On 3/27/2023 10:08 AM, AugustA wrote:
On Mon, 27 Mar 2023 08:17:46 -0400
"Newyana2" <Newyana2@invalid.nospam> wrote:
N> "AugustA" <nospam@thankyou.invalid> wrote
N>
N> | Anyone here have their XP pc equipped with an SSD as their main drive?
N>
N> I've been using them for years and find them notably
N> faster. But it may depend on the vintage of your other
N> hardware.
The SSD is in a T60 Thinkpad. The interface is SATA-I. The writing speed is comparable for SATA-I, but the reading speed is slower? -- that's odd. Not sure if I can trust CrystalDiskMark.
"The result depends on Test File Size, Test File Position,
Fragmentation, IDE(PATA)/SATA/RAID/SCSI/NVMe controller and CPU speed etc…"
https://crystalmark.info/en/software/crystaldiskmark/crystaldiskmark-main-window/
"How to start Benchmark
1. Select Number of Test Runs, Test Size and Test Drive.
Test Drive
C:\, D:\ and etc...
If you would like to test Network Drive, please execute CrystalDiskMark
by standard user (UAC: No)
"
That makes it a partition-based test. And the results would
depend on WinXP (unaligned) or Win7 (aligned) partitioning.
WinXP uses CHS, and CHS numbers are divisible by 63, whereas
SSDs are based on 2**n numbers, and 63 does not divide into
2**n. ( 63 = 3*3*7 ) . The clusters do not end up on 2**n
boundaries as a result.
Normally, what should happen, is write should be slower.
Just a guess. That's because fractional writes, need two writes
for an unaligned cluster. Reads should hit on the track buffer
or the cache, and the impact on mis-alignment should be minimal.
*******
To bench at device level (having nothing to do with partitions),
try HDTune. It is a read-only benchmark, so should not touch
any data. I use this all the time, and on my daily driver. No damage.
https://www.hdtune.com/download.html
https://www.hdtune.com/files/hdtune_255.exe
The test has a limited "span" and on a 20TB drive, it would
only be testing a small portion of it. Giving exaggerated
speeds consistent with the outer diameter of the platters.
On a 1TB drive, HDTune should give representative numbers.
The paid version of HDTune does both Read and Write bench.
the free version does Read only.
In your case, we seek to see a faster read result (telling us
the hose pipe and the pumps are working normally :-) ). And this
would indicate the slowness has something to do with the partitioning
or the file system. If a partition was almost full, and you ran Crystal on it, perhaps the result would be pessimistic (the test file might be wedged
between fragments).
I'm not an expert on Crystal, so these are just ideas.
The only thing I don't like about "aligning" a mis-aligned
WinXP drive, is the impact on write life. It's wasteful
to have to do this, and it's better to get the alignment
right before installing an OS. While there is a standalone
utility that "aligns" and costs $39.95, normally you can
do this sort of thing for $0 by using backup/restore
utilities that have working re-alignment on restore.
But backup and restore is just as expensive as an
alignment utility.
If we were scientists, we would prepare two drives,
one aligned, one misaligned, and compare the crystal
numbers. I can't really do this in a VM, so it has
to be done on physical hardware. And my WinXP motherboard
died a year ago.
Paul
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)
