How does Multicore works on U-He Synths? still a mystery
-
- KVRist
- 482 posts since 10 Mar, 2013
-
- KVRAF
- 13224 posts since 16 Feb, 2005 from Kingston, Jamaica
I don't use synths as demonstrated in the video, but I have all those used in the video and use them for their sonic strengths....
using patches/presets.
I think it would be fair to say Sylenth is untouchable for cpu usage, I would give Dune 3 second (Richard is a wiz for low cpu usage/vs sonic quality), but in my use Hive (2) comes in next, then consider Spire next and Serum the heaviest.
That is my experience, qualitatively in actual use.
Like I said, for me a video comparing those is of no use for me cause I would never use any of them as used in the video.
Cubendo Mac and PC, low asio buffer (64 sample size)
rsp
using patches/presets.
I think it would be fair to say Sylenth is untouchable for cpu usage, I would give Dune 3 second (Richard is a wiz for low cpu usage/vs sonic quality), but in my use Hive (2) comes in next, then consider Spire next and Serum the heaviest.
That is my experience, qualitatively in actual use.
Like I said, for me a video comparing those is of no use for me cause I would never use any of them as used in the video.
Cubendo Mac and PC, low asio buffer (64 sample size)
rsp
sound sculptist
-
- KVRist
- 482 posts since 10 Mar, 2013
-
- u-he
- 28065 posts since 8 Aug, 2002 from Berlin
Yes, Dune has always used less CPU than Hive. We never said otherwise.
We will never rewrite Hive‘s 1x unison to work a different way since it would need to sound different to use less CPU.
(Q: Wasn't Dune 3 *always* processing on multiple cores, no matter what? I can't quite remember but I thought this was one fo its features?)
We will never rewrite Hive‘s 1x unison to work a different way since it would need to sound different to use less CPU.
(Q: Wasn't Dune 3 *always* processing on multiple cores, no matter what? I can't quite remember but I thought this was one fo its features?)
-
- u-he
- 28065 posts since 8 Aug, 2002 from Berlin
I did run a little check with the same setting as the first example in the video (copied here from the comments section):
1st example - Hive actually plays 10 oscillators. All Oscs process in multiples of 4, Subs process 1 x unless set to "same". And while you can't hear most of them, they're still processed. That was done because we wanted to keep it open to automate/modulate the filter input switches. Also, it never occurred to us that in real world examples people would be an issue - particularly since on a 2019 MBP each oscillator in this very setting takes up 0.15% CPU (measuring with 44.1kHz). That's right, a single CPU Core could process 600 of these oscillators, at 4x unison. At 16x unison, each oscillator takes 0.32%, measure freshly with our microsecond exact timer we have for debugging. Once you play 5 note chords, the 16x unison goes further down to 0.28% per oscillator. That's less than 3% CPU in total for a 5-note chord with 32 sawtooth voices + 2 sub oscillators. I'm not sure what the definition of CPU friendly is, if that isn't it.
But yeah, I put it on my todo list that for the next larger update, we stop processing oscillators one can not hear, even if for the academic value of such videos.
Now, can we close this?
1st example - Hive actually plays 10 oscillators. All Oscs process in multiples of 4, Subs process 1 x unless set to "same". And while you can't hear most of them, they're still processed. That was done because we wanted to keep it open to automate/modulate the filter input switches. Also, it never occurred to us that in real world examples people would be an issue - particularly since on a 2019 MBP each oscillator in this very setting takes up 0.15% CPU (measuring with 44.1kHz). That's right, a single CPU Core could process 600 of these oscillators, at 4x unison. At 16x unison, each oscillator takes 0.32%, measure freshly with our microsecond exact timer we have for debugging. Once you play 5 note chords, the 16x unison goes further down to 0.28% per oscillator. That's less than 3% CPU in total for a 5-note chord with 32 sawtooth voices + 2 sub oscillators. I'm not sure what the definition of CPU friendly is, if that isn't it.
But yeah, I put it on my todo list that for the next larger update, we stop processing oscillators one can not hear, even if for the academic value of such videos.
Now, can we close this?
-
- Banned
- Topic Starter
- 252 posts since 14 Oct, 2020
In real world Hive is low CPU. People might not seem to not notice, but i did noticed when multiple instances come in to play. Also as Urs mentioned Unison, Serum is really much more demanding when unison is activated where's Hive is not.
But anyways if Urs there is room for optimisation, imo its always welcome.
But anyways if Urs there is room for optimisation, imo its always welcome.
-
- KVRAF
- 25446 posts since 3 Feb, 2005 from in the wilds
I just put an instance of Hive 2.1 into Bitwig... enabled 2 filters, 2 Osc's one to each filter and each with 7 unison voices and played 8 notes and the CPU use is quite low.claudedefaren wrote: ↑Sun Mar 14, 2021 4:31 pm Nice attitude.
