Is my machine underperforming?

[alpha]

Hey all,

I've just picked up SoB again, and am nearly halfway through a full test. Just wondering if my machine is underperforming - how does this sound:

XP 1700+
192MB PC133
Windows XP Pro
k = 4847
n = 3824607

cEMs/sec peaked at the beginning of the test at around 166 kcEMs/sec and is now at 144 kcEMs/sec and hovering.

Maybe I'm just being paranoid about seeing other peoples numbers on other k's and n's but is this machine underperforming? It's been tweaked to the max, so save the 'background tasks eating CPU' suggestions ;)

[mklasson]

That does seem a little low. I've got an XP1800+ running on a linux system, getting 270K cEMs/s. That machine has DDR ram though, which probably makes a significant difference. Another machine, 750 MHz athlon with PC133 and win2k, peaks at around 120K. I would think your machine should score about twice that.

Mikael

[alpha]

But don't the k and/or n values affect the cEMs/sec rate? So unless your 750MHz Athlon has a test with similar(?) k and/or n values to my machine, they are incomparable... no? If not, does anyone have any suggestions?

Also I should mention that on my personal user stats, the equivalent power (last day) recently was around 2GHz, which is way over the 1.46GHz that the XP 1700+ runs at. So maybe everything is OK afterall.

[jjjjL]

the GHz estimate is in P3s so that's why most people see a higher GHz equiv then they'd expect.

-Louie

[alpha]

Ah OK - thanks Louie.

And in response to my own post, I left the machine alone for a while and it crept back up to 160 kcEMs/sec, so the performance hasn't dropped drastically throughout the test as I made out before :)

[mklasson]

Originally posted by alpha
But don't the k and/or n values affect the cEMs/sec rate? So unless your 750MHz Athlon has a test with similar(?) k and/or n values to my machine, they are incomparable... no? If not, does anyone have any suggestions?
Yes, the n value certainly affects the rate, but not terribly much for roughly similar values. The 120K figure is for n around 3.8M. The computer also has to be left fairly alone without a bunch of programs running, otherwise the rate [naturally] drops.

Mikael

[alpha]

Originally posted by mklasson
Yes, the n value certainly affects the rate, but not terribly much for roughly similar values. The 120K figure is for n around 3.8M.

Ah OK. Then it suggests something is not as it should be.


The computer also has to be left fairly alone without a bunch of programs running, otherwise the rate [naturally] drops.

Mikael

Of course, and I thought this was so when I observed 144 kcEMs/sec, but I was mistaken.

Obviously the cEMs/sec section of the client does not display an instantaneous value - so how about having this toggleable in a future version, Louie? You could then choose between average (which I assume is currently being used) and instantaneous.