Diva with Rosetta2 on Apple M1
- KVRAF
- 2120 posts since 10 Apr, 2002 from Saint Germain en Laye, France
-
- KVRist
- 49 posts since 14 Mar, 2010
On the MacBook Air it says up to 8 GPU cores. The cheapest Air only has 7 GPU cores active whilst the more expensive one has all 8 cores active.
The main issue is thermal throttling on the Air, as it does not have a fan.
For comparison of the M1 Air and M1 Pro see https://www.youtube.com/watch?v=CmMOJTs7Pu8
At 8:26 he runs Cinebench which is CPU intensive and the Air is then throttling whilst the Pro seems to not be throttling. In his comparison the Air gets a score of about 1,000 less in multi-threading.
At 13:49 he runs a Logic test where he doesn't fully complete the runs and instead mentions tests he's run on previous videos. The Air did not handle as many tracks as the Pro in this particular video comparison, but in previous tests they performed just about the same. I think that is because previously he didn't run the Logic tests long enough to cause the Air to throttle whilst in this video the system may already be on the border of throttling due to earlier tests. But that is just my guess.
So, if your sessions do not use the CPU enough to cause the Air to heat up enough so it has to throttle, you are unlikely to see any difference between the Air and the Pro. But if your session is busy enough to cause throttling then the Pro will perform better.
The main issue is thermal throttling on the Air, as it does not have a fan.
For comparison of the M1 Air and M1 Pro see https://www.youtube.com/watch?v=CmMOJTs7Pu8
At 8:26 he runs Cinebench which is CPU intensive and the Air is then throttling whilst the Pro seems to not be throttling. In his comparison the Air gets a score of about 1,000 less in multi-threading.
At 13:49 he runs a Logic test where he doesn't fully complete the runs and instead mentions tests he's run on previous videos. The Air did not handle as many tracks as the Pro in this particular video comparison, but in previous tests they performed just about the same. I think that is because previously he didn't run the Logic tests long enough to cause the Air to throttle whilst in this video the system may already be on the border of throttling due to earlier tests. But that is just my guess.
So, if your sessions do not use the CPU enough to cause the Air to heat up enough so it has to throttle, you are unlikely to see any difference between the Air and the Pro. But if your session is busy enough to cause throttling then the Pro will perform better.
- KVRAF
- 2032 posts since 15 Mar, 2002 from Seattle, WA - USA
The potential to thermally throttle is certainly true, but in my experience so far with the M1 Air (8CPU/8GPU core model) and MBP I haven’t seen it in the context of typical music production work. They’ve been roughly the same for me. I’ve had a hard time even getting the MBP to spin up its fan at all, at least in any way I can hear and I’ve placed my ear right up against the chassis. I’ve only gotten it to audibly spin up after 20-30min of World of Warcraft with resolution and graphics features set high. Since audio work doesn’t tend to max out the GPU cores and much of music work is more bursty with lots of stops and starts, I haven’t felt a difference on the Air. Then again, my projects don’t typically sit there idling in the 70+% range.
If I was doing work that requires a lot of 100% utilization encoding, video export, rendering etc... or needed guaranteed performance for mission critical situations like performing on stage or something, then the MBP is the safer bet. But, who would lean on such a new system architecture for anything ‘mission critical’ right now anyway? Considering the M1 Air can already play all of my past projects without issue, outperforming my 2017 15” MacBook Pro, I’d say it’s a plenty viable option for many users.
If I was doing work that requires a lot of 100% utilization encoding, video export, rendering etc... or needed guaranteed performance for mission critical situations like performing on stage or something, then the MBP is the safer bet. But, who would lean on such a new system architecture for anything ‘mission critical’ right now anyway? Considering the M1 Air can already play all of my past projects without issue, outperforming my 2017 15” MacBook Pro, I’d say it’s a plenty viable option for many users.
-
- KVRAF
- 2111 posts since 25 Jun, 2008 from Montreal, Canada
How does this compare to recent desktop CPUs like new i9-10xxxK or r7-5800x?
