del@pilchuck.UUCP (Erik ) (05/21/86)
{}
I have lost the original message (our news gets purged SO fast around here),
but this is in response to a posted question about why AT disk operations
benchmarked so FAST.
1) As stated in the article, the AT hard disks seek at least twice as
fast as the XT hard disks.
2) The XT disk format uses a sector interleave of 6, while the AT uses
a sector interleave of 2. This means an XT has to wait for 6 full
disk revolutions to get a full track, while an AT gets it in 2.
3) The AT is just a faster processor (no kidding :-), so time between
disk accesses is less, too.
I don't know what mathematical algorithm relates the above parameters,
but when you combine them, you do see a speedup of 6-10 times that of
an XT.
Speaking of mathematical relationships, one that I don't understand is
why the disk access (copy to nul for example) is that fast, and Norton
sysinfo says my 9Mhz AT cranks along at 8.7 * PC performance, but my
compiles only go 4-5 times as fast.
I'll just have to live with it, at least till I can buy a 386 machine.mo@well.UUCP (Maurice Weitman) (05/22/86)
In article <346@pilchuck.UUCP> del@pilchuck.UUCP (Erik ) writes: >[] > >I don't know what mathematical algorithm relates the above parameters, >but when you combine them, you do see a speedup of 6-10 times that of >an XT. > >Speaking of mathematical relationships, one that I don't understand is >why the disk access (copy to nul for example) is that fast, and Norton >sysinfo says my 9Mhz AT cranks along at 8.7 * PC performance, but my >compiles only go 4-5 times as fast. > First of all, even Norton admits that his SI ratings are misleading at best, and I say they're worse. He uses an algorithm which is heavily weighted for ADDs, I believe, and not really representative of what a machine will do under many other circumstances. Next, there are so many variables here (such as disk interleave factors, number of buffers in your config.sys files, disk controller efficiencies, etc.) that looking for a mathematical relationship is quite an ambitious and maybe nebulous project. If you're really serious about improving throughput in your AT, look into using one of the disk cacheing programs, consider playing around with the disk interleave factor, using a ram disk for many of your compiler files. If you'd like more specific info, send me e-mail, and I'll try to answer. (I'm waiting for my '386, too %-) ) maurice -- Maurice Weitman 9600 ..!{hplabs,lll-crg,ptsfa,glacier}!well!mo | 57600 (415)549-0280 voice (415)549-0388 modem-2400 this^is not 300 mcimail mweitman a pipe 110 P. O. Box 10019 Berkeley, CA 94709