Mac Studio - performance test

[wonshu]

Hi guys,

for those who have gotten their Mac Studio already:

Over in a German Logic user forum (logicuser.de) we have this little test song that just runs a fast flute line (creates many voices) through 3 Chroma Verb plugins and we check how many of those tracks our machines can handle.

This gives a very comparable performance gauge, because you can run that song on your current machine and see at how many tracks it will choke.

It would be great if people could run this on their Mac Studio.

You can download that Logic Projectfile here:

EDIT: changed the link to the current agreed upon version of this test (thank you djordji post #11):
SPUMD Logic CPU Test

Looking forward to results...

Oh and also: are the fans running crazy loud when running at say 75% CPU performance (check the meter and mute activate the number of tracks that sort of level out at 75%)

PS: On my MacBoorAir with M1 I'm getting 17 tracks. At 18 the machine chokes.

EDIT:
On that new file my results are:
MacOS: Monterey 12.3
Mac Model: MacBookAir
Prozessor: M1 (8-core / 7-core GPU)
RAM: 16 GB
Logic Version: 10.7.3

clean playback: 45 tracks (64 before measure 38) with interal speakers Output Device

-H

PS: For transparency reason:
I had posted an older version of the test song which will result in significantly lower track count (there are 3 reverbs on each channel, so much more CPU intensive) and the first two results on this page were done with this test.
For transparency reasons I'm posting that link again. Please use the other test above, as results for that are comparable across a couple of other forums.

http://www.hanshafner.de/00_static/Logic_CPU_Performance_Test.zip

 

[F-Train]

I haven't installed Logic on my Mac Studio/M1 Max yet, but I'm interested in this test and will try it after installation. Apart from your own result, is there a thread or chart at logicuser.de that has results from various computers?

 

[Jorbanead]

Just tested it out. My Mac Studio can run 31 tracks. It overloads at 32.

My setup:
M1 Max Processor
10-core CPU
32-core GPU
32GB ram
1TB SSD

The fans didn’t ramp up a single time while doing this. I honestly don’t know of any workload that causes the fans to actually ramp up. I think the cooling in this is already so good that it’s idle cooling is good enough for most workloads. Insane.

 

[flossiezoo]

Hi, was curious and had a go too.

M1 Ultra, 20-core CPU, 48-Core GPU, 64GB RAM, 1TB SSD

Got 68 tracks consistently, tested for 15+ minutes. Fans didn't ramp up. Interestingly it doesn't tend to use the efficiency cores for this test. CPU was at c. 70% the whole time. Did manage 70 tracks but it stopped after 10m.

 

[F-Train]

My Mac Studio is one step down from @Jorbanead's (post #3):

Mac Studio M1 Max
CPU: 10-core
GPU: 24-core
Unified Memory: 32GB
SSD: 512GB

I also got 31 tracks. I have my buffer at 256 samples. The only apps I had running in addition to Logic were iMessage and Screen Capture.

 

[Jorbanead]

My Mac Studio is one step down from @Jorbanead's (post #3):

Mac Studio M1 Max
CPU: 10-core
GPU: 24-core
Unified Memory: 32GB
SSD: 512GB

I also got 31 tracks. I have my buffer at 256 samples.

View attachment 1976627

That would make sense as this doesn’t really utilize the GPU at all. I’m very impressed by the performance.

 

[F-Train]

I had to mute track 31 when I decided to open Activity Monitor in a window and take screen captures of Memory and CPU.

I wasn't paying attention to fan noise, but didn't notice any.


Memory:

CPU:

 

[F-Train]

The results above are roughly in line with the amount of unified memory...

MacBook Air, 16GB Memory: 17 tracks
Mac Studio Max, 32GB Memory: 31 tracks
Mac Studio Ultra, 64GB Memory: 68 tracks

 

[wonshu]

Thank you! This looks very promising indeed.

We've been doing this comparison for many years... this is the original thread (in German of course..), at the time we used a song with an ES2 (Logics built in synths) playing a four voice chord, trackcount was going up to and beyond 200 so we decided on this new songfile, but I can't seem to find the thread for it... ?

