Message boards :
Number crunching :
time prosses of work unit wrong?
Message board moderation
Author | Message |
---|---|
guillermo Send message Joined: 2 Jun 99 Posts: 23 Credit: 9,007,580 RAC: 0 |
y see somthing Curious sence y put some laptop double proscesors (2.6GHz for each proscesor)to work y have: 5236 floating point MIPS (Whetstone) per CPU, 2534 integer MIPS (Dhrystone) per CPU, is more than the other cpus pentium 4 single procesor, but the initial time for work unit is 7 hours aprox, but when the work unit is in prosces the real time to complete are 2 hours, and y see that the initial time is the same all time, the progran cant calculate well? this is an error o this is ok? thanks sorry for my english SETIHOME USER SINCE 1999 |
Clyde C. Phillips, III Send message Joined: 2 Aug 00 Posts: 1851 Credit: 5,955,047 RAC: 0 |
El programa de BOINC adivina el tiempo de procesar cada unidad. Para las primeras unidades estas conjecturas muchas veces son incorrectas. Después de se calculen más unidades se serán más correctas las conjeturas. The Boinc program guesses the processing time of each unit. For the first units these guesses are often wrong. After more units are calculated the guesses will be more (nearly) correct. |
Alinator Send message Joined: 19 Apr 05 Posts: 4178 Credit: 4,647,982 RAC: 0 |
That's normal. For a variety of reasons, 'synthetic' benchmarks like the kind BOINC uses for a number of purposes don't really reflect the actual performance of a machine on a 'real world' application, like SAH WU's. This is why BOINC also tracks a metric called the Task Duration Correction Factor (TDCF) to compensate for that. However the host has to run a number of tasks in order for BOINC to collect enough information about the hosts actual performance to calculate this value. Normally it takes around 10 to 15 tasks for this to get in the ballpark, and then your runtime estimates should be much better. HTH, Alinator |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.