CompuServe Messages

John, Those data rates are burst rates…not sustained rates. The limiting factor is the drive itself. Let's take a fairly typical large, fast SCSI HD. 15 heads, 35 sectors per track, 1:1 interleave, essentially 0 time to switch heads and spinning at 3600 rpm. That means it is spinning at 60 times per second…thus you could read any one track 60 times in once second. 35 sectors x 512 bytes per sector x 60 means the maximum speed you can get data off the drive is 1,075,200 bytes per second. But since you don't constantly read a single track, then you also have to account for head stepping/settling time. With our 15 heads, as long as you stay in the same cylinder…that time is essentially 0. So you can read the first 15 tracks (idealy) in 15 revolutions. But now you need to step and that can take anywhere from ~15ms on up. That's at least 1 revolution (16.67ms per rev). Since 15 (heads) goes into 60 (revs) 4 times…you lose 4 revs every second if you only step 1 cylinder at a time. Thus instead of being able to read 60 tracks, you can only read 56 each second. That lowers the max data rate to 56 x 35 x 512 or 1,003,520 bytes per second. And that presumes that all the data is laid out sequentially. In real life, that almost never happens. So your real data rate will always be lower. With disk cacheing schemes, or intelligent on-drive data buffering, it's possible for the drive to store up to X amount of data from the read, and then send it down the bus at high speed (as you mentioned up to 4 Meg per sec)…but that's only in brief bursts. Sustained rates on drives like described above will always be much lower than the burst rate. So, if you can do fast anims with sustained data rates of 750K per sec or so…yes. Other wise, no matter what controller…it's just not possible. Don