KVR Audio

Sascha Franck · Post by **Sascha Franck** » Thu Nov 19, 2020 2:35 pm

Correct, Logic is using a higher buffer on playback tracks (1024 samples fixed, IIRC).
As far as overall performance of the M1 machines goes, both the Mini and MBP will perform slightly better than my 2x2.66 MP (which, in fact, is quite incredible as it's a true workhorse) in terms of multithread performance, the Air will come very close - and all three of them will completely slaughter this trusty cheesgrater when it comes to singlecore tasks - like being 3-4 times as juicy.
And as I never need all the multithreading power, one of these will be *the* perfect fit for me (minus the poor connectivity...).
Now all that needs to happen is Corona to f*** off.

Ploki · Post by **Ploki** » Thu Nov 19, 2020 2:54 pm

ericj23 wrote: Thu Nov 19, 2020 10:02 am Doesn't Logic background render tracks so that the current track is the only one that is "live"? Unless this has been turned off these test are not measuring computational power for processing VST but ability to stream audio files. While that is a useful metric it isn't really the one I want to know about!

no, it doesn't. Unless you use freeze function

Ploki · Post by **Ploki** » Thu Nov 19, 2020 2:58 pm

Etienne1973 wrote: Wed Nov 18, 2020 7:01 pm iMac 27 2020 i9 3,6Ghz 10-core reaches 317 Alchemy tracks.

Mac Mini 2020 M1 8-core reaches 126 Alchemy tracks.

I'm close to tears. Straight disappointment this real-world comparison.

You think?
it's an entry level Mini designed to replace the i3.
i9-core has much more cooling and it has effectively 6 more cores (forget about the high-efficiency cores here). Also costs 4 times as much.

Every bench so far i saw i think is impressive af. Have you seen the teardown? There's basically nothing in there. They can put three times as much computer inside that Mini enclosure.

Markus Krause · Post by **Markus Krause** » Thu Nov 19, 2020 3:17 pm

Here are my benchmark results and a lot more info:

viewtopic.php?f=16&t=555863

The CPU is faster than my Intel I3, but slower than my Ryzen 3800. For audio processing I'd say it is comparable to a strong Intel I7, which has 4 cores (+4 virtual ones for hyperthreading).

Ploki · Post by **Ploki** » Thu Nov 19, 2020 3:25 pm

Thanks!

zvenx · Post by **zvenx** » Thu Nov 19, 2020 3:28 pm

Thanks. I dont' see benchmark results though.
Do you Ploki?
rsp

Ploki · Post by **Ploki** » Thu Nov 19, 2020 3:33 pm

no, it just seems to be a general rundown of everything.
What impressed me the most so far is this guy from GS:

gs guy wrote:EDIT: I just performed the test with the "NewLogicBenchmarkTest" file that is elsewhere on Gearslutz (the one with Sculpture instead of the FM synth), and the scores, under similar buffer settings, were 70 for my iMac, and 99 tracks for the MacBook Air running native ARM. The iMac's fan really got going on this test, and the score on the iMac would lower when the fan was working harder, but I couldn't get the MacBook Air to throttle back. And after several repetitions, it was only warm to the touch.

My mac mini get's about 80 on that benchmark.

It's a good indication of what performance CAN BE when it's compiled for ARM and optimised. And it's absolutely crazy - 4-core i7's get about 40-45 track tops in that bench.

Markus Krause · Post by **Markus Krause** » Thu Nov 19, 2020 3:38 pm

Sorry. I did not have the time to do videos or detailed statistics. I need to take care of porting my stuff to M1.
The measurements were done with ElectraX 2.8, Icarus 2.0 and Saurus 2.6 (64 bit) running in Reaper under Rosetta2. All plugins do not use AVX. The code uses integer, floating-point and SSE. It should be rather representative for an average plugin.
Around 70 instances can be loaded at the same time.

The performance drop from Rosetta compared to native M1 seems to be around 30% for audio stuff.