I'm glad, no one was having fan issues so far, this is my biggest concern at the moment.

 

[wonshu]

I just ran it again on a Hackintosh with an i9 10-core running @ 3.7Ghz 64 GB RAM.

Getting 41 tracks there reliably, at nearly 100% CPU load. 30 tracks sort of run around 75% load.

I'm still amazed that the little fanless MacBookAir manages to do a little less than half that. At a third of the cost not even considering power consumption...

 

[Djordji]

I was the one who setup this test and the one wonshu linked is an early stage of it.
You can find the final version of this test here in german and here in english

Referenc for this test was the following Mac setup

MacOS: Mojave 10.14.6
Mac Model: Mac Pro (Early 2009)
Prozessor: 12 x 3.46 Intel Xeon
RAM: 48 GB
Logic Version: 10.4.8 - 10.5.1


89 clean playback tracks with Loopback Output Device
91 clean playback tracks with Merging VAD Output Device

Use the following audio settings in logic if you want to test and make it comparable:
--------------------------------------
I/O Buffersize: 128 Samples
Processing Threads: Automatic (change this setting to maximum cores too to compare)
Processor Buffer Range: Large
Multithreading: Playback & Live Tracks
Summing: Standard Precision (32 Bit)
Rewire Behavior: Off

---------------------------------------
In the linked english thread I tested it with my M1 MacBook Air 16 GB

MacOS: Big Sur 11.6.1
Mac Model: MacBook Air (2020)
Prozessor: M1 8 core
RAM: 16 GB
Logic Version: 10.7.2
Output Device: Internal Speaker

40 clean playback tracks with Internal Speaker Output Device

By changing the processor threads to 8 in Logics audio settings i had an improvement.
67 clean playback tracks with internal speakers as output device


You can see that the MacBook Air M1 still outputs less in this test than my old Mac Pro Intel even if the M1 Geekbenches much higher.

Download Link SPUMD Logic CPU Test

Stop the playback, duplicate the Alchemy tracks (best in 10s), copy the midi region to the new alchemy tracks, playback again till the cpu drops out. Write the results in this thread providing the following informations: (p.s. close all other applications like safari while testing) make sure the reverbs on the send are all turned on.

---
MacOS: xxx
Mac Model: xxx
Prozessor: xxx
RAM: xxx
Logic Version: xxx

clean playback: xxx tracks with xxx Output Device
---


Thanks for participating.

 

[F-Train]

Hi @Djordji,

When I told your your Logic file to open in Logic, I got this:

 

[Djordji]

just skip it. its about virtual midi ports (IAC)

p.s. my answers have to be approved by an admin so there is a delay ... sorry

 

[wonshu]

Thank you Djordji!

I've edited post #1 to link to that file so that we're as much on the same page as possible.

I'm a bit surprised I had to manually turn on the ChromaVerbs in the bus. That may confuse people... Too many options will confuse the results.

But hey, here we are...

 

[Djordji]

Yeah, you are right. But back then i forgot to turn on the verbs, before i uploaded and linked the test.
**** from the past mudding the present. I guess i still hope it will turn in somekind of feature.

 

[Sharky II]

Must be something weird going on here as my 5,1 is getting 2-3x the numbers of the Studios. I don’t think we’re all doing the same test, and the logic file is quite messy, chromaverbs not enabled etc…

 

[wonshu]

Must be something weird going on here as my 5,1 is getting 2-3x the numbers of the Studios. I don’t think we’re all doing the same test, and the logic file is quite messy, chromaverbs not enabled etc…

Yes.

I had posted an old version of the song that was setup way different and then the original creator of the test came in, so I changed it in the original post.

It really doesn't matter as long as we know which version we used...

Perhaps the people who ran it on their Mac Studios could repost the results with this song file that is comparable across a couple of other forums as well...

 

[heden]

The results above are roughly in line with the amount of unified memory...

