How to increase system system performance

[John John - MVP]

You don't need an extremely high write speed. A lot of times temp files
are just empty files, many are 0 bytes. Almost all are under 700KB.
Even at a write speed of of say a low of 5MB/s is still only a fraction
of a sec.

This keeps the read/write head from thrashing about creating and
updating file records.

And just to up the ante, I enabled disk compression on the USB drives to
reduce the size of the writes.

This notion of yours is getting loopier with each of your replies. Now
you think that enabling compression is going to compensate for the
slower USB flash drives and increase performance because the writes are
going to be smaller, yet you fail to take into consideration the
overhead involved into compressing and decompressing files.

John

 

[Tae Song]

This notion of yours is getting loopier with each of your replies. Now
you think that enabling compression is going to compensate for the slower
USB flash drives and increase performance because the writes are going to
be smaller, yet you fail to take into consideration the overhead involved
into compressing and decompressing files.

John

Over head is insignificant, even for a single core CPUs. 8.5GB/s transfer
rate from RAM is slow for a CPU. Compression is just something to do while
it waits for the next byte to come by. And I'm running on a Core2 Quad.
Overhead for compression is next to nothing.

 

[John John - MVP]

Over head is insignificant, even for a single core CPUs. 8.5GB/s
transfer rate from RAM is slow for a CPU. Compression is just
something to do while it waits for the next byte to come by. And I'm
running on a Core2 Quad. Overhead for compression is next to nothing.

So that will make your USB drive faster than an internal hard drive? If
that is the case why not just compress the files on the hard drive and
make the internal hard drive that much faster than the USB flash drive?
You are grasping at straws, the plain facts are that USB flash drives
are slower than internal hard disks and whether you want to admit it or
not there is an overhead when file compression is involved and even if
you compress the files the USB drive will still be slower than the
internal drive. The subject of your post is "How to increase system
system performance" yet everything that you propose (including
compression) has the opposite effect!

John

 

[Pegasus [MVP]]

Over head is insignificant, even for a single core CPUs. 8.5GB/s
transfer rate from RAM is slow for a CPU. Compression is just something
to do while it waits for the next byte to come by. And I'm running on a
Core2 Quad. Overhead for compression is next to nothing.

It's been suggested several times before: How about backing up some of your
exotic suggestions with actual and reproducible measurements? Without those
your posts are little else than a soap box oratory: Great for you if you
like to hear yourself speak but not taken seriously by anyone in the
audience.

 

[Peter Foldes]

Tea Song

Drop this issue while you are still ahead (maybe not). You are beating a dead horse
with this crap. Move on and go back to answering posts without the CBS.log for every
issue.

 

[Tae Song]

So that will make your USB drive faster than an internal hard drive? If
that is the case why not just compress the files on the hard drive and
make the internal hard drive that much faster than the USB flash drive?
You are grasping at straws, the plain facts are that USB flash drives are
slower than internal hard disks and whether you want to admit it or not
there is an overhead when file compression is involved and even if you
compress the files the USB drive will still be slower than the internal
drive. The subject of your post is "How to increase system system
performance" yet everything that you propose (including compression) has
the opposite effect!

John

OK very last post on this subject... hopefully.

Yes, in a certain specific case compression could make even a slow USB drive
faster than even a hard drive. The requirement would be the file would have
to be very compressible.

I did try compressing the whole hard drive to see if would improve
performance on a Windows XP machine. I didn't know it at the time, but
Windows will compress the Bootmgr if you don't exclude it. Nice of
Microsoft to at least tell you what the problem is when you try to reboot.
Bootmgr is compressed, LOL!

I will give that another try, one folder at a time on this machine.

I'm sorry I don't have benchmarks to back up the any claim... I'm just too
lazy to do one and I can't seem to find a free storage benchmark program
that I like. Actually I don't really care, I just like this setup, it
works for me. People have a hard time keeping the computer up and running
as it is, they don't really need another factor to complicate their setup.

Gnu Image Manipulation Program 2.6.6 takes the longest time to start for me
on this machine. Used to take over 10sec, now 4.33sec on a stopwatch.

Not very scientific but there you go.

 

[Pegasus [MVP]]

OK very last post on this subject... hopefully.

I'm sorry I don't have benchmarks to back up the any claim... I'm just too
lazy to do one and I can't seem to find a free storage benchmark program
that I like. Actually I don't really care, I just like this setup, it
works for me.

Not very scientific but there you go.

In other words: You like the slow-down you designed for your machine and you
insist telling everyone about it, dressing it up as the greatest thing since
sliced bread. Do you really expect anyone to believe your claims when, by
your own admission, you're too lazy to verify them?

 

[John John - MVP]

... People have a hard time keeping the computer up
and running as it is, they don't really need another factor to
complicate their setup.

Exactly. All the more reason why they shouldn't bother with your
"performance tweaks".

