[G4] iTunes Speed/Performance Feedback

Mon May 26 13:59:24 PDT 2003

Kunga,
I'm not a real expert, but there may be some things you misunderstand 
about CD ripping. First of all, you will only get 52x at the extreme 
outside of a disk, because the the drive mechanism will not spin the CD 
fast enough to get 52x on the inside tracks. This is called constant 
angular velocity (think of how a record spins faster on the outside 
than by the label). Unless your drive will read at 52x constant linear 
velocity (how fast the data moves past the read head on the drive), you 
won't get that speed toward the inside of the CD where the tracks 
start. Most CDs don't have data all the way to the outside, so you may 
never get 52x data read. You can observe this during a rip by watching 
the speed go up as you get to the later tracks. My Quicksilver 933 
starts about 8x and proceeds up to about you 13x - 14x max that you 
have seen.
Next, you have to think about how dual processors work. To use dual 
processors efficiently, you must have a task that can be subdivided, 
but ripping a CD is inherently a linear task. For the CD drive to feed 
data to two different processors, it would have to constantly shift 
where it was reading, which would slow down the operation, not increase 
it.
Finally, the bus and drive are going to limit the top end on ripping 
speed much more so than the processor. Your Cube has only a 100 MHz bus 
and a 5400 rpm disk drive (if it's still stock) which adds more of a 
bottleneck. Now, a 1.2 GHz processor may give you some boost, but since 
I don't see much faster operation on a machine with a 133 MHz bus and a 
7200 rpm drive, the problem may be more with the ripping operation than 
the processor. Your dual will certainly improve your system response 
from a user interface perspective, since the OS can readily use both 
processors and interface tasks are not nearly as linear as CD ripping.
I would be very interested in hearing other perspectives on this 
problem.

Mel

On Monday, May 26, 2003, at 02:50  PM, Kunga wrote:

> Wish I had done the single 1.2 GHz Cube upgrade instead of buying a 
> dual 867.
>
> Just sent this to the iTunes team. Thought you would all like to know 
> this, if you don't already.
>
> I'm using OS 10.2.6. Ripping AAC files @ 192kbps.
>
> Dual processors DO NOT help increase ripping speed. iTunes rips faster 
> with one fast processor than it does with two slower ones that add up 
> to more than the speed of the single one. I found that out the hard 
> way.
>
> Bought a dual 867 thinking I was going to get rips more than 3 times 
> the 8x I get on my old G4 500 Cube (100 bus). Get no more than 13.5x 
> even with the same 52x reader. Both processors are working (can tell 
> via Cee Pee You).
>
> Cee Pee You 1.1.1 <http://www.unsanity.com/products.php> near the 
> bottom and FREE.
>
> But they only total the work of one processor and the system overhead 
> makes even that not as fast as it would do with one 867 processor.
>
> So for ripping AAC or MP3 files, it's all about a fast single 
> processor with a fast reader so far. Please try to develop a version 
> of iTunes that will use the full capability of both processors 
> continually. When I do an 8x rip on my G4 500 with a 52x FW reader, 
> the processor is working at a full 97%-99% load the entire rip. When I 
> do the same rip on my dual 867, both processors are flipping all over 
> the map between 20% and 80% both totaling little more than 100% at any 
> moment and yeilding a measly maximum 13.5x rip speed.
>
> I feel like I got ripped off buying this dual 867 for ripping because 
> iTunes is not really a dual processor program thus far. There is 
> something wrong with the code that prevents it from using both 
> processors up to their full capacity. If you will please fix that 
> code, I believe I will be able to get almost 28x rips on this dual 867 
> which would be the logical multiple from my single 500 to a dual 867 
> or 1734 GHz or 3.47 times 8 (3.47 is the multiple of 500 that my 1734 
> adds up to).
>
> Is the fact that iTunes doesn't use both processors fully OS X's fault 
> or iTunes' fault? Either way, I wish you would make this capability a 
> priority in the next version. I have over 3,000 CDs I must rip and the 
> process is still way too slow for me to want to do it thus far. I am 
> extremely upset to discover this weakness in iTunes which I thought 
> was a multiprocessor application. While it uses both processors, it 
> does not do so in a way that totals more than the power of one of 
> them. My dual 867 is ripping at a multiple of one 867 compared to one 
> 500 (on a slower buss too). (500 + 367/500) 1.73 times 8 is 14. All my 
> dual 867 rips occur at a maximum of 13.5x NEVER 14x. I understand 
> system overhead. But on a faster bus from the same 52x reader? Your 
> team should really be concerned about this. I would expect a dual 1.8 
> GH 970 PowerMac to deliver an iTunes ripping speed of at least the 52x 
> reading speed - about 1.1 minute per one hour album or the speed of 
> the 52x reader. The bottleneck should consistently be the processors 
> until the speed of the reader is reached. Please try to deliver that.
>
> Sincerely yours since SoundJam 2.0.
>
> Kunga
>
>
> ----------
> G4ist, a listserv for discussion of everything G4.
> FAQ at <http://www.themacintoshguy.com/lists/G4.html>.
> Send a message to <G4-DIGEST at themacintoshguy.com> to switch to the 
> digest.
>
> XRouter | Share your DSL or cable modem between multiple computers! 
> Dr. Bott | Now $139.99      <http://www.drbott.com/prod/xrouter.html>
>
> Dr. Bott LLC   4-port USB Hub in translucent Graphite!
> Dr. Bott LLC   <http://www.drbott.com/prod/ghub.html>
>
> Small Dog Electronics   http://www.smalldog.com  | Canon Digital 
> Cameras |
> --Quicksilver & refurbished G4's arriving daily! |   and Camcorders!!  
>   |
>
> MacResQ Specials: LaCie SCSI CDR From $99! PowerBook 3400/200 Only 
> $879! Norton AntiVirus 6 Only $19! We Stock PARTS! 
> <http://www.macresq.com>
>
> OS X News, Dr.Mac, Forums, Tutorials, Tips, Hints, FAQ?s - 
> http://www.osxfaq.com
>