MacBook Air, 16GB Memory: 17 tracks
Mac Studio Max, 32GB Memory: 31 tracks
Mac Studio Ultra, 64GB Memory: 68 tracks

Can someone maybe explain how the performance differs when the single core power is more or less the same across all M1's (from M1 to Ultra)? Or is this test based on sample libraries (kontakt etc.)?

 

[F-Train]

Must be something weird going on here as my 5,1 is getting 2-3x the numbers of the Studios. I don’t think we’re all doing the same test, and the logic file is quite messy, chromaverbs not enabled etc…

Yeah, I'm happy to do this if I can simply load a file into Logic without tweaking that I know is identical to what others are loading.

 

[wonshu]

The point of this tr

Can someone maybe explain how the performance differs when the single core power is more or less the same across all M1's (from M1 to Ultra)? Or is this test based on sample libraries (kontakt etc.)?

it's about the number of voices the machine can play. And to make it comparable it's an instrument that comes with Logic so everyone has it.

 

[Jorbanead]

Can someone maybe explain how the performance differs when the single core power is more or less the same across all M1's (from M1 to Ultra)? Or is this test based on sample libraries (kontakt etc.)?

This test is more concerned with multi core performance. That’s why you see doubling of performance with twice as many cores from the Max to the Ultra. My guess is the ultra gets a bit more than a true double due to more ram. It’s a common method to test a CPU and Apple often does a very similar test for their own benchmarks. Since each track can be processed by a thread, it demonstrates multithreaded workloads pretty well.

That being said higher single core performance would also improve results as well.

This test uses stock plugins.

 

[heden]

This test is more concerned with multi core performance. That’s why you see doubling of performance with twice as many cores from the Max to the Ultra. My guess is the ultra gets a bit more than a true double due to more ram. It’s a common method to test a CPU and Apple often does a very similar test for their own benchmarks. Since each track can be processed by a thread, it demonstrates multithreaded workloads pretty well.

That being said higher single core performance would also improve results as well.

This test uses stock plugins.

Very much appreciate you took the time to elaborate a bit on this. Thx!

 

[Horselover Fat]

MacOS: Big Sur 11.6.4
Mac Model: MP 4,1 -> 5,1 dual CPU
Prozessor: 2 x 5690
RAM: 12 GB
Logic Version: 10.5

clean playback: 91 tracks with internal device / 91 tracks via RME Babyface



I know that synthetic benchmarks are not everything, but since this test scrutinizes multi core performance, shouldn't the Mac Studio Max deliver at least something like 150? Given these results, I am hesitant to pull the trigger and more inclined to believe there is something wrong. Either with the test design or that Logic is not native AS (which it is) or something else.

Where's the famous Apple Silicon performance? These are not the results that confirm statements like "The Mac Studio is more power than you'll ever need..." It feels like the audio guys were left out of the equation, because all this fuzz about performance only seems to hold true for video, 3D and photo workflows. But this doesn't make sense either. I'm clueless since I expected a considerable leap in performance.

EDIT: Over at gearspace.com (Link: 596) someone's Studio Ultra delivers impressive results based on a different Logic benchmark test: https://music-prod.com/logic-pro-benchmarks/
There is also a chart that places my computer where I expected it, compared to the new Macs.

 

[wonshu]

It was a mistake to change the songfile. At the time of posting I did not know that there was a newer version of the test. This is very confusing now.

I'll try to make it clearer:
In the first Logic song I had posted every track contained an Alchemy Instrument and 3 Chromaverbs. This testsong will always lead to less tracks as there are many many more Reverbs to be calculated. That's why the number of tracks from the first two people who ran the test are lower than what will be surfacing now.

I decided to change to the newer project file because there are many more results in other forums (logicprohelp.com logicuser.de) that were done and posted using that version.

But once you look at and run the tests yourself you will quickly see what's what and what is comparable to what.

I will repost the link to the "wrong" project file again with a disclaimer in the first post so that people can see for themselves what's going on.

 

[Sharky II]

Maybe just start another thread