To once again entertain your point, if we put literally any of the other vsts in this video at 4 voice unison, Hive still uses way more CPU than all of them. This is, again, demonstrated to an extreme in that video when Dune is set to 16 voices unison and oversampled filter enabled. It uses, again, FAR less CPU than Hive does at 16 voices with any filter enabled.
It's a shame that you don't believe that the use of single voice unison setting on an oscillator isn't something that anyone ever does, since that's obviously not based in reality, because if you took the complaints about Hive's CPU usage seriously it could be an even better product. It's also a shame that you get so defensive about feedback and choose to act so flippant towards your customers over it.
Now I just took that same preset, made a clip with 10 notes, then made duplicates of the track until there were 30 tracks total, each with 10 notes all playing at the same time (300 simultaneous notes) with plenty CPU to spare on my MBP.
Out of curiosity, I made it 60 tracks, so 600 simultaneous notes and now the CPU meter is pretty full, but no crackles.
I would say calling Hive CPU friendly is an accurate statement. I have a bunch of synths that use significantly more CPU.
-
- u-he
- 28065 posts since 8 Aug, 2002 from Berlin
In Logic, oddly, there's some kind of dual CPU meter. One part shows "thread activity" while another shows the actual usage of the CPU cores processing. Maybe Live does the same and in Hive 2.0.1 (rev 9709) - the version used for this video, which is *not* Hive 2.1 - we still had a few more threads? Maybe the issue is related to the UI struggles which we fixed in 2.1? I don't know. But in Logic I see nothing out of the ordinary. Hive has some "base CPU" which might be higher than some of these other synths, but once you crank up unison, the CPU usage barely slips.
-
- u-he
- 28065 posts since 8 Aug, 2002 from Berlin
You mean like, I should be eternally grateful to someone for accusing us of false advertisement? I think you're in the wrong film if you expect my gratitude.
-
- KVRist
- 482 posts since 10 Mar, 2013
-
- KVRAF
- 25446 posts since 3 Feb, 2005 from in the wilds
You claimed u-he was falsely advertising Hive as CPU friendly... I just proved you wrong. 600 simultaneous voices is plenty friendly (I could have added more and there are much faster computers than mine). I will never have an actual project with anywhere near that many concurrent notes.claudedefaren wrote: ↑Sun Mar 14, 2021 6:26 pmHey I've done that too! And I also did it with a handful of these other synths too, same settings, to compare. Hive lost the comparison again.
At that point CPU ceases to be a concern except of course for pedantic argument.
-
- KVRist
- 482 posts since 10 Mar, 2013
-
- KVRAF
- 25446 posts since 3 Feb, 2005 from in the wilds
My mistake... I thought that video was yours. Still, you posted it and the video claims that so you kind did too, even if inadvertently.
-
- KVRAF
- 25446 posts since 3 Feb, 2005 from in the wilds
I can make lots of tracks with Hive and not run out of CPU. So no, I don't believe there is a real world issue or problem. To me, that makes it pedantic unless your computer setup is some sort of outlier and you are having a specific problem that others aren't.claudedefaren wrote: ↑Sun Mar 14, 2021 7:28 pmI'm not being pedantic, I'm talking real world usage here. Hive is less CPU efficient than a large amount of its competitors, full stop.
You are also using manipulative language. Like saying thanks to Urs for 'fixing' something that isn't broke in the first place. It is working as intended.
And saying it is less CPU efficient than a 'large amount of its competitors' is also deceptive language when you are going by an outdated video using only a single use case and then declaring some definitive statement (that does not stand up to scrutiny).
I just tested Dune 3 with 2 osc's each with a wavetable and 7 voice osc unison and an LFO sweeping the wavetables and 2 filters with no FX. I created the same setup in Hive with 2 Osc's each with a wavetable being swept going into 2 filters and no FX. I made 30 tracks with Dune and got CPU crackles. I made 30 tracks with Hive and no CPU crackles. So in this example, Hive is using less CPU than Dune 3 for a similar preset.
In past tests, Serum does not use less CPU than Hive but I don't have it installed right now to duplicate this specific test. I just checked Vital and it definitely uses more CPU than Hive.
So there are 2 currently popular wavetable synths that are 'competitors' with Hive and neither one is using less CPU in the specific test I just made, they are both using more.
So contrary to your assertions, nothing is wrong with Hive, it isn't broken or having some issue. It is a well crafted synth that provides fast workflow, wide sonic diversity, excellent sound quality and is cpu friendly.
-
- KVRist
- 482 posts since 10 Mar, 2013
