Hard drive speed problems

[John Carlyle-Clarke]

I long suspected I had a problem with hard disk speed and benchmarking
confirms it. I am getting read/write rates of about 1 or 2 MB per
second up to maybe 10 at best, from a UDMA 100 drive. The drive is a
Seagate 20GB drive, and the Mobo is a Microstar NForce based one.

I run XP home and am totally up to date on all patches, drives and
SPs. I do have an 80 wire connector, and I can't see any damage to
it. I've tried removing the CD burner on channel 2, the only other
device. The HDD is the primary master. I've tried uninstalling all
the devices (IDE controller etc.), resetiing the ECSD (or is that
ESCD?) and letting it redetect everything. DMA is enabled in the
device manager. I see no conflicts there. I also see no errors in
the event log. I have run chkdsk, and the disk is only 1% fragmented.

I am going to try some different benchmark software, just in case. I
also have access to another identical PC, so I am going to try the
benchmarks on that to compare. Depending on those results, I am going
to try a different ribbon cable, maybe try a different drive in my PC,
and try my drive in a different PC.

I am really open to any helpful suggestion anyone can give me, but I
also have a couple of specific questions.

I realise that in the event of problems, XP will drop down to the
slowest access for drives. If this happens, would I always see
messages in the event log, or anywhere else? Can I check if this has
happened?

Also, does anyone know of any self-contained HDD benchmark software
that runs on DOS or Linux (preferably boot from floppy) that I could
try, to down whether I have a soft- or hard-ware problem?

Thanks in advance.

 

[John Carlyle-Clarke]

Most so-called HD benchmarks that run under WinWhatever are really
filesystem benchmarks, rather than HD benchmarks, because they
merely issue R/W commands through WinWhatever's filesystem; hence,
the results are clouded by caches and by the extra baggage of
metadata accesses.

Bob, thanks for that. Interesting stuff. Unfortunately, my drive is
NTFS Your info is still food for thought though.

Another question: I managed to run the same benchmark on another PC
with identical hardware. I saw about a 3 to 4 times better score on
the other machine. That machine has the 20G drive partitioned into 3
FAT32 partitions. Mine has a single NTFS partition. What kind of
speed difference should I expect these FS differences to account for?

 

[Bob Willard]

Bob, thanks for that. Interesting stuff. Unfortunately, my drive is
NTFS Your info is still food for thought though.

Another question: I managed to run the same benchmark on another PC
with identical hardware. I saw about a 3 to 4 times better score on
the other machine. That machine has the 20G drive partitioned into 3
FAT32 partitions. Mine has a single NTFS partition. What kind of
speed difference should I expect these FS differences to account for?

I've never done a heads-up comparison on the same hardware. I use NTFS
whenever possible because it is more crash-proof than FAT and because
it handles large HDs simply and supports much larger HDs.

I have seen other notes indicating that FAT is somewhat faster for small
HDs, but I am skeptical of reports of factor-of-2/3/4 differences either
way. I suspect that there are other differences between those two PCs
(the OS, cache setup within the OS, etc.); the way the HD is part'ed
can make a difference with some benchmarks.