Hive NOT that light ?
- KVRist
- Topic Starter
- 162 posts since 14 Dec, 2013
I made quick comparison of hive, dune2 and spire.
These are performance graphs of each synth playing 4 note chord with 4 oscillators set to x9 detuned saw wave.
Dune2:
Spire:
Hive:
So the results of average load are:
1. Dune 0.920 ms
2. Spire 2.117 ms
3. Hive 2.701 ms
So here is my question: is this real CPU usage of Hive or maybe Iam having some technical issues ?
Hive's cpu usage is its main advert slogan so I find it weird that Dune2 is over 3 times lighter......
These are performance graphs of each synth playing 4 note chord with 4 oscillators set to x9 detuned saw wave.
Dune2:
Spire:
Hive:
So the results of average load are:
1. Dune 0.920 ms
2. Spire 2.117 ms
3. Hive 2.701 ms
So here is my question: is this real CPU usage of Hive or maybe Iam having some technical issues ?
Hive's cpu usage is its main advert slogan so I find it weird that Dune2 is over 3 times lighter......
- u-he
- 28062 posts since 8 Aug, 2002 from Berlin
What is the "DSP Performance Graph" taken from? Is it built inside a DAW? If so, which DAW...? Were any FX turned on? Which engine was used? Were both filters set to bypass? - It's really hard to imagine how these figures come about without such details.
That said, Hive's core oscillators have always been slower than Dune2's, mostly because Hive's oscillators have better anti-aliasing from what I figured. There were compromises that we wouldn't want to make for sake of CPU. However, we managed to put non-linear, oversampled zero delay feedback filters in there and much better FX than we've seen elsewhere. So yeah, we market Hive as low CPU because we think it's the best CPU for sound quality you can possibly get
That said, Hive's core oscillators have always been slower than Dune2's, mostly because Hive's oscillators have better anti-aliasing from what I figured. There were compromises that we wouldn't want to make for sake of CPU. However, we managed to put non-linear, oversampled zero delay feedback filters in there and much better FX than we've seen elsewhere. So yeah, we market Hive as low CPU because we think it's the best CPU for sound quality you can possibly get
- KVRist
- Topic Starter
- 162 posts since 14 Dec, 2013
@Urs
You can find DSP graph in Bitwig inside "View" menu.
All three synths were set from init patch.FX were off. Setting all fx ON in hive doubles the average load. Changing synth engine from normal to clean makes it about 2.300 ms
I love Hive, I know it sounds greate but I was just wondering if I can make it less cpu hungry.
ps: I did my test on i5-480M with assio4all v2 set to 2048 samples
You can find DSP graph in Bitwig inside "View" menu.
All three synths were set from init patch.FX were off. Setting all fx ON in hive doubles the average load. Changing synth engine from normal to clean makes it about 2.300 ms
I love Hive, I know it sounds greate but I was just wondering if I can make it less cpu hungry.
ps: I did my test on i5-480M with assio4all v2 set to 2048 samples
- u-he
- 28062 posts since 8 Aug, 2002 from Berlin
Hmmm, let's see:
Pulsewave takes up more CPU than other waveforms.
For Osc unison, chose 4, 8, 12 or 16 for optimum results. If you can, do 16x unison in a main oscillator rather than 2 times 8x in Osc and Sub. Likewise, if you can get away with one oscillator, switch the other oscillator off (solo buttons).
Clean engine has fastest filters.
Setting both filters to same mode may be faster than having two different modes.
Buffersize makes hardly any difference. Hive should use pretty much the same CPU with 128 samples as with 512 or 2048. However, 64 or may take up a tad more while below 64 may be notably more CPU intensive.
Try to avoid opening too many plug-in editor windows. I guess that's a general rule for anything that has something blinking in accordance to the audio (In Hive that's MIDI Led, Output Meter, Sequencer progression, Compression meter etc.)
That's all I can think of for now...
What's weird are those intermittent spikes. I might fire Bitwig up and try to figure out what's happening there. Or are these Note On events?
Pulsewave takes up more CPU than other waveforms.
For Osc unison, chose 4, 8, 12 or 16 for optimum results. If you can, do 16x unison in a main oscillator rather than 2 times 8x in Osc and Sub. Likewise, if you can get away with one oscillator, switch the other oscillator off (solo buttons).
Clean engine has fastest filters.
Setting both filters to same mode may be faster than having two different modes.
Buffersize makes hardly any difference. Hive should use pretty much the same CPU with 128 samples as with 512 or 2048. However, 64 or may take up a tad more while below 64 may be notably more CPU intensive.
Try to avoid opening too many plug-in editor windows. I guess that's a general rule for anything that has something blinking in accordance to the audio (In Hive that's MIDI Led, Output Meter, Sequencer progression, Compression meter etc.)
That's all I can think of for now...
What's weird are those intermittent spikes. I might fire Bitwig up and try to figure out what's happening there. Or are these Note On events?
- KVRist
- Topic Starter
- 162 posts since 14 Dec, 2013
Its 2 bar chord looping. Spike apears everytime loop starts.Urs wrote:What's weird are those intermittent spikes. I might fire Bitwig up and try to figure out what's happening there. Or are these Note On events?
