>From: mbbrutman-cctalk at brutman.com
>I revived an old 6Mhz PC AT today by removing the original Western Digital
>hard disk controller and replacing it with a newer, generic IDE controller.
>The large boat-anchor class 5.25" full height hard disk had died, and without a
>suitable replacement a small IDE drive was chosen. The rest of the machine is
>original, including the EGA card with the extra memory daughterboard and the
>full slot monochrome card.
>While discussing this beast, the topic of putting a CD-ROM on it came up. (I
>would never do such a thing except for giggles.) The topic of the CD-ROM led
>to a throughput question - could this old monster even do it? As in, is the
>ISA bus up to reading from the CD-ROM at a reasonable speed?
>Well, let's go throught some quick math. 6 million clock cycles per second is
>lot, but when you consider that each transation on the bus takes 4 to 5 cycles,
>and instructions take 4 to 5 cycles (average) on a 80285, the throughput might
>not be so hot.
>Forget the CD-ROM and let's just pretend it's an I/O device on the bus, like a
>fast hard disk. Given a PC AT running at 6Mhz, what can reasonably be expected
>for bus throughput? (Obviously this depends on the loop and the instructions
>used.) I would have guess 500K or so per second, but now I'm second guessing
>myself and I'll actually have to write a small benchmark to try it.
>When writing to a hard disk, would this machine have used a tight processor
>loop, or would it have used DMA? Under what circumstances would it use DMA to
>transfer data to a hard disk?
Hard disk controller have at lest one sector RAM buffer and most
newer ones have at least a track worth of RAM. The controller doesn't
write the sector until the buffer has the new data. Reads are buffered
the same so that bus speeds are not issues, like for floppies.
>And one last question .. Unlike the PC and XT, the AT BIOS handles hard
> It didn't blink when I removed the crusty WD based controller and replaced it
>with a no-name WinBond based controller. Does the new IDE controller really
>look that much like the old controller that the BIOS can't tell?