- KVRAF
- 2032 posts since 15 Mar, 2002 from Seattle, WA - USA
It’s still hard to say until more software is M1 native, but I bet they’re a good bit better in multicore. I have the 8c/16t i9-9900K and according to Geekbench M1 is over 35% faster in single core, but 10% slower in multicore. I don’t know how trustworthy that is for audio though. The i9 can still run nearly 15-20% more tracks, at least in the new Logic benchmark everyone is using these days, so it does seem to prefer better multicore performance.xx JPRacer xx wrote: ↑Fri Jan 08, 2021 12:07 pm How does this compare to recent desktop CPUs like new i9-10xxxK or r7-5800x?
-
- KVRist
- 448 posts since 18 Nov, 2002
I have been following with interest the results that are being published of the new Apple cpu's processing audio and this is the most impressive result that I think I have read about the processing capacity of this new CPU.tasmaniandevil wrote: ↑Thu Dec 03, 2020 2:48 pm
Results:
Multicore option in Diva disabled:
On the 2015 Intel MBP, 3 instances with 18 voices ran without glitches. The fourth instance already produced too many dropouts to continue working.
On the 2020 M1 MBP, 20 instances with 120 voices ran glitch free. With one more instance and 126 voices, there were some dropouts, and it finally glitched heavily when adding instance 22.
Our jaws dropped collectively at that moment.
With Diva's multicore option enabled in each instance:
On the 2015 Intel MBP, six instances with 36 voices ran glitch free, a seventh instance produced occasional dropouts, and the eighth instance brought it to its knees.
On the 2020 M1 MBP, only five instances with 30 voices played without glitches, the sixth instance already produced heavy crackling.
Update December 8:
We now have a first version running natively on the M1 (i.e. without using Rosetta2).
With this version, we ran the same test project again, and could now run up to 32 instances of Diva with 192 voices without dropouts.
x7 under Rosseta and x10 in native mode, is absolutely amazing.
These tests are done at 512 samples, how is the performance at 64 samples or other low latencies? because if what I have seen in other tests is confirmed where the results at low latency (even with a 32 sample buffer) were practically equal to that of a high 512 sample buffer, then the performance difference would be even greater.
If this is going to be the performance with the most process demanding audio plugins, then I will have to buy a Mac after 25 years of my la
-
tasmaniandevil tasmaniandevil https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=62450
- KVRAF
- Topic Starter
- 1735 posts since 22 Mar, 2005 from a planet called u-he
Please be aware that the difference is so huge, because the only computer we had to compare it to was a Macbook Pro from 2015, that's already quite old.
We haven't done excessive tests at lower buffer sizes yet, but since lower buffer sizes are more demanding, the performance will likely be decreased.
We haven't done excessive tests at lower buffer sizes yet, but since lower buffer sizes are more demanding, the performance will likely be decreased.
That QA guy from planet u-he.
-
- KVRist
- 448 posts since 18 Nov, 2002
I think the evolution in single thread performance has not been very significant in the latest 5 years. Also I have read reports where Diva is around 6-7 instances in relative modern desktop cpu, like i7 at 3.5 ghz, so 32 instances is equally stagering.
The reports about low latency in other sources are also very surprising, thats why I was asking.
Hope we can have some reports about low latency when you have it.
Thank you.
The reports about low latency in other sources are also very surprising, thats why I was asking.
Hope we can have some reports about low latency when you have it.
Thank you.
-
- KVRAF
- 2111 posts since 25 Jun, 2008 from Montreal, Canada
I have an old i5-3450 CPU from 2011 and on Windows 10 in Bitwig with a cheap external Sound Blaster gaming sound card I can do 11 instances of Diva with the mentioned test @ 48 Khz, 480 samples.
I'm sure a modern desktop i7, i9 or Ryzen 5 or 7 would do A LOT better. When I read about an i7 doing 4-5 instances I had a WTF moment just to realize that was on a laptop. I just can get my head around people making music on laptops.
