I\'m just wondering, because CPDN crashed often on my PC I clocked it a bit down, but the s per timestep where rising. Maybe this has something to do with the CPU temperature? At first I run my PC hat 2300 HHz (AMD Athlon XP Barton 2800 ->2080 MHz). Then I clocked first down. THe bencmarks sunk. Then I clocked once more down, the bencmarks sunk but the time which is nessecary to run a S@h wu sunk also.

Can somebody help?
Greetings from Germany!
Basti

The seconds per timestep measure is difficult to interpret because it is the average over the whole WU, and you will get some variation in speed anyway over the life of a model. The only way to measure current speed is to take the elapsed time between the last two trickles and divide it by 10802 (the number of timesteps). That assumes, of course, that you have not made any changes between trickles. The lower the number, the faster you are processing the model.

Climate modelling is a very intensive application, and this makes overclocking difficult. CPDN requires absolute stability over a long time, and CPU temperature, memory timings etc can all disturb that. There has been much discussion of this sort of thing on the <a href="http://www.climateprediction.net/board/index.php">PHP board</a>.

Thx for your answere!


Greetings from Germany!
Basti

&gt; Thx for your answere!
&gt;
&gt;
&gt; Greetings from Germany!
&gt; Basti
&gt;
The short answer is the Seconds / Time Segment is a measure of how fast you are crunching, and lower numbers indicate higher speed. But stability is more important.
<a href="http://www.boinc.dk/index.php?page=user_statistics&amp;project=cpdn&amp;userid=13"><img border="0" height="80" src="http://13.cpdn.sig.boinc.dk?188"></a>