Jonny said:
The idea of offloading the pagefile from other than the XP system partition
is multi-fold. Primarily, you're looking for an alternate bus to allow some
semblance of dual access as in scsi. That is, doing more than one thing at
one time. A single partition exclusively for the swapfile at front of a
hard drive on such a bus, will provide that. A hard drive reads outside
first, then inwards. Opposite on CD/DVD. Such a partition is created at
the beginning of hard drive by XP, msdos, and many 3rd party softwares. It
does not have to be relocated.
A secondary benefit is less fragmentation. The swapfile, over time, will
fragment the system partition, and cause new file or modified file writes to
be fragmented.
If photoshop is writing temp files of some sort, they do not have to be
relocated. If you have some other bus system with a hard drive attached,
moving the default location for these temp files to that may improve
performance and cause less fragmentation of both the temp files on that
location, and the system partition files as the temp files aren't in that
location any longer. Mixing these with the new swapfile location defeats
the purpose of.
If photoshop is using temp files to create a final product, leave the temp
files on the system partition. Rather, assign the location of the final
product in another location instead. When satisfied with the final product,
delete the temp files if not done automatically.
And, don't mess with the swapfile min/max sizes. Let it play in the sandbox
to its heart's content. 4GB partition works for me.
Good points. You did miss one significant point of "theory": The primary
gain to be had by locating to a non-system and essentially "used" disk, I
would create partitions for rarely used achives & such, is two fold: (1) the
idea the the head will always be at of very near the track that it wants to
use thereby miniminze trach search/seak time and (2) This is new, with the
newer NCQ features the command queuing really does promote concurrency
especially if they are on separate interfaces, in this case ones of Pata and
the other on an SATA..
I did implement my idea and tested it to a degree: that is, placed my swap
file at the beginning of a separate SATA drive. I tested with HD-Tuner,
PCMark05, and 3DMark2001 before and after with no other changes:
Observations:
1. The Maxtor Diamond 10 100GB SATA with 8M buffer is not such a hot
product. Thats one reason it was only $79. It does show an average
throughput of 56MB/sec, far from the "spec'd" 150 MB/s but considerable
better than by older PATA system disk which only does 27 MB/s. The HD-Tune
results are very disturbing, over several runs, it consistently showed HUGE
plunges in performance going from 65-ish to 10 MBS. None of my other disks
show that behavior and I am thinking about giginf it back.
2. There was essentially no improvement measured by 3DMark2001.
3. A few test in PCMark5 showed a significant improvement related to some
very disk centric tests like encryption/decryption BUT the WIn Boot time was
significantly slower. In all, nothing to really get excited about.
4. One thing that these tests don't appear to measure is context switching
time. I know that because it was so markedly improved that it didn't need
decimal points to express it. Let's call it a 50%+ improvement. Very snappy
task/context swapping - I'd have to say the Windows must use pagin
extensively in context swapping (just like you'd guess.)
Bottom line: I don't have any Photoshop related tests, but I don't see much
improvement at all.
Currently: I put the slower PATA drive in the Paging mode and transfered
the old image to the new SATA disk. Thinks seem a little better yet, but I
have not rerun the tests yet.
RAM Disk is starting to look pretty interesting Huh?