Get a Wavecenter PCI card + tango24 DA converter. Latency has a lot to do with a particular sound card\'s GSIF driver and the Frontier is the best in my opinion. Using the Digital outs on my Wavecenter I\'m getting under 5 ms. Even the RME gives around 8-9 ms.
I don\'t use Windows ME - according to what
I\'ve heard it bad in general and really bad for Audio stuff.
When you set latency to very low on any
card its vital that the harddisks are set
to use DMA transfer - otherwise the won\'t keep up and you get cracks
My giga/Delta1010 combo performs as well as vany of my hardware samplers ( Akai S5000, Emu S6400 ). I get between 6-8ms no problem.
Use a good solid seperate midi interface ( don\'t use the Delta1010 midi ) for best performance. I bought a 2 port midi flyer for $ 60...works great.
You don\'t need to buy any other soundcard...the GSIF driver for the 1010 is excellent...again, just use a dedicated Midi interface.
My main recording rig is a ProTools mix +. Just program a simple groove for about 5 min using either PT midi or Logic audio ( under DAE ). use giga as sound source and record the original midi groove as audio back in PT...zoom in and compare to the original midi groove data to the audio groove...this should give you the latency of the Giga + the A/D going in Pt ( wich is blazing fast under TDM ).
Hmmm. I\'m not quite sure from your description if you are actually measuring the difference between the MIDI signal at the input and the audio signal at the output of your GS machine.
What I\'ve done is set up a test jig similiar to what John Lloyd suggested at: http://groups.yahoo.com/group/NemesysMusic/message/2660 I sent the MIDI IN signal as an audio signal to the left input channel and the audio out of the GS machine to the right input channel on my main rig. From there I made several tests on the target GS machine. I created a single key region, mono triangle wave, 1 sec GIG file with no envelopes and sent out a single Note On/Note Off MIDI command. I measured the distance between the leading edge of the MIDI signal and the start of the wave.
800MHz PIII, BX chipset, M-Audio Audiophile, Roland MPU-401, latest VxD drivers, Windows ME, GS 2.5 beta.
Just for kicks, I pulled out an old sound module I had, an EMU Soundengine (basically a Proteus Lite) and tested that.
Direct: 6.5 - 7.5ms
I also took some preliminary measurements with Windows XP and GS 2.5 beta on a different machine (1GHz PIII, VIA chipset). Not suprisingly:
MIDI Input device -
Audiophile WDM: 20ms
Windows ME on the same machine:
MIDI Input device -
Audiophile VxD: 10.5ms
That\'s nearly double the latency. I also tested CPU utilization, but not using the GS meter. I attempted a more realistic test varying CPU speeds and listening for cracks and pops at 160 voices. Preliminary results show that XP requires a roughly 20-25% higher CPU frequency to achieve crack and pop free output. I need to repeat this using an unlocked Athlon CPU for more accurate results.