It is known that Cubase SX (starting from release 3) reports strange (but real?) latency values with some audio cards, Creamware included.
Here is a test we (a group of users and tech) have done in the past days trying to understand the problem.
We have tested the MIDI ports also.
The PC used for the test is a well-used, rock-solid machine:
MB: Asus P4P800
CPU: Pentium 4 HT 3.2 GHz 800 MHz FSB
RAM: 1GB Corsair
Video: Matrox G450 AGP in dual monitor mode
Supply: Enermax
Audio: Creamware Pulsar 2
OS: Windows XP with SP1, optimised for audio workstation, ACPI mode
DirectX 9.0c
Scope 4.0 - driver from 3.1a
Cubase SX 2.2.0
Cubase SX 3.0.2.623 (latest)
Scope settings:
Sample rate: 44100
ULLI: 4 mS @ 44100
ASIO1 Flt 64 source and destination.
Hiperthreading disabled during the test.
For the audio loop-back test we have routed a couple of ASIO source channels directly to the ASIO destination module; we have played a track while recording itself on another track thru the ASIO driver.
A test project ready to load is available from Steinberg: see their FTP server under documents folder.
Cubase 2.2.0 test results:
Reported latency: Input 4.376 mS, Output 4.376 mS (correct, about 192 samples).
Audio loop-back test:
Offset 0 samples; the latency reported by the program is true as there is no offset.
Cubase 3.0.2.623
Reported latency:
VST Audio bay: Input 4.376 mS, Output 7.279 (?)
ASIO Scope: Input 4.376 mS, Output 4.376 mS
Audio loop-back test:
Offset -128 samples; Haaa… the output latency reported is incorrect as the compensation introduced by the program; in fact, 128 samples are equal to 7.279 - 4.376 = 2.903 mS.
Fortunately, this error can be cancelled out by setting the record placement offset to -128 (Device Setup-VST Audio Bay-Expert-Record Placement Offset).
Obviously this value is dependent to the ULLI settings.
But the question here is: why Cubase calculate this erroneous value (and only with some audio cards/drivers)?
Or, is it a Steinberg problem or a Creamware problem?
For us the most important thing is that THIS ADDED LATENCY IS NOT REAL and a workaround exist for the recording offset; we will see…
About Creamware MIDI ports:
We have done a MIDI loop-back test also, to see the behaviour of emulated Directmusic vs. Windows MIDI ports (why Creamware don’t make a real Directmusic driver?).
We have had the best results with the emulated ports and the ”Use system timestamp” option unchecked.
The available time precision is sufficient but inferior in respect to other MIDI interfaces (the MIDI ports of the M-Audio Audiophile 2496 for example).
The Windows MIDI ports are simple unusable (too much time offset).
Be aware that MIDI performance can vary from system to system due to different chipsets and clocks used.
Hope this can be helpful…
Best Regards