John

 

[Tae Song]

In other words: You like the slow-down you designed for your machine and
you insist telling everyone about it, dressing it up as the greatest thing
since sliced bread. Do you really expect anyone to believe your claims
when, by your own admission, you're too lazy to verify them?

I thought up a better way to illustrate my point, so I didn't resist posting
this... shame on me.

When loading an application, Windows reads files from the hard drive into
memory while creating temp files. That's read throughput (hard drive) +
write throughput (USB drive) is greater than the throughput of hard drive
alone. It's not a competition about which drive is faster. And read and
write operations described previously can occur side-by-side, where as on
the hard drive read and write would have to be queued.

(Temp files on flash drive)
USB ++++*
HD ----------

(Temp files on hard drive)
USB
HD --o++-------

o = overhead for extra seeks associated with having the temp files on the
hard drive.

* I put in an extra ++ for temp files on flash drive since writes are slower
and also to illustrate why it doesn't matter.


I did think up of what I thought was the best way to benchmark and
illustrate the performance difference. It requires the setup of identical
hardware to isolate any variables down to just the temp file locations. It
would be preferable to have one set of input devices for both computers so
they are getting the same input at the same time. But I just don't have
that kind of money.

Anyways, even if I did tons of benchmarks on MY hardware. It doesn't mean
you will get the same results on YOUR hardware. You will just have to test
out the idea for yourself, unless you don't want to find out on your own.

 

[Pegasus [MVP]]

I thought up a better way to illustrate my point, so I didn't resist
posting this... shame on me.

When loading an application, Windows reads files from the hard drive into
memory while creating temp files. That's read throughput (hard drive) +
write throughput (USB drive) is greater than the throughput of hard drive
alone. It's not a competition about which drive is faster. And read and
write operations described previously can occur side-by-side, where as on
the hard drive read and write would have to be queued.

(Temp files on flash drive)
USB ++++*
HD ----------

(Temp files on hard drive)
USB
HD --o++-------

o = overhead for extra seeks associated with having the temp files on the
hard drive.

* I put in an extra ++ for temp files on flash drive since writes are
slower and also to illustrate why it doesn't matter.


I did think up of what I thought was the best way to benchmark and
illustrate the performance difference. It requires the setup of identical
hardware to isolate any variables down to just the temp file locations.
It would be preferable to have one set of input devices for both computers
so they are getting the same input at the same time. But I just don't
have that kind of money.

Anyways, even if I did tons of benchmarks on MY hardware. It doesn't mean
you will get the same results on YOUR hardware. You will just have to
test out the idea for yourself, unless you don't want to find out on your
own.

What you describe is some general consideration about a benchmark test. It
is not a step-by-step recipe that anyone can test on his own machine to
verify your claims. To stand up to scrutiny, your test would need to consist
of a detailed set of precise instructions. Let's see them!

 

[Curious]

I have never seen any evidence or even ever heard of Windows creating any
"Temp" files on any drive when loading an application. If you run Excel it
loads excel.exe in memory and then loads or creates a new .xls spreadsheet
file in memory there are no other/temp files created.

 

[Gordon]

I have never seen any evidence or even ever heard of Windows creating any
"Temp" files on any drive when loading an application. If you run Excel
it loads excel.exe in memory and then loads or creates a new .xls
spreadsheet file in memory there are no other/temp files created.

If you open a Word Document, Word will create a temp copy of the file in the
same folder that the original document exists. That's one of the main
reasons why users are advised NOT to open Word documents direct from
removable media....

 

[Curious]

I understand that current releases of some office products create temp
versions of their own documents when you start editing them. But it is the
application program(Word) creating them and not Windows itself as Tae Song
claimed.

 

[Gord Dibben]

Excel does the same thing.

A temp file is created for each workbook opened.

When the workbook is closed the temp file is deleted.............most
times<g>

With a workbook open browse to

C:\Documents and Settings\username\Application Data\Microsoft\Excel

You will find an ~123x456.xar file


Gord Dibben MS Excel MVP

 

[Curious]

You are correct. I should not have used an Office product as an example of
what happens with any Windows based application since the Office
applications themselves create temp files as you describe.

 

[Patrick Keenan]

You don't need an extremely high write speed. A lot of times temp files
are just empty files, many are 0 bytes. Almost all are under 700KB. Even
at a write speed of of say a low of 5MB/s is still only a fraction of a
sec.

This keeps the read/write head from thrashing about creating and updating
file records.

And just to up the ante, I enabled disk compression on the USB drives to
reduce the size of the writes.

Reducing the size of the writes won't affect the time it takes and certainly
will not alter the fact that Flash technology has a limited number of write
cycles. If you're using it as a temp drive, you are ensuring that a flash
drive will fail *sooner* rather than later.

Flash drives aren't appropriate for filesystem utility use. They can only
be relied on for convenient transfer of data that exists elsewhere.