Well, my first R2 wu finished on my 2.8GHz P4 in 82400 seconds, (22.8 hours), and has claimed 259 credit, (not granted yet). This gives 11.32 credits per hour, whereas this machine normally earns 15.3 credits per hour with R1 wu's.
From a purely runtime perspective, at least one of my machines will have to stop crunching Einstein.
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.
First 3 done, 1 granted 135.47 for 20.82 hours. The other 2 are pending. They claimed the same credit as the first one.
This machine was doing 18-18.5 WU's per day with HT on, and 17-17.5 with HT off. RAC was in the 950-980 range on R1's. If the R2's are all like the first 4, and I know they will not be, you are looking at about 625 a day for 4 WU's per day, plus a little on the next 4.
From a purely runtime perspective, at least one of my machines will have to stop crunching Einstein.
As long as the deadlines remain roughly proportional to the crunching time, what difference does it make how long the individual tasks take? Taking arbitrary numbers out of the air, suppose an old and sporadically available system could only be sure of completing one ten-hour WU in two weeks. Shouldn’t it be equally capable of finishing a forty-hour run in two months?
From a purely runtime perspective, at least one of my machines will have to stop crunching Einstein.
As long as the deadlines remain roughly proportional to the crunching time, what difference does it make how long the individual tasks take? Taking arbitrary numbers out of the air, suppose an old and sporadically available system could only be sure of completing one ten-hour WU in two weeks. Shouldn’t it be equally capable of finishing a forty-hour run in two months?
Deadlines aren't changing. I'm now getting a 149 hour item, still two weeks.
Is Chilango's result (the only one I've seen linked so far) likely to be typical in terms of production run-time - i.e. 5.4 times S5RI run-time, in his estimation? (I stress production run-time, since it looks as if it still has a lot of debug output).
If so, you might like to consider whether 14 days is still the appropriate deadline: my Celeron, allowing for its current 50% share with SETI, would take about 17 days pro-rata. Not that the project should be run for the benefit of one superannuated cruncher, of course!
Our goal is run times in the range from 6 to 24 hours. Note that some optimization is expected in the future so the apps will get faster.
Cheers,
Bruce
I'm getting an estimate of 37+ hours to do the one I just received today. I think my average turnaround (distorted by some outages) was around 4 days for WU that were estimated to take 11-13 hours, so this should be interesting.
I just tried something. This is on an IBM x360, four 1.9Ghz HT MP Xeons 512 L2 and 1024 L3 cache, 2GB of ECC DDR2100 ChipKill.
Turned off HT, watched the time to completion rise as it had been with HT on. I then set processor affinity for each WU and noticed that the rate of climb started to fall off, by 12sec per 1 minute