I'm sure a modern desktop i7, i9 or Ryzen 5 or 7 would do A LOT better. When I read about an i7 doing 4-5 instances I had a WTF moment just to realize that was on a laptop. I just can get my head around people making music on laptops.
- KVRAF
- 13223 posts since 16 Feb, 2005 from Kingston, Jamaica
Depends on the quality setting in Diva.
My PC desktop is in every way spec'ed better than my macbook pro. Yet their performances aren't vastly different.
rsp
My PC desktop is in every way spec'ed better than my macbook pro. Yet their performances aren't vastly different.
rsp
sound sculptist
-
- KVRAF
- 2111 posts since 25 Jun, 2008 from Montreal, Canada
I did the test like mentioned on the 1st page, quality: great.
-
- KVRist
- 448 posts since 18 Nov, 2002
Had the oportunity to try this "Diva test" in Reaper with a i7 7820 8 cores at 4ghz (all the cores were active in the win10 performance meter) and with 10 Diva instances some crackling start to sound when changing the 6 notes chord, if I let the chord sound forever (not changing any note) I can reach 13 instances, but i thing its not how the test was done?.xx JPRacer xx wrote: ↑Tue Jan 19, 2021 12:52 pm I have an old i5-3450 CPU from 2011 and on Windows 10 in Bitwig with a cheap external Sound Blaster gaming sound card I can do 11 instances of Diva with the mentioned test @ 48 Khz, 480 samples.
I'm sure a modern desktop i7, i9 or Ryzen 5 or 7 would do A LOT better. When I read about an i7 doing 4-5 instances I had a WTF moment just to realize that was on a laptop. I just can get my head around people making music on laptops.
So 10/13 instances in a 8 core 7820x @ 4ghz vs 32 instances in the M1, still surprissing this M1 Diva performance.
-
- KVRAF
- 2111 posts since 25 Jun, 2008 from Montreal, Canada
^ Damn! I'll try the test in Reaper but from what you wrote Bitwig seems to really do an awesome job with my CPU! I was expecting a CPU like yours to be able to support AT LEAST twice more instances since you have twice the cores. (+ more GHz)
Quick question: Did you enabled multi-core option in Diva? Because if I do my performance drop to 5 instances max. For my system the performance is better with the option off.
Quick question: Did you enabled multi-core option in Diva? Because if I do my performance drop to 5 instances max. For my system the performance is better with the option off.
-
midi_transmission midi_transmission https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=298730
- KVRian
- 989 posts since 13 Feb, 2013
Thanks for testing. Yes, it's clear that the m1 is super strong and probably the best cpu design currently. I maybe wrong, but I find the huge difference in this test hard to belief and not similar to other benchmark results I saw. It's not about questioning that the m1 might be faster, just a check of the order of magnitude it's faster.krraqk wrote: ↑Wed Jan 20, 2021 1:50 amHad the oportunity to try this "Diva test" in Reaper with a i7 7820 8 cores at 4ghz (all the cores were active in the win10 performance meter) and with 10 Diva instances some crackling start to sound when changing the 6 notes chord, if I let the chord sound forever (not changing any note) I can reach 13 instances, but i thing its not how the test was done?.xx JPRacer xx wrote: ↑Tue Jan 19, 2021 12:52 pm I have an old i5-3450 CPU from 2011 and on Windows 10 in Bitwig with a cheap external Sound Blaster gaming sound card I can do 11 instances of Diva with the mentioned test @ 48 Khz, 480 samples.
I'm sure a modern desktop i7, i9 or Ryzen 5 or 7 would do A LOT better. When I read about an i7 doing 4-5 instances I had a WTF moment just to realize that was on a laptop. I just can get my head around people making music on laptops.
So 10/13 instances in a 8 core 7820x @ 4ghz vs 32 instances in the M1, still surprissing this M1 Diva performance.
Not saying that the u-he test is wrong, just that there might be something not right with the test or something we understand wrongly.
Last edited by midi_transmission on Wed Jan 20, 2021 6:22 pm, edited 2 times in total.