ARM updates galore beta round 01 revision 11152

[submerge]

I am happy to confirm that (as a test) expanding the Diva package, updating the Descriptor as Lady Gaia stated and flattening the package again, I was able to install the updated package natively without Rosetta and Diva runs fine in Logic

Thanks both for your time and helpful comments. Every day is a school day!

[Lady Gaia]

Nice to have this experimentally confirmed, @submerge! I see the current option set actually specifies the hostArchitectures set as "i386" which should probably be updated exactly as suggested in my prior post since it looks like these binaries don't contain a 32-bit slice.

[Urs]

Wow, thanks. Yeah, I think our build server is on Catalina atm. I guess we're missing some things to be compatible with Big Sur before we can go there, and build servers are just too crucial an infrastructure to update prematurely.

[Lady Gaia]

Wow, thanks. Yeah, I think our build server is on Catalina atm. I guess we're missing some things to be compatible with Big Sur before we can go there, and build servers are just too crucial an infrastructure to update prematurely.

Completely understandable. I'm sure you can still change the option set to produce the desired result, but it explains why there's an extra hoop to jump through. Do feel free to reach out if there's anything I can do to help - I used to manage part of the development tools group at Apple, but I'm retired and focusing on music these days.

[licasto2]

Performance at 96k is equivalent or better on my 13” M1 MBP compared to my iMac Pro. I can run the same amount of any of plugs (I have all the M1 updated U-he ones) on both systems, but the plug-in Interfaces become unresponsive and the Diva, etc. waveforms don’t refresh on the iMac Pro when it starts to get maxed. On the M1 MBP everything runs as smooth as butter even right up to the edge of being maxed out. It’s kind of amazing the load the thing will take whist remaining silent and cool to the touch. Being able to use hella tape and comp @HQ at 96k is going to be great

[Tronam]

96K? Wow, I haven't ever considered running my DAWs at that high of a sample rate, especially on a laptop. Impressive.

[david.beholder]

It's only twice 48k, not as impressive as 384k

[RyanCo]

Hey All-

I'm going to be doing some comparisons this Sunday with a M1 MBP and 5950x, are there any sort of agreed upon diva tests running? My interest is mostly in low latency/recording performance.

In the "Deep Space Diva" test from the other tread, I can get about 250 voices at the 512 sample buffer, and maybe more impressively 192 voices at the 32 sample buffer (RME HDSPe in TB3 chassis, which should also work with the M1).

Thanks

[pdxindy]

96K? Wow, I haven't ever considered running my DAWs at that high of a sample rate, especially on a laptop. Impressive.

I run at 96khz for Bazille... it sounds better at the higher sample rate.

[Funkybot's Evil Twin]

Hey All-

I'm going to be doing some comparisons this Sunday with a M1 MBP and 5950x, are there any sort of agreed upon diva tests running? My interest is mostly in low latency/recording performance.

In the "Deep Space Diva" test from the other tread, I can get about 250 voices at the 512 sample buffer, and maybe more impressively 192 voices at the 32 sample buffer (RME HDSPe in TB3 chassis, which should also work with the M1).

Thanks

Be sure to also post about the experience with fan noise, heat, and any throttling if any.

[Urs]

Please note that we're squeezing out a few more cycles on Intel and ARM as we go, the final versions may get more voices on either platform.

(a recent refactoring which was meant to improve performance on M1 has actually improved performance on Intel even more... to be expected with the next round of betas)

[martinjuenke]

96K? Wow, I haven't ever considered running my DAWs at that high of a sample rate, especially on a laptop. Impressive.

I run at 96khz for Bazille... it sounds better at the higher sample rate.

Are there "official" recommendations by U-He which sample rate to use?

[pdxindy]

96K? Wow, I haven't ever considered running my DAWs at that high of a sample rate, especially on a laptop. Impressive.

I run at 96khz for Bazille... it sounds better at the higher sample rate.

Are there "official" recommendations by U-He which sample rate to use?

Not that I know of... I do lots of complex modulations at audio rate so 96 makes a difference. For a basic saw bass, not really.

[nilhartman]

96K? Wow, I haven't ever considered running my DAWs at that high of a sample rate, especially on a laptop. Impressive.

I run at 96khz for Bazille... it sounds better at the higher sample rate.

Are there "official" recommendations by U-He which sample rate to use?

Not that I know of... I do lots of complex modulations at audio rate so 96 makes a difference. For a basic saw bass, not really.

+100, exactly. And for that very same reason, I can't help wondering how Bazille would sound with a Ultra HQ mode.

[RyanCo]

Be sure to also post about the experience with fan noise, heat, and any throttling if any.

Not a peep from the MBP, the 5950x is an ac controlled machine room, so it was dead silent too

Happy to report the HDSPe MADI PCIe card in a TB3 enclosure works great on both.

Both impressive machines, 5950x is the clear winner at every buffer size, even in single core performance measured with instances of Pro-R (AS native). Unfortunately the CPU/Thread restriction in the m1 build in Reaper doesn't appear to be working. With both of them easily surpassing 16 voices on one track even in Devine mode at 32 samples, I couldn't figure out a way to do single core tests for Diva.

In the deep space diva test (6 voices per track, great, voices set to 6, 48k), the 5950x can match the M1's 192 voices, AT THE 32 sample buffer! I could only get around 24 voices from Diva on the M1 with the same settings, but I was tinkering with the stack option and something went weird with Diva- so I don't know if that was correct (at some point in there it stopped being able to do one voice...), then my time with the M1 was up. Of course this is multicore, so 16c +SMT vs 4+4...

Can anyone else tell me how many voices of Diva they get with the deep space diva test from the other thread at 32 samples 48k?

Both amazing machines, I think I can replace a 7980xe hack with an M1 when the native code dust settles. The 5950x with the HDSPe is like a farm of HDX cards for native...

One point there, on windows I see a big performance difference at 32 samples between RME and basically everything else. On the M1, a Quantum was pretty close to matching the HDSPe card. I spent like 2 hours getting the dang drivers installed (the allow button wouldn't come up in SIP) so ran out of time to get proper numbers on that.

Amazing era for native audio....