IDE Transfer rate

Greetings!

Does anyone (and this question is addressed to the math experts out there) have a general feeling for the throughput you would see on a disk to disk copy on an empeg? I know the limiting factor is the IDE bus speed, but has anyone ever calculated it out?

I don't know whether it could be easily calculated or whether real life measurements are the only way to go. ISTR that the IDE interface on the empeg is CPU bound, so I don't know whether the IDE specified speeds can be obtained.

My empeg's in the car right now... does it have the "time" command? I'd be interested to see the output of some speed tests. These are the kind of tests I do when I get a new HD and so far, the limiting factor in my P2@400 has always been the drives (each new one outperforming the previous).

Disclaimer: I'm not currently logged in to my empeg so I can't verify the file names I'm talking about
Write test:

# time dd if=/dev/zero of=/drive0/testfile bs=1048576 count=1024

Read test:

# time dd if=/drive0/testfile of=/dev/null bs=1048576 count=1024

If the drive is fairly full, it will slow these numbers down (due to the testfile being closer to the end of the drive [always slower] and possible fragmentation). An alternate read test that would read from the drive sequentially is:
Alt read test:

# time dd if=/dev/hda of=/dev/null bs=1048576 count=1024

Then try going between the drives:
Cross-drive test:

# time dd if=/drive0/testfile of=/drive1/testfile bs=1048576 count=1024

All the tests above are using 1GB (1073741824 bytes) of data so you can get MB/s by taking 1024/time (e.g. 20 seconds to complete = 51.2MBps). Stopping the player application and quitting other hacks that may be running will help get better performance.

On systems without DMA, the CPU is usually the limiting factor and this can be seen in the output of the time command listing 100% CPU.

P.S. How do I make my CPU speed not be linked like an email address?

Do you mean between two drives in a dual-drive empeg, or if you were to take a drive out of the empeg and mount it in a desktop Linux PC?

I doubt that time is installed on the empeg. It shouldn't be too hard to put there though.

For some of these tests, I'm not sure why you don't just invoke hdparm -Tt It won't do inter-drive read-writes though, you'll still need another method as described.

P.S. How do I make my CPU speed not be linked like an email address?

Good question...
P2@400 ?

Yep. Like this;
P2[i][/i]@400

...which is also the handy method that allows you to curse... shit damn poo!

Does anyone (and this question is addressed to the math experts out there) have a general feeling for the throughput you would see on a disk to disk copy on an empeg?

I haven't timed disk-to-disk, but memory-to-disk gets about 4Mbytes/sec, and it's CPU-bound, so I wouldn't expect more than 2Mbytes/sec disk-to-disk.

This is much less than the head rate of modern laptop disks, so you should be able to do much better if you put them into a Linux PC with an adaptor.

Peter

Thanks. I was just looking for an estimate, and this gives me a good basis.