savantcreative said:
I have had Windows XP Pro running on a machine for about 4 years and have
noticed a definite slow down. My system is a pretty high end w/raid array
and is virus and spyware clean. Should I try to clean the registry? If so
how? Should I reinstalled Windows? Will that speed it up? I also have a
program called PC Angel LE. Will using that clean my registry?
What is the best protocol to get a machine running like new again?
Where can I read about all of this?
Thank you so much for the answers to my many questions.
There are some things you can try on the hardware side.
First, get a copy of CPUZ.
http://www.cpuid.com/cpuz.php
It will tell you the speed at which your processor and
memory are running. In some cases, like a motherboard
with an Athlon or AthlonXP on S462 socket, the CPU input
clock can be set too low, resulting in lost performance.
That can happen sometimes, on a computer crash, followed
by the BIOS using its "safe" settings. A trip to the BIOS
is needed to correct it. And to know the correct value
to use, I have a table of values I got from the old
qdi.nl site.
This is a picture of the result for a P4 processor at 3.8GHz.
The FSB is FSB800. If I had a P4 3.8GHz, I'd be pretty
happy with the resulting display.
http://img.tomshardware.com/us/2004/11/15/3/cpuz_cpu.gif
There is 1GB of physical memory in the machine, running at DDR2-533.
(The frequency value should be doubled, to get the more familiar
rating.)
http://img.tomshardware.com/us/2004/11/15/3/cpuz_mem.gif
In Task Manager, you'd expect that P4 processor to have two
graphs. That implies the multiprocessor HAL was being used.
Sometimes, a user's machine is only using one core, instead
of all available virtual or physical cores.
http://www.r-tech.com/images/task_manager_image.jpg
To test the disks, you need a benchmarking tool. The
disk transfer rate varies over the surface of the disk,
and the disk tends to be "zoned". Mine transfers at
60MB/sec near the beginning of the disk, and about 30MB/sec
near the end. The benchmark of the disk, displays a curve with
stairsteps in it. When there are sectors substituted on the
disk, there can be downward spikes in the graph. I tested a
brand new drive, and it was spike-free. A disk which is
six months to one year old, is already showing spikes.
And that implies that occasionally, there will be some
additional delay to get data from the disk.
(Download the free one on the right.)
http://www.hdtune.com/download.html
Other useful functions are the "Info" tab, and the "Error Scan"
tab. If you have a simple IDE drive, the Info tab can tell you
what transfer mode the disk is in. Sometimes, a disk will be
operating in PIO mode, which gives a flat line in the benchmark
screen, and very low numbers. That is the usual reason for
performance complaints, that PIO is being used.
The "Error Scan" is something you should run, before subjecting a
disk to strenous activity. For example, if you were going to
defragment a drive, you'd want to use some tool to scan the
entire surface of the disk, to make sure it was healthy.
If any problems are detected, then running a defrag is asking
for trouble.
There can be occasional problems caused by driver updates, such
as getting drivers from Windows Update. The best place to get
drivers, is from the manufacturer's site. A driver updates are
generally only required, if there is some problem. If you play
a lot of games, there might be more need to update video card
drivers to the latest version, as there can be bug fixes in there.
But for a lot of other peripherals, a "cleaning fetish" is
not the right answer.
If you're certain the drive is healthy, there are no "SMART"
problems reports, and the surface scan is clean, you can try
defragmenting. That might give some performance.
The Task Manager will show what is currently running, and
you may see something using compute cycles in the display.
You might investigate what is going on there.
Benchmarking is also something you can do, but if you
didn't benchmark the machine when it was healthy, you
have no baseline against which to measure what is it
doing currently. To give an example, I use SuperPI for a
mainly compute-bound benchmark.
http://www.xtremesystems.com/pi/super_pi_mod-1.5.zip
(
http://www.xtremesystems.com/pi/ )
I usually use the "1 million digit" test, as a test case.
If the processor in question has an extremely large
cache, then a larger number of digits can be used.
I find the "32 million" setting is just too long to wait.
On my previous computer, when it was new, I got
45 seconds to compute 1 million digits. After I
added my antivirus software, this dropped to 50 seconds,
even though there was nothing obvious to see in
Task Manager. So the AV software seemed to be
sapping some of my performance. In that case,
you can see the value of me being lucky enough
to have run the program before the software
was run. Otherwise, I wouldn't have been in
a position to blame the AV software for it.
There are certainly other benchmarking suites, but
some of them mix both computing and disk aspects
together, which can make it harder to separate the
two aspects of performance. SuperPI is pure
computing, and HDTune concentrates on the disk.
Some things can be hard to quantify. For example,
a certain SIS chipset, caused a five second delay
(an internal timeout), every time the user tried
to do something with an attached disk. This was
some kind of driver problem. A disk benchmark
would not have shown a problem. But in terms of
interactive performance, such a behavior would
drive you nuts. (Imagine a five second delay,
every time you go to do something.) I'm not aware
of a test that checks responsiveness from
"click to results", so for that kind of problem,
I'd have to rely on a user describing it to me.
I'm not a big fan of "automated performance tools",
like buying a software for $39.95 from the Internet
and expecting it to fix everything. At the very
least, you'd want to research how many users had
machine disabled by the use of such a program,
before running it.
HTH,
Paul
