SOS Forum Post
PC music specialist Martin Walker replies: I'm not surprised that you're confused, since there's lots of conflicting information around, and much of it is out of date. Although you've presumably just bought Gigastudio 2.5, that quote is actually from an FAQ document dated October 2001, which also says that Gigastudio is compatible with Creative Labs' SB Live soundcard using Direct Sound drivers (which is no longer true, since Gigastudio no longer supports Direct Sound under Windows 2000 and XP), and that laptops are not recommended due to the lack of GSIF-compatible PCMCIA soundcards (no longer true either, thanks to Echo's excellent Indigo range).
However, the most obvious giveaway is the mention of ISA-based MIDI interfaces, since I don't know of any modern PC motherboard that still has any of the now extremely elderly ISA expansion slots that you'd need to plug one in — the last time I bought one was back in 1998!
You could complain that manufacturers' support documents should be updated more often, but there is nevertheless still a grain of truth in the recommendation of PCI over USB, as the results of my two-part investigation into 'The Truth About Latency' in SOS September and October 2002 proved. My PCI-based MIDI interface gave me around 3.6ms input latency when capturing a keyboard performance, with a latency jitter of just 0.2ms; a serial-port interface gave around 4.2ms with jitter of 1.2ms; and a USB interface gave about 4.8ms with jitter of up to 1.9ms, all running under Windows XP.
This should hopefully prove to your satisfaction that USB isn't too slow, since an increase of just 1.2ms over PCI is simply not discernible while playing — a MIDI Note On message will itself last nearly 1ms, and a six-note chord could therefore emerge spread over 6ms.
I personally doubt that an increased jitter of up to 2ms would be noticeable in most situations either — many musicians can apparently detect timing jitter when it exceeds about 5ms, but below this it's likely to go unnoticed. It's possible that during dense MIDI passages the situation could get worse, but I don't think you should worry too much. Moreover, when playing software synths in 'real time' via MIDI, their timing jitter can be two to three times that of the MIDI interface.
Sadly it's extremely rare to find a PCI-based multi-channel MIDI interface nowadays — nearly all are USB devices. But there are various things you can do to minimise timing jitter problems with a USB MIDI interface. It's important to plug it into a dedicated USB port rather than a USB hub (powered or otherwise) so that the interface isn't fighting for its share of the USB bandwidth with other devices. Always use the latest interface drivers, and try not to use too many USB devices simultaneously, even if they are plugged into separate ports. Also, I personally still think it's tempting providence to try to run separate USB audio and USB MIDI interfaces simultaneously, since their drivers may end up fighting for supremacy.
The bottom line is that loads of musicians (including me) are now running multi-channel USB MIDI interfaces with no obvious timing problems, while some of the problems that others run into aren't necessarily due to the interface, but to other issues with their computers. When Nemesys wrote that FAQ, USB MIDI was still in its early stages, and things have improved since then.
It's also important to remember that while the capturing of a MIDI performance may be subject to a couple of milliseconds of timing uncertainty, many of us are relying more and more on software synths, whose playback timing is always accurate.