Time for the true. NI Kompakt against Steinberg Halion.

[MacTron]

so, how's the final result
what are their true numbers of polypony?

The test show me a thing first and later the opposite. So there isn't a truly test  ...

[VMS]

so, how's the final result
what are their true numbers of polypony?

The test show me a thing first and later the opposite. So there isn't a truly test  ...

do you mean that halion couldn't reach 76 voices in the 2nd test?

[VMS]

ok I have just tested kompakt & halion2 on a windows 10 laptop (cpu i3 2348m)
their performance are quite close
the main difference is how they handle pitch-transposition

setting:
results are the average of 20 takes
no. of voices in effect = 61
filters are bypassed
volume env. are on with long hold long sus.
a 5mins 44/16 sample is used which covers 61 keys
direct from disc
GUI are visible

cpu usage result 1  [non pitch tracking (will not transpose from the root key)]

kompakt = 10-11%
halion2 (good quality+ good resampling quality) = 12%
halion2 (mid quality+ good resampling quality) = 10-11%

cpu usage result 2  [pitch tracking (will transpose from the root key)]

kompakt = 12-13%
halion2 (good quality+ good resampling quality) = 11-12%
halion2 (mid quality+ good resampling quality) = 10%

[Protools5LEGuy]

Thanks for your Windows 10 benchmark, but we always are talking about performance on Mac OS 9, because x86 and PowerPC software depends on optimizations, mostly Altivec (G4) vs Pentium IV (SSE) related.

[VMS]

Thanks for your Windows 10 benchmark, but we always are talking about performance on Mac OS 9, because x86 and PowerPC software depends on optimizations, mostly Altivec (G4) vs Pentium IV (SSE) related.

yea I know but I left my powerbook at home
and I wont go home until xmas
anyway the results should be consistent with os9
ie: kompakt needs more cpu power for pitch-transposition

[IIO]

how much CPU is required for the antialiasing of transposed samples might eventually depend on the version of the product. i´d doubt that the OS9 kontakt is doing it at all.

[VMS]

how much CPU is required for the antialiasing of transposed samples might eventually depend on the version of the product. i´d doubt that the OS9 kontakt is doing it at all.

um...we are talking about kompakt, not kontakt
I don't think those lazy NI stuff had updated anything regarding antialiasing for kompakt
here's the changelogs

KOMPAKT 1.0.8
Improvements
Overload detection at 95 % now cuts 10 % of played voices (oldest first)
Fixes
Load menu not sticky in "Remove" section (Mac only)
About screen source problem (Mac only)
Tiger AU validation issue (Mac only)
More than 2 GB RAM stopped Kompakt from loading in Cubase SX 3. (PC only)
Offline bounce dropouts
Hanging notes in logic 5.5.1. (PC only)
KOMPAKT 1.0.3
Improvements
DFD performance has been vastly improved.
The latest Kompakt update contains a new button in the options menu: "Lock Memory (No Swap)". After running the updater, this option will be DISABLED by default, which should result in faster loading times for some users. If you experience clicks during playback, simply enable the option by clicking the button. Most users should experience improved performance with this option enabled, but patches will load more slowly.
Fixes
Pressing group edit button crashes Pro Tools
Poor performance (RTAS)
DFD not available in RTAS / HTDM plug-in
Error on exiting Nuendo or Cubase SX
Crash in sonar 3 while playing midi-keyboard and resetting the instrument
Enable reverb while playback in host crashes engine
Selection list for delay in sync mode
Crash when re-activating plug-in
Sustain pedal settings are not stored in memory
Crash when switching DFD off
Crash when in group mode and loading a new patch
Crash in NI_DFD during multi-instance playback
Whole instrument turns off when switching DFD off
Distortion after letting Kompakt run for a few hours
VST version crashes Cakewalk Project5 on insertion
KOMPAKT 1.0.1
Fixes
Several problems and crashes regarding creating and importing sample files fixed
Browser refresh problems fixed
Known issues
The effect send produce a zipping noise when moved.
Kompakt can lock Logic key commands.
Mouse scrollwheel does not work in all hosts programs.
The level meter could be more accurate.
Giga import can produce large amounts of data.
DXi causes problems upon exit.
KOMPAKT 1.0
Notes
Hard disk energy save mode of the content hard disk will cause crackles. Make sure your HD does not go into sleep mode to have the best possible performance.
The library hard disk needs to be as defragmented as possible to deliver good load results.
Logic only allows to automate 128 parameters. Every parameter above that limit is clipped.
SF2, AKAI, Giga files are handled like folder. Simply open them and load Instruments or single samples from them.
The option settings are global for all instances of KOMPAKT. Changing them for one instance, will change them for all instances. For the effect to take place all instances of KOMPAKT need to be restarted. A simple solution here is to reload the song.

[IIO]

then what else could make a sampler require so much more cpu for playing a file compared to other samplers.

the result of 200% almost sounds like 16 to 32 bits or 44 khz to 96 khz...

[VMS]

then what else could make a sampler require so much more cpu for playing a file compared to other samplers.

the result of 200% almost sounds like 16 to 32 bits or 44 khz to 96 khz...

Not sure....
Maybe halion was optimized for pitch-transposition ?
We cant tell unless we analysis the transposed audio
Unfortunately i am not familir with such thing

Anyway kompakt does use less cpu power than halion for non transposed thing
I guess thats why MacTron got contradictions in the tests.

[IIO]

i believe there is not much to optimize. it is just converting rates... maybe one of them has an (unneccessary) high precision mixer and the other one not...

[VMS]

i believe there is not much to optimize. it is just converting rates... maybe one of them has an (unneccessary) high precision mixer and the other one not...

all can say is that they do sound quite different on highly transposed stuff

let's hope that someone could make an audio analysis to clear things up