The SETI-Spy processing time calculator

I'm puzzled by the result I get when I plug in my original configuration
to the "processor efficiency" calculator at

http://cox-internet.com/setispy/efficiency.htm

Here's what I entered:

Processor type Intel Pentium II/III (512 kB L2)
Processor speed 500 MHz
Memory speed 100 MHz
Angle range 0.417 degree

With these values, I get an estimated completion time of 11.3 hours.

My real-world experience with this same configuration is that it took
more than three times that long to complete a work unit -- almost 44 hours,
in fact, averaged over my first four WUs.

For my next ten WUs, with 512 MB of RAM vs. 128 MB, the average was near
40 hours.

My question is: Am I making the wrong choices for inputs to the
calculator, or does it have a bug in it?

Thanks,

Chris

This sounds like you are using the GUI client with graphics turned on. Turn
off the graphics or switch to the CLI client.

Mike Ober.

"Christopher P. Winter" wrote in message
...
I'm puzzled by the result I get when I plug in my original
configuration
to the "processor efficiency" calculator at

http://cox-internet.com/setispy/efficiency.htm

Here's what I entered:

Processor type Intel Pentium II/III (512 kB L2)
Processor speed 500 MHz
Memory speed 100 MHz
Angle range 0.417 degree

With these values, I get an estimated completion time of 11.3 hours.

My real-world experience with this same configuration is that it took
more than three times that long to complete a work unit -- almost 44
hours,
in fact, averaged over my first four WUs.

For my next ten WUs, with 512 MB of RAM vs. 128 MB, the average was
near
40 hours.

My question is: Am I making the wrong choices for inputs to the
calculator, or does it have a bug in it?

Thanks,

Chris

Message-ID: . net,
"Michael D. Ober " wrote:

This sounds like you are using the GUI client with graphics turned on. Turn
off the graphics or switch to the CLI client.

The other thing that will cause this sort of discrepancy is motherboard
based graphics. These are often included in low end machines because they
reduce costs without reducing the numbers (processor clock and memory size)
on which the average consumer buys. They do have a very severe impact on
real performance.

Cheap motherboards can also impact performance in other ways, but motherboard
graphics is the biggest hit.

Using a, for the time, high end motherboard (BX chipset), a 350MHz P II and
the Linux command line client, I seem to remember I would get a typical
processing time of about 13 hours. The memory was 128MB but nothing like
that was needed to run the SETI@Home code.

I haven't tried this, but, if you don't want to improve your OS, or motherboard,
you may want to try setting the machine into full screen console (DOS) mode
whilst running the client, as that may minimise the graphics load on the
main memory. (Don't boot in DOS mode. You need to start a "DOS" box and
then maximise it (ALT-Enter).)

"Christopher P. Winter" wrote in message
...

For my next ten WUs, with 512 MB of RAM vs. 128 MB, the average was
near

The client uses about 13MB. Once you have overcome your OS' overhead
(about 4MB for a custom Linux, maybe 100MB for XP), and any motherboard
graphics allowance (maybe 4MB) you only need an extra 13MB to run the
client at full efficiency.

On Sat, 12 Mar 2005 10:45:13 +0000, (David Woolley)
wrote:

Message-ID: . net,
"Michael D. Ober " wrote:

This sounds like you are using the GUI client with graphics turned on. Turn
off the graphics or switch to the CLI client.

The other thing that will cause this sort of discrepancy is motherboard
based graphics. These are often included in low end machines because they
reduce costs without reducing the numbers (processor clock and memory size)
on which the average consumer buys. They do have a very severe impact on
real performance.

Cheap motherboards can also impact performance in other ways, but motherboard
graphics is the biggest hit.

Using a, for the time, high end motherboard (BX chipset), a 350MHz P II and
the Linux command line client, I seem to remember I would get a typical
processing time of about 13 hours. The memory was 128MB but nothing like
that was needed to run the SETI@Home code.

I haven't tried this, but, if you don't want to improve your OS, or motherboard,
you may want to try setting the machine into full screen console (DOS) mode
whilst running the client, as that may minimise the graphics load on the
main memory. (Don't boot in DOS mode. You need to start a "DOS" box and
then maximise it (ALT-Enter).)

"Christopher P. Winter" wrote in message
...

For my next ten WUs, with 512 MB of RAM vs. 128 MB, the average was
near

The client uses about 13MB. Once you have overcome your OS' overhead
(about 4MB for a custom Linux, maybe 100MB for XP), and any motherboard
graphics allowance (maybe 4MB) you only need an extra 13MB to run the
client at full efficiency.

Thanks for the detailed explanation. Yes, I based my real-world numbers
on the "screen saver" version. I guess I missed the fact that the calculator
is intended for the CLI version.

Motherboard graphics is not a factor. My system is an HP Kayak XU 7/500,
with the 440BX chipset; for video it uses an ELSA Gloria Synergy AGP card
with 8MB of VRAM.

Actually, the bottleneck was the lack of a second processor. I first
boosted system RAM to 1GB (from the original 128MB) and got only a modest
increase in performance. Then I added a second 500MHz processor. That brought
my times down to the 12-13 hour range -- just what the calculator predicts.

It seems the screen-saver version of SETI@Home /does/ make good use of a
second processor.

Chris