Different speeds on different computers

[kozikowski]

if I fix the delay at the end, in the middle there is still a delay.

To add a few words, you got the beginning and end to match, and the middle is still off?
It's very unlikely that the clock signals would change within one performance. They almost always go their own way consistently.

What could do that...?

Now you're into supernatural territory. I have no good answers for that one. I could make something up like the drive filling up caused one computer to slow for the first half of the show and then a humidity increase caused the other one to start dropping samples.

Koz

[steve]

They almost always go their own way consistently.

I agree that clock signals usually run one way or the other, though I have seen a sound card gradually drift one way and then gradually drift back again.
Jack Audio has a neat workaround to this, which is to allocate one device as "master" and the other as "slave". If the slave becomes out of step from the master, Jack will duplicate or drop one sample from the slave to bring it back into step. The inevitable "glitch" from doing so is tiny and likely to be undetectable. I think that Jack on Windows is still classed as experimental so is perhaps not a viable option for non-geeks.

Heat can certainly have an effect on the clock rate and there is likely to be a greater change in clock speed during the first hour of use (see: http://www.qsl.net/d/dl4yhf///////speclab/frqcalib.htm)

[Learning2record]

You might want to change the tempo on one recording to match the other -- that way they are at the same speed in the end.

[kozikowski]

Yes, but the current problem is the ends match but not the middle. Koz

[arabtomten]

It's very difficult to say how much delay there is, but normally after an hour there's about 1/10 of a second to 1/2 of a second delay. I hope that helps.

[steve]

1/10 second drift over an hour is within what I would expect, and can be managed quite easily with just a little editing.
1/2 second over an hour is toward the maximum that I would thing reasonable, and I agree that is getting a bit awkward to manage, but nowhere near as bad as back in the days of tape recordings where the drift might be several seconds per minute!

The bottom line is that in the absence of active synchronisation, they will drift apart. What we can do is to try and minimise that drift so that it is more easily managed.
Try switching on the computers and mics at least half an hour before use, and try setting 48000 sample rate in both Audacity (Edit > Preferences > Quality) and in the Windows Sound Control Panel. Is that any improvement?