Ploki · Post by **Ploki** » Thu Nov 19, 2020 3:40 pm

Markus Krause wrote: Thu Nov 19, 2020 3:38 pm Sorry. I did not have the time to do videos or detailed statistics. I need to take care of porting my stuff to M1.
The measurements were done with ElectraX 2.8, Icarus 2.0 and Saurus 2.6 (64 bit) running in Reaper under Rosetta2. All plugins do not use AVX. The code uses integer, floating-point and SSE. It should be rather representative for an average plugin.

The performance drop from Rosetta compared to native M1 seems to be around 30% for audio stuff.

that's actually exactly what that guy got with NewLogicBenchmark test. 99 tracks via native, and ~70 via Rosetta

mind you, my mini i7 struggles with 80 tracks in that benchmark, i get 82-83 on a good day. So a passively cooled MBA getting 99 is pretty crazy.

Markus Krause · Post by **Markus Krause** » Thu Nov 19, 2020 3:41 pm

Rosetta works pretty well. The Apple guys did a good job here.
Note: I did compare towards my Intel I7 on PC. The Intel I3 is in a Mac mini.

Daags · Post by **Daags** » Thu Nov 19, 2020 3:41 pm

Tim Cook be like ... "oh, you want more RAM ?"

Ploki · Post by **Ploki** » Thu Nov 19, 2020 3:44 pm

Markus Krause wrote: Thu Nov 19, 2020 3:41 pm Rosetta works pretty well. The Apple guys did a good job here.
Note: I did compare towards my Intel I7 on PC. The Intel I3 is in a Mac mini.

tbh passively cooled MBA getting twice as many tracks over Rosetta, as my 2018 13" macbook pro gets natively, is crazy on its own right lol

i'm seriously impressed - if they scale this, it's gonna be crazy

Daags wrote: Thu Nov 19, 2020 3:41 pm Tim Cook be like ... "oh, you want more RAM ?"

how is this in ANY way relevant lol?

Daags · Post by **Daags** » Thu Nov 19, 2020 3:58 pm

ya you're right ... with unified memory, being locked in to Apple's RAM prices at the point of purchase isn't relevant. €227 for a Mac Mini 8gb RAM upgrade (8gb ->16gb) isn't relevant. In contrast, the fact I got a 64gb RAM upgrade for my laptop for €240 isn't relevant either.

because in the walled garden that is Apple, money is irrelevant.

GOTCHA.

Ploki · Post by **Ploki** » Thu Nov 19, 2020 4:12 pm

Daags wrote: Thu Nov 19, 2020 3:58 pm ya you're right ... with unified memory, being locked in to Apple's RAM prices at the point of purchase isn't relevant. €227 for a Mac Mini 8gb RAM upgrade (8gb ->16gb) isn't relevant. In contrast, the fact I got a 64gb RAM upgrade for my laptop for €240 isn't relevant either.

because in the walled garden that is Apple, money is irrelevant.

GOTCHA.

Apple locking memory is nothing new, it's been going on since 2012. We had luck with Mini '18... So you're late for the party. Mini 2014 had locked memory, and all macbooks since 2012 had locked memory.
Why is it such a surprise that entry-level Macs have locked memory?

And secondly, comparing unified memory with memory sticks 1:1 is just pointless.

So no, your comment isn't relevant at all

zvenx · Post by **zvenx** » Thu Nov 19, 2020 4:30 pm

Markus Krause wrote: Thu Nov 19, 2020 3:38 pm Sorry. I did not have the time to do videos or detailed statistics. I need to take care of porting my stuff to M1.
The measurements were done with ElectraX 2.8, Icarus 2.0 and Saurus 2.6 (64 bit) running in Reaper under Rosetta2. All plugins do not use AVX. The code uses integer, floating-point and SSE. It should be rather representative for an average plugin.
Around 70 instances can be loaded at the same time.

The performance drop from Rosetta compared to native M1 seems to be around 30% for audio stuff.

Thanks Markus.
Appreciated.
Rsp

Apple announces new Mac Mini, Air + 13" MBP featuring their own M1 chip.