-
- Banned
- 511 posts since 5 Jul, 2014
Curious about this, Urs. Wouldn't buffersize increases, decrease CPU use? Isn't that the purpose of increasing ones buffersize? So plugins will use less CPU and run with less audio glitches? Is there something particular to Hive that makes it this way, or is it some sort of common misunderstanding about how buffersizes work?Urs wrote: Buffersize makes hardly any difference. Hive should use pretty much the same CPU with 128 samples as with 512 or 2048. However, 64 or may take up a tad more while below 64 may be notably more CPU intensive.
- KVRAF
- 23101 posts since 7 Jan, 2009 from Croatia
At some point the benefits for CPU disappear. In reality I don't think anything above 512/768 really matters to today's CPUs. But as you go below 128, it becomes much more straining to CPUs.
- u-he
- 28062 posts since 8 Aug, 2002 from Berlin
Our stuff internally runs at slices of 64 samples. This allows for very good optimization of complex processes, i.e. we keep a large number of temporary buffers small.KingTuck wrote:Curious about this, Urs. Wouldn't buffersize increases, decrease CPU use? Isn't that the purpose of increasing ones buffersize? So plugins will use less CPU and run with less audio glitches? Is there something particular to Hive that makes it this way, or is it some sort of common misunderstanding about how buffersizes work?Urs wrote: Buffersize makes hardly any difference. Hive should use pretty much the same CPU with 128 samples as with 512 or 2048. However, 64 or may take up a tad more while below 64 may be notably more CPU intensive.
OTOH if you have simple processes that don't require many temp buffers (or any temp buffers at all), it's good to run as big chunks as possible.
There's always a tradeoff of memory requirements versus size/number of process loops. I've done some experiments with running our stuff at 128 and 256 sample slices internally, but apart from worsening the control rate of modulation matrices, it didn't much good. Hence I think 64 samples is pretty good in terms of CPU vs. algorithmic flexibility.
- U
- KVRAF
- 3053 posts since 25 Apr, 2011
Also, dune2 runs at the worst setting by default. I wonder if the quality (e.g. aliasing) is worse in dune2 compared to hive at the default. Usage is one thing, but what about the other side; the "quality".
-
- KVRer
- 3 posts since 10 Oct, 2012
I am running an XPS13 9360 (i7 7500U dual core 2.9Ghz) with Bitwig and Hive.... and had similar problems. I `downgraded` to this small touch screen laptop and Hive was kind of part of the strategy . Troubleshooting:
1) I tried using the solo buttons (and filter input buttons) to reduce the DSP load as suggested by Urs but this had no effect.
2) If the sub has the same waveform ("=") then the unison applies to the sub.... doubling the number of voices. And the INIT patch has the "=" waveform.
3) The init patch also has polyphony of 8 voices... Im guessing this means for a new 4 note chord immediately after an old 4 note chord with a tail I have 2 x 4 x unison x 2 voices (if sub included)
On the hardware side in the BIOS I removed the Intel Speedstep and ensured multicore was enabled. This seemed to help also.
Thanks for the help Urs. Would be nice if we could turn off Osc2/Filt2 as suggested?
1) I tried using the solo buttons (and filter input buttons) to reduce the DSP load as suggested by Urs but this had no effect.
2) If the sub has the same waveform ("=") then the unison applies to the sub.... doubling the number of voices. And the INIT patch has the "=" waveform.
3) The init patch also has polyphony of 8 voices... Im guessing this means for a new 4 note chord immediately after an old 4 note chord with a tail I have 2 x 4 x unison x 2 voices (if sub included)
On the hardware side in the BIOS I removed the Intel Speedstep and ensured multicore was enabled. This seemed to help also.
Thanks for the help Urs. Would be nice if we could turn off Osc2/Filt2 as suggested?
- Banned
- 6129 posts since 9 Oct, 2007 from an inharmonious society
Dune 2 has a multi-core switch.
When it's off it uses way more cpu percentage.
Hive doesn't seem to have that function, so the comparison doesn't say much to me.
I use both, and do keep my eyes on cpu usage quite a bit.
They both use about equal cpu percentage for the majority of sounds.
The Spire demo I tried, used way more cpu than either Dune2 or Hive, and Spire's cpu was all over the place.
When it's off it uses way more cpu percentage.
Hive doesn't seem to have that function, so the comparison doesn't say much to me.
I use both, and do keep my eyes on cpu usage quite a bit.
They both use about equal cpu percentage for the majority of sounds.
The Spire demo I tried, used way more cpu than either Dune2 or Hive, and Spire's cpu was all over the place.
-
Perimeter Sound Perimeter Sound https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=162713
- KVRAF
- 1744 posts since 11 Oct, 2007
Hive is not a cpu *free* synth ha-ha, but I've yet to get the quality of sound it can do for the low cpu usage from much other than well, older & far less capable synths, and equally amazing ones such as Zebra (duh ha-ha), and Dune 2. My opinion/experience is of course non-scientific, but come on, any great sounding synth that barely spikes the cpu ? What year is it ? What century is it that we're complaining about such minor differences in computing power needed ?