A follow up to my original post: In late March the performance of this machine went up dramatically in terms of crunching rate. I had complained originally that it seemed to be under performing, crunching about 3600 credits per day on Einstein. Since that time, the computer has increased its rate to averaging over 5500 credits/day for the last 17 days. I originally expected more than twice the rate of my son's Q6600 system and now am getting slightly more than twice his system's credit rate.
Because of Einstein's credit temporal variability, I have been waiting for this "peak" behavior to dip back down to the lower rate but the current behavior seems not to be a statistical variation. I am puzzled by this change but certainly not unhappy about it either.
A follow up to my original post: In late March the performance of this machine went up dramatically in terms of crunching rate. I had complained originally that it seemed to be under performing, crunching about 3600 credits per day on Einstein. Since that time, the computer has increased its rate to averaging over 5500 credits/day for the last 17 days. I originally expected more than twice the rate of my son's Q6600 system and now am getting slightly more than twice his system's credit rate.
The two new science apps ABP2 and S5GCE both earn significantly higher credit than S5R6 did on intel systems. AMD systems OTOH saw almost no gain. I don't know why the new app is doing so much better on intel, but the reason for intel gaining instead of amd losing is probably due primarily to E@H using AMD systems internally.
A follow up to my original
)
A follow up to my original post: In late March the performance of this machine went up dramatically in terms of crunching rate. I had complained originally that it seemed to be under performing, crunching about 3600 credits per day on Einstein. Since that time, the computer has increased its rate to averaging over 5500 credits/day for the last 17 days. I originally expected more than twice the rate of my son's Q6600 system and now am getting slightly more than twice his system's credit rate.
Because of Einstein's credit temporal variability, I have been waiting for this "peak" behavior to dip back down to the lower rate but the current behavior seems not to be a statistical variation. I am puzzled by this change but certainly not unhappy about it either.
Thanks for all the help,
Rod
RE: A follow up to my
)
The two new science apps ABP2 and S5GCE both earn significantly higher credit than S5R6 did on intel systems. AMD systems OTOH saw almost no gain. I don't know why the new app is doing so much better on intel, but the reason for intel gaining instead of amd losing is probably due primarily to E@H using AMD systems internally.