|
Forum Index | Read LUG: Policy/Rules Messages Threads Digests | Post New Message | Search!
There seem to be two threads about performance that have come up in
this forum: Latency (you want very low latency for tracking via
software monitoring, if you need to track through effects such as a
guitar amp) and Disk/Track count (which you want very high for large
projects).
This is NOT a test for track count. I set out to do a latency
comparison between Logic 7 and Cubase SX3. First of all, my testing
methodology (curtesy of Dennis Gunn):
1) Rec-Enable a stereo track
2) Plug guitar into one of the stereo inputs (in my case, right)
3) Route output (in my case, right) into the other input (in my case,
left)
4) Record transients (hitting pickup with piece of metal)
5) Open region in SE and count the sample distance between the right
and left transients.
This gives a pretty accurate reading of round trip latency. I am
recording 24-bit, 44.1k files.
I am using an RME HDSP 9632 card which feeds a SPDIF signal to Metric
Halo Mobile I/O ULN2 which serves as AD/DA and Mic-Preamp for the
signals.
After each of the latency tests, I do a "reality check playback
test"
in which I run a demo song at that buffer setting, as a test to see if
that buffer setting is realistic for recording/playing back a project.
I chose demo songs because since they are included with the package,
everyone can verify my results on their machines, and because the demos
are usually optimized to run well.
Test machine: PowerMac G5 2Ghz dual processor, 1GB RAM, OSX 10.3.5
I'll start with Logic 7.
1st Logic Test:
Core Audio Buffer: 64 samples
Samples of Latency: 480 samples
CPU usage of single stereo 24/44.1 track armed w/guitar amp pro
inserted: less than 5%
2nd Logic Test:
Core Audio Buffer: 32 samples
Samples of Latency: 376 samples
CPU usage of single stereo 24/44.1 track armed w/guitar amp pro
inserted: about 5%
Reality Check Playback Test: For my reality test, I ran the Logic 7
Demo Song (12" Full Track version). This song consists of:
50 tracks
17 Audio Instruments
100+ effects
This song runs at 50% at 128 samples, with one peak at 80%. At 64
sample buffer, this song ran at about 60% CPU with one peak at 90%. At
32 sample buffer, this song ran at about 70% CPU with one peak at about
100% (both meters full up) but there was no audio dropout at all. In
other words, recording tracks at 64 samples with a song like this could
work, but recording tracks at 32 samples with a song like this might
not work.
Now for Cubase SX3. Cubase does not allow for buffer settings below 64
samples.
Buffer: 64 samples
Samples of latency: 545 samples
However: Cubase SX 3 has an "Expert" mode in its driver section so
I
ran more tests:
Audio Priority to Very High: 545 samples
Lower Latency Checked: 416 samples
Multi-Processing Unchecked: 545 samples
Lower Latency, MP off, Very High Audio Priority: 416 samples.
CPU usage of single stereo 24/44.1 track track armed w/overdrive effect
inserted at 64 sample, low latency checked: 15%.
Reality Check Playback Test: I used one of the SX3 demos, Blofeld's
Return, because it also ran at 50% CPU at a 128 sample buffer. The
song consists of:
42 tracks
4 instruments
5 plug-ins
With my buffers set at 64 samples, this song would simply not play. At
64 samples and low latency the song seemed to be playing at varispeed,
way slow and with lots of crackles (and peaked the meter regularly).
With low latency off and audio priority set to high, the song played
back normally, but there were still a couple peaks that caused audio
dropouts. Even setting the preload to 6 seconds (the maximum allowed)
the song simply would not play back at all. At 128 samples, however,
even with Low Latency checked, the song ran at 50% CPU. From these
results, I'm not sure if 64 sample recording is feasible in anything
but a very small project using Cubase SX3 on a Macintosh. I'm assuming
that its PC ASIO results would be better, however.
(REMEMBER: this is not a "track count" test. With the buffer at
512
samples, the CPU use for the Cubase SX demo song drops to 20% and
playback is not an issue. This is ONLY a low latency test).
Analysis: Cubase/Nuendo may be the track count king, but for low
latency recording with effects, Logic Pro 7 gets the nod.
Orren
--
http://www.mertonfolio.com
Author of:
• Logic_6_Power
• GarageBand Ignite
http://www.courseptr.com/ptr_catalog.cfm?group=Music%20Technology
Forum Index | Read LUG: Policy/Rules Messages Threads Digests | Post New Message | Search! © 1994-2008, All Rights Reserved. |