This article is from the CD-Recordable FAQ, by Andy McFadden (firstname.lastname@example.org) with numerous contributions by others.
It depends on your recorder, media, and who you talk to. For example,
some informal testing with the venerable Yamaha CDR-100 determined
that it worked best at 4x speed with media certified for 4x writes.
1x worked almost as well, but 2x would occasionally produce discs with
With audio CDs, the results are more subjective. Some people have asserted
that you should always write at 1x, others have stated that 2x may actually
be better. It depends on the recorder, media, player, and your ears. Try
it both ways and listen. See section (4-18) for some notes on how you can
write the same set of bits to two CDs and still have audible differences.
CD-R media is written by heating up tiny sections of the disc. When the
disc spins faster, the laser has less time to shine on a particular spot,
so the laser has to be controlled differently. Different formulations of
media may require a different "write strategy" at certain speeds, and each
recorder may adjust its write strategy differently to accommodate those
speeds. This can potentially result in combinations of recorder and media
that work perfectly at one speed but fail miserably at another.
Put simply, there's more to writing at high speed than just spinning faster.
It's entirely possible that writing slowly to "high-speed" media will
produce significantly worse results than writing to it quickly.
There is no One True Answer to this question. Do what works best for
what you have. Some experimentation may be required.
See "The Speed of Sound: How Safe is High-Speed CD-Audio Recording?" at
http://www.emedialive.com/EM2000/starrett5.html, for a very thorough
analysis of audio disc quality at several different speeds. With some
recorders and some media, it's actually better to write faster -- but in
none of the tests performed did the error rate get anywhere near danger
levels, regardless of speed.
See the graphs in the article "Glenn Meadows' CDR Tests" at
http://www.digido.com/ for an examination of BLER (BLock Error Rate) with
different recorders, different media, and different recording speeds.
A few of the graphs show the same recorder and same media at different
speeds, and in some cases the BLER increased at higher speeds, while in
others it decreased.
There is some cause to believe that recording at higher speeds can result
in increasing "jitter". This doesn't cause any difference in BLER or in
the extracted audio, but is audible during playback. See section (2-41).
See http://www.cd-info.com/CDIC/History/Commentary/Parker/stcroix.html for
commentary about "write strategy" selection and different media types.