I actualy missed a point ! If ripping to WAV or low CPU-greedy encoding (such as APE), thus using less than 100% CPU per data stream (or, differently put, where DAE speed is the limiting factor), it may be useful to rip simultaneously from multiple drives. I had stucked in my head
that ripping always went along with MP3 encoding (that's MY prefered format), which is particularly CPU intensiv (especially for high variable bitrates).
@ kiwi & nameless : you got a point
PS: I used to rip to WAV then convert to MP3... but at the end of the day, you end up with the same amount of time until your files are done with. And ripping "directly" to MP3 allows me to lower down the CD-drive read speed (since it is encoding that slowers down the process) and thus have a much more reliable DAE (especially with bad-shape CDs).
As for the encoding speed, with my "old" PIII @ 1 Ghz + 512MB, I encode MP3 (LAME --preset standard) no faster than 2.5x.
I'm
curious how fast can the new CPUs be, especially with the augmented bus speed and high-performance RAM. Anyone interested in comparing (for the "LAME 3.93.1 --preset standard" scenario, for example) ?
Personally, I had the feeling in was not worth spending several hundreds bucks (CPU+RAM+board) to achieve no more than 2-odd times speed gain. But if we're speaking 5-odd times or more...
UPDATE:
Just went to have a look @
http://www.spec.org ; using the latest P4 @ 3.2 Ghz (lotta $...), it seems I could expect a 3x (integer ops) / 5x (floating ops) perforance gain compared to my PIII... now, as for actual
application speed gain (mix of int ops, float ops, I/O access, etc..), I wonder... and I don't know if SPEC's benchmark go deeply into the MMX domain (one would need to read and understand their benchmark's description better than I do)
But sorry ! I'm disgressing from the original topic of this thread. If everyone interested to follow the matter, let's start a new thread.