Well I have had a peek and its an Intel Celeron 1.7GHz thing. Ho hum, still it was free so I shouldn't complain.

I noiticed .

I also noted that I am getting about 1.9 times up lift with Boinc as opposed to David's 2.5 times. But my box has a 112 UD performance rating and Davids has boxes range from 126-156. Which to me suggests the difference between the two points awarding systems is not a straight fixed factor and given your evidence it would appear that there is a cross overpoint.

What would be useful to know Ady is what is the perfromance rating UD gives your machines, for the more data that can be plotted the greater the chance of seeing whether the factor is linear.

Just for reference:
AMD AlthonXP 2000+:

• Windows XP UD: 112
  
• Windows XP/Linux Boinc: 1563 Whetstone, 2627 Dhrystone
  

Intel Celeron 1.7Hz:

• Linux Boinc: 434 Whetstone, 1189 Dhrystone
  

Given that the Boinc scoring system uses a product of the CPU time and the Whetstone and Dhrystone Benchmarks, this would explain the difference between the average points per work unit returned between my AMD and Intel boxes given their differing Whetstone:Dhrystone ratios.

Also given that Boinc scoring system only awards the points when the same WU has been returned by several Computers and has been validated, and then the lowest score is awarded to all of computers, it would explain why I am seeing no major fluctuation in the average points per work unit returned for Intel Celeron box (low scoring), and a fall off for the AMD AthlonXP box (high Scoring) 390 yesterday to 340 today. This lowest points scored system means that with a grid of predominantly high spec machines running Boinc you would expect predominently higher points returns than if you have a grid of predominently low spec machines.

If my logic is correct, the nett result is that if the machine mix of the Boinc grid changes to more higher spec machines you are going to get an increase in the points per unit returned. As a result as the general performace of the grid improves the conversion factor between the UD points scheme and the Boinc points scheme will drift out for high spec machines, meaning that they scored higher points with Boinc than UD.

Edit ---------------
*Back from lunch*

However the high scoring machines will occasionally get a low score for a work unit, when a low scoring machine is part of the comparission group, but the low scoring machines will always get a low score. Thus altering the UD:Boinc scoring correction factor whilst reining in the scores of the high scoring machines and reseting the nominal power of the grid, will penalise the low scoring machines.

Thanks for this Batchoy - very valid points (pun?) - something which I had forgotten about, and indeed effects the point scoring.

One slight thing to note, that in the quorum of results required for validation, the top and bottom scores are discarded, and the average of the middle two used as the score awarded to all 4, rather than the lowest.

However, as you rightly say, if you have 3 high performance machines together with 1 low as the quorum, then the low scoring machine should get a higher than usual score. Likewise if there are 3 low performing machines and 1 high in the quorum, then the high performers score will suffer.

The 'Credit Claimed' and 'Granted' scores can be seen under the boinc results status on your WCG page.

I notice whilst browsing through some of mine, that occasionally I get a full 100% credit granted, more often around 50% and sometimes lower. All this needs to be taken into consideration when trying to compare Boinc with UD which makes things even more complicated!!

One thing to note, which will help, is that in about a weeks time a page is going to be added which lets us see the status of the other units required for your validation:


Ady

I counted 25 out and 25 results came back - that put's the double accounting theory to bed

I checked out the credit granted and most of the time I'm in the 90-100% bracket maybe my 2800+-3200+'s are really old hat now

It's looking like it really is impossible

I'm now getting an average of 72.56 points per hour with BOINC whereas my current average hourly score is 28.78 and I started crunching with just a 3200+ and have added 3000+ and 2800+'s to the mix since then so even 28.78 is not a true reflection of my crunching output today, it is slightly high.

Looks like we might have lost this battle but will win the war on some of Mankind's most serious ailments.

It is a pity though that this element of fun through the comparing of like with like will be lost with the introduction of the new crunching system

I would still like to know how we are managing to crunch more Work Units with BOINC than the general UD population too

Looks like the answer to your question Bill is "they are not"



Dave