"Steve James" said in :
Hi

I have two HD's on the same IDE channel, I think one is faster than
the other (ATA and RPM), will the fastest one be 'held back' by the
slowest one or not ? I have my O/S and program files on the main
(fastest) drive and only use the slower one for storage and the
pagefile. (to reduce head travel)

MTIA

Steve

It depends on what motherboard chipset you use. I had a 3+ year-old
Pentium 3 800MHz Slot1 (SECC2) system that used the AOpen AX6BC
motherboard which uses Intel's 440BX chipset. The IDE controller
operated only in single mode; i.e., it would get set for whatever was
the slowest reported device. Doesn't matter what the IDE ports were set
for in the OS. The IDE controller provides 2 channels. For that
system, it was important to match the ATA spec of the devices on each
IDE port and try to NOT mix ATA (hard drive) and ATAPI (CD-ROM drive)
devices on the same port. CD-ROM drives then and still only operate up
to UDMA-33. Putting an UDMA-33/66/100/133 hard drive on the same IDE
channel as, say, a UDMA-33 CD-ROM drive, had the controller operate at
the least-common-denominator of whatever was the slowest device on that
channel so the hard drive would also operate at UDMA-33. So I had my 2
hard drives on IDE0 and the CD-RW and DVD-ROM drives on IDE1 to keep the
slower UDMA-33 CD-type drives off the IDE channel for the faster hard
drives. One of my hard drives was UDMA-100 while the other was UDMA-66
but beyond UDMA-66 there is little real speed increase on end-user host
whose primary function is to run applications rather than operate as a
file server. The interrupted use of the hard drive along with all the
head travel obviates the momentary but short-lived burst mode of the
drive. One drive spun at 7200 RPM and the other at 5400 RPM but that
only affects how long to get to a particular sector on the same track
(so keeping your drive defragmented helps).

So for the old chipsets, they could operate the 2 IDE channels at
different modes (i.e., speeds) but they could only use the lowest
device's mode for all devices on the same channel. One channel could
run in one mode and the other channel could run in a different mode, but
a channel itself could only operate in one mode. Now comes along newer
chipsets that can support multiple concurrent modes on the same IDE
channel. If you visit Intel's web site to look at, say, the hardware
specs for their 845E chipset, the doc for its paired ICH4 (southbridge)
chip notes, "Each IDE device can have independent timings". So you can
have a UDMA-100 hard drive on the same IDE port as a UDMA-33 CD-ROM
drive and the controller will use the best mode to support the highest
transfer rate of each device. There's no harm in configuring the
devices as was needed under the old 1-mode-per-channel chipsets but
there's no point, either. So the better setup now is to have the hard
drives on separate IDE channels (to maximum concurrent bandwidth
capacity for the most used devices) and put the CD-type drives on
whichever channel you want (but preferrably on whichever is not the IDE
channel being used for the Windows "boot" partition, if possible).

Microsoft's terminology regarding which is the boot and system partition
is a bit backwards of common sense. To Microsoft, the "system
partition" is where are located the bootup files and the program located
in the boot sector of the hard drive partition that the bootstrap code
in the MBR uses to start loading the OS. For Windows, this is wherever
are placed the NTLDR, NTDETECT.COM, and BOOT.INI files. The "boot
partition" is wherever is the rest of Windows that gets loaded, like
NTOSKRNL.EXE. So you have boot files on the system partition and system
files on the boot partition. Yeah, go figure. See
http://support.microsoft.com/?kbid=314470.

The system and boot OS partitions can be in the same hard drive
partition and is the typical scenario with Windows installed wholly on,
say, the C: drive. The system BIOS can only read the bootstrap code
from MBR (master boot record, or sector 0) on the first detected
physical hard drive and uses the partition table in that MBR to
determine which is the current active primary partition (which is only
on that same drive since the partition table only has entries for that
drive). The MBR bootstrap program then loads the boot program in first
sector of that partition. Since the system BIOS can only read the MBR
on the first detected physical hard drive, the "system partition" for
Windows must be on that same first hard drive. You can change this
behavior if you replace the standard bootstrap program in the MBR with a
more robust program, like a boot manager (IBM BootManager, Powerquest's
BootMagic, etc.). However, the "boot partition" for the rest of Windows
can be in the same partition on that hard drive, in another partition on
that same hard drive, or in a partition on a completely different hard
drive.

If you have 2 hard drives then you should place pagefiles on BOTH hard
drives. The "boot partition" (where Windows is read from after it got
started) should have a pagefile set to 64KB larger than your system RAM
size but can be larger. Then configure a pagefile on the other hard
drive of the same size or larger, typically 1.5 to 2.5 times your RAM
size. Windows will detect that a pagefile exists on a hard drive other
than its boot partition and give it higher priority for use. To reduce
fragmentation of the pagefiles, set their min and max size to the same
value. The defrag program included in Windows will not defrag the
pagefiles, so you need to use the trick of rebooting into Recovery
Console mode to delete the pagefiles (after already having defragged the
drives) so they get recreated on Window startup. Some 3rd party
defraggers, like Diskeeeper and PerfectDisk, will defrag the
pagefile(s). Some folks will even create a separate partition on the
non-boot partition hard drive just for that pagefile to keep it from
getting fragmented with data and applications that normally show up on
the other hard drive. I've never bothered going that extreme but then I
regularly defrag my drives. The pagefile being 1 piece might have a
tiny performance advantage to a pagefile sliced into 3 or 5 pieces but
that's insignificant to the performance loss if the pagefile was sliced
into hundreds of pieces.

Even if the drives spin at different speeds, you'll want to allocate a
pagefile on the non-boot partition to allow Windows to page out virtual
memory while it can still concurrently read/write files on its own boot
partition. If the hard drives are UDMA-66 or higher, it really doesn't
matter that the primarily used pagefile on the non-boot partition is
spec'ed to use a slightly slower burst mode. However, once you get a
difference of UDMA-100 (nothing really for UDMA-133) against UDMA-33 for
2 hard drives, and especially if the UDMA-33 spins at 5400 RPM and the
UDMA-100 drive spins at 7200 RPM, you'll probably want to forego using a
pagefile on that slower non-boot partition hard drive.