[__david__]

Funny, my experience is the opposite--FireWire drivers are generally very DMA oriented and require little to no CPU at all, while USB drivers tend to need more CPU for the protocol overhead. This is because FireWire is an address oriented bus and the hardware can map FireWire addresses to CPU addresses and DMA in and out with (literally) zero CPU overhead. A driver write would really have to go out of their way to make a FireWire protocol become CPU bound. I suppose the audio protocol is isochronous based which would be slightly different, but that is still DMA based in every FW implementation I know of.

8 channels of 24 bit/192KHz audio is only 2MB/s which is basically nothing. I'm very surprised that you would see any issues.

I have about 3 or 4 devices in my FW chain (including disks and scanners) and my sub $1000 FW audio interface works just fine.

[swift]

Same here. I have a $250 FireWire audio interface that imposes basically no load on the system even with upwards of 10 channels in use.

I chose FireWire over USB because of all the complaints I saw in the Amazon reviews for every USB audio interface I looked at; many of the FireWire audio interfaces had absolutely glowing reviews.

I've had similarly excellent experiences with disks connected via FireWire.

[icarus_drowning]

Yeah, this has been my experience as well. Even cheap Firewire audio interfaces are light-years better than their USB counterparts.

I still use a USB 2.0 MIDI controller and the latency is absolutely killer. Press a key, a beat later, see the note appear on piano roll...

[nitrogen]

That's interesting. I've had no problems with USB MIDI adapters. As long as I have less than 5ms latency in my audio path, I'm able to play comfortably over the cheapest USB-MIDI adapter on Amazon. I've also had reasonable success with USB audio interfaces, with latency around 5ms (IIRC - I switched back to my ancient emu10k1 card because of the DSP).