. The GUI/mixer everything is buttery smooth. The good news is my M1 Macbook Air is using around 10% of its CPU. Extremely pleased with this machine.

Just makes me think of 4 track Studers. Who, other than film/orchestral composers needs that many tracks? I don't. I don't think I've ever used more than 27, tops, VIs and audio.

 

I am currently working on a rock song with a friend, collaborating overseas. I have like 10 alternatives now in the session and trying out different guitar sounds so I'm hiding tracks etc. I am already around 70 tracks. Simple rock song. This includes Logic's Drummer tracks (in a summing track) etc. There is compression EQ delays reverbs virtual amps etc. all over the place. Flextime, and some flex pitch used. The GUI/mixer everything is buttery smooth. The good news is my M1 Macbook Air is using around 10% of its CPU. Extremely pleased with this machine.

Exactly. I've been working on some stuff for the past 10 years, some of it being above my average pay grade and whenever you came across serious sessions, always less was more, especially sonically. There is also a major artist (post audio mixing), who's name I leave in the unknown, who has probably 30% of applied inserts by making the right choices versus the new kid, applying 15 inserts on a vocal chain not knowing what their doing.

All I'm saying, is that even I, doing serious sessions myself, wouldn't even go beyond 40% of the M1's processing power so why should I wait for the next chip? It's gonna be more expensive anyways and the M1 as it stands already surpasses the most expensive Macbook Pro 8-Core from part year in pure processing performance. What more do you need?

If I do heavy VIs for orchestra sessions during pre-peoduction, that's what my desktop hackintosh is for but I rarely come across that, maybe 10%.

Most professional Vi users have a very ergonomical way of using virtual instruments and don't need crazy amounts of RAM.

 

I dunno, 180 tracks sound pretty good for 16 cores of that Intel Type processor, given my 9900k manages around 130 to 140, tracks, which is akin to a 12 Core Pro 2019.

 

My 9900k is OCd to run at 5.1 GHz on all cores during AVX and it is the same speed as the 12 Core Mac Pro from 2019, doing 140 tracks. Now, since the 16Core is 20% faster than your 12, according to geek bench, it is 130 x 1.2, which is , 168, which means 180 is actually very realistic and good.

 

The numbers posted on music-prod are wrong.

 

For example, the Air says 110 there, but in reality, it is 94/95 tracks. Oddly enough the Pro is correct, with 105 tracks.

 

But 180 tracks with the 16 core xeon is pretty good. Xeons are made for stability not performance. Using a 10 Core 11 series processor at 5GHz on every core will be very close to your 16 Core Xeon.

 

Thanks, that makes sense to me.

I found another video of someone managing 309 tracks on their 16-core MP 2019. I wonder if they're just adding audio/midi tracks without duplicating all the plug-ins? Who knows....

 

Just makes me think of 4 track Studers. Who, other than film/orchestral composers needs that many tracks? I don't. I don't think I've ever used more than 27, tops, VIs and audio.

 

Things move on. Music now doesn't sound like it did in the 1960's. Technology is one of the things that changes - it's musicians as well as engineers who push the boundaries of technology, in order to find new, novel sounds. Even the Beatles linked two 4-tracks together.

Lots of people who do film / orchestral composition need 70-80 tracks for each 'articulation' - a whole orchestra (or more normally just a string section) playing legato, or pizzicato, or tremolo. So not all tracks are sounding at the same time, but the VIs are loaded and ready to play on multiple articulations. In the 1960's, you needed an orchestra, and a hall. But probably only two mics.

For pop/rock/country, stuff gets shipped around as stems - 5-10 takes per instrument. Easy to get high track counts for that stuff too, although not as high as film.

And in the middle there are fusion bands and world music with choirs and tens of percussion players or multi-miked drum kits. It's really easy to swallow up the tracks.

Yeah I mean, the 2019 Mac Pro is a nice machine but you'd probably would've been better off running an Intel or Ryzen Hackintosh for music production.

But it is still a great machine, however you paid a hefty Apple Premium on it since Ryzen 5000 series CPU at 700 bucks equals in performance with the fastest Mac Pro from 2019.

And these Benchmarks with 300+ tracks are flawed or frozen. Intel Xeons don't stack up that well against the K series such as the 9900K or 11900k running with 5 GHz on all cores with nice cooling.

I'm scoring about 11k on Geekbench with my 9900K running at 5.1 GHz on all cores with a nice Noctua Cooler. If I look at the Geekbench Mac Scores, the 12 Core 2019 Mac Pro sits at 12.000, so I'm fairly happy.

My friend scores 14.500 points with his Ryzen 5950x but Geekbench is very Intel optimized and ways to cheat the scores, Cinebench is a better point of reference.

Cinebench is purely CPU focused.

The M1 scores roughly 7800 points (7200) for the Air, which is amazing.

My 9900KS sits at 14.000 and I'm barely doing more than 20% usage on hefty sessions.

In Cinebench R20, the 28 Core Mac Pro 2019 is surpassed by the 5950X with 10409 points, dialing in at 700 Euros plus a TB3 Motherboard for 500.

My 9900KS runs at 4800 points in stock, which at 5.1 GHz on every core, reaching 6000 points, very close to your 16 Core MP at 6800 points.

If we translate the M1 results to Cinebench R20, it would be at 2900 points.

