interesting

If you have the 9800gt ee version could you test that also.I bet that would be top dog on your list.

This one
http://www.newegg.com/Product/Product.aspx?Item=N82E16814133279

rklapp
So I'm playing with my new wattmeter and here's what I found. I believe that if I divide the PPD by watts

What system hardware specs did you test with?  Someone with a different configuration but the same cards may not get the sale results of PPD/Watt as you have.  Also, which wattage?  If you have an outlet power meter, you're including the power the PSU uses to convert AC to DC that shouldn't count.  
 
For example, if you read 400W from an outlet using an 80% eff PSU, actual system wattage is 320W.  This is important because not everyone owns the same PSU with the same efficiency rating, as ajr23 mentioned.  An 87% eff PSU would read 367W from the wall, while a 75% eff PSU would read 426W from the wall.
 
This would either make those numbers you gather more appealing to someone with a better PSU to invest in more graphics cards for folding, than someone with a less efficient PSU to go out and waste more power for the same task.

lehpron
What system hardware specs did you test with?  Someone with a different configuration but the same cards may not get the sale results of PPD/Watt as you have.  Also, which wattage?  If you have an outlet power meter, you're including the power the PSU uses to convert AC to DC that shouldn't count.  

With the wattmeter plugged into the outlet, I turned on each client and measured against the baseline. I agree that it's difficult to compare system to system. The variables are the other components running and the PPD for each client.
 
If I run them all and shut down each one at a time, I get the following in this order.
 
9800 = 51W and 98 rklapps
 
250 = 94W and 64 rklapps
 
470 = 187W and 72 rklapps
 
smp = 115W and 57 rklapps
 
Again, the 9800gt comes out on top.

Punchy

Baseline being an idle system?

Yes. I give it half a minute for the wattage to stabalize before measuring.

Understanding there can be differences (%+- in accuracy) in the wattage readers we are using, I'm seeing the following with my EVGA GTX470 SC+ (default clock) and using MSI Afterburner to control fan speeds.

PC at Idle before starting the GPU Folding client = 194W

PC at the point the first 250 (out of a 25000 total) job starts = 338W

PC a few minutes after the job has been running (settled out) = 345W

So the maximum that my GTX470 is drawing as relates to running the Folding GPU client is 151W.

Rick

i think i have 2 of his 9600's, one died, but the other guy is folding nicely at a steady 3k ppd..

i couldn't tell ya 

so then, my 9600 has a rklapp rating of 28.57 right?

700_up 
 I'm very interested what the 460 will get as far as performance per watt. 

When I added to gtx460s to my rig it's added 280W for 20-21K (slightly overclocked)

rklapp
If I run them all and shut down each one at a time, I get the following in this order.
 
9800 = 51W and 98 rklapps
 
250 = 94W and 64 rklapps
 
470 = 187W and 72 rklapps
 
smp = 115W and 57 rklapps 

I measured them again in reverse order, shutting down each at a time.
 
smp = 81W and 80 rklapps
 
470 = 200W and 67 rklapps
 
250 = 93W and 64 rklapps
 
9800 - 57W and 87 rklapps
 
Same results except that I'm surprised the smp is using less watts this time. Not sure why. Also, I noticed that the 470 systray client still uses 127W if placed on pause instead of quit. Does the systray usually use energy when on paused?

rklapp

I'm running at 1450 shaders compared to the default 1250. When I return from my trip, I'll have to compare OC to default. Perhaps that's why I'm getting 30W more.

It looks like the change in the shaders increases by 18 watts.