That's pretty much half as fast as your 16-Core Mac Pro, which I find impressive no?

And about 40% slower than my stock 9900KS which I don't use beyond 30% of usage anyways in demanding sessions on God damn awful Pro Tools Ultimate 2021.3.0.

 

Yeah I mean, the 2019 Mac Pro is a nice machine but you'd probably would've been better off running an Intel or Ryzen Hackintosh for music production.

 

But it is still a great machine, however you paid a hefty Apple Premium on it since Ryzen 5000 series CPU at 700 bucks equals in performance with the fastest Mac Pro from 2019.

 

And these Benchmarks with 300+ tracks are flawed or frozen. Intel Xeons don't stack up that well against the K series such as the 9900K or 11900k running with 5 GHz on all cores with nice cooling.

 

I'm scoring about 11k on Geekbench with my 9900K running at 5.1 GHz on all cores with a nice Noctua Cooler. If I look at the Geekbench Mac Scores, the 12 Core 2019 Mac Pro sits at 12.000, so I'm fairly happy.

 

My friend scores 14.500 points with his Ryzen 5950x but Geekbench is very Intel optimized and ways to cheat the scores, Cinebench is a better point of reference.

 

Cinebench is purely CPU focused.

 

The M1 scores roughly 7800 points (7200) for the Air, which is amazing.

 

My 9900KS sits at 14.000 and I'm barely doing more than 20% usage on hefty sessions.

 

In Cinebench R20, the 28 Core Mac Pro 2019 is surpassed by the 5950X with 10409 points, dialing in at 700 Euros plus a TB3 Motherboard for 500.

 

My 9900KS runs at 4800 points in stock, which at 5.1 GHz on every core, reaching 6000 points, very close to your 16 Core MP at 6800 points.

 

If we translate the M1 results to Cinebench R20, it would be at 2900 points.

 

That's pretty much half as fast as your 16-Core Mac Pro, which I find impressive no?

And about 40% slower than my stock 9900KS which I don't use beyond 30% of usage anyways in demanding sessions on God damn awful Pro Tools Ultimate 2021.3.0.

Yes, I got myself caught in a corner and by last September I had to buy something. My 2015 MBP was out of steam despite doing an SSD upgrade and an external eGPU didn't work out, but did waste a lot of time. A new 2019 MBP 16" was a disaster with my 49" 32:9 screen - fans were on all the time so I sent it back. A new (Intel) Mac Mini had a different problem with the same screen. So I bought a Mac Pro and it's been great apart from the lack of portability and the absurd price. Runs Pro Tools, which the 2015 MBP just couldn't cope with and has been super-stable. I've worked in IT most of my life but had no appetite for a Hack, but I agree that they represent far better value for money. I might have been tempted by one, or even a Win10 machine, if it hadn't been for my UAD stuff which needs TB. I was in a hurry so I paid the premium.

The M1 Air I tried was fantastic and easily powerful enough for the music I do but the 32:9 screen was a problem again. No sign of 5140x1440 or 3840x1080, instead the max resolution was 3008x846 ! A few other things were a bit flaky but in general it was very good, especially for the price. If I hadn't already got the Mac Pro I would probably have kept it. As things stand, I can't really sell the Pro, the bottom has dropped out the market and eBay have some very difficult sellers limits too unless you have recent high-value history. But it does what it does well, if you ignore the price point.

https://9to5mac.com/2020/12/24/m1-mac-with-ultrawide/

Yes, but not fixed three months later. Will probably never be fixed in Catalina - i.e. never fixed on the Intel Mac Mini.

I don't trust Apple to fix anything that doesn't work out of the gate anymore. When they do, it's just good luck.

They also seem to break things that previously were fine a lot more than they used to. And then not acknowledge the problem and not put it right.

https://9to5mac.com/2020/12/24/m1-mac-with-ultrawide/

Yes, but not fixed three months later. Will probably never be fixed in Catalina - i.e. never fixed on the Intel Mac Mini.

I don't trust Apple to fix anything that doesn't work out of the gate anymore. When they do, it's just good luck.

They also seem to break things that previously were fine a lot more than they used to. And then not acknowledge the problem and not put it right.

Has it been posted in the official apple support forum?

With hardware issues or upgrades I agree but when it comes to software they do usually fix these things. I mean what's the official supported maximum resolution for M1?

It's an easy fix tbh.

Lots of threads on the Apple forums, and one thread with hundreds of posts on MacRumors. It’s not an easy fix if you have to wait for Apple and they won’t fix it. Similar issues with the 5700MPX modules causing crashes on the Mac Pro - endless complaints from video professionals, and no progress from Apple.

The M1 should support 4K and 5K screens. The specs are not specific about 32:9 screens but they work on the Mac Pros and some of the Intel laptops but not on the M1s or the Intel Minis.

https://9to5mac.com/2020/12/24/m1-mac-with-ultrawide/

Yes, but not fixed three months later. Will probably never be fixed in Catalina - i.e. never fixed on the Intel Mac Mini.

I don't trust Apple to fix anything that doesn't work out of the gate anymore. When they do, it's just good luck.

They also seem to break things that previously were fine a lot more than they used to. And then not acknowledge the problem and not put it right.

it was fixed on big sur in February.

use switchresx for catalina

i don't share your sentiment about apple, i think it generally works pretty well and given all the replacement programmes they also acknowledge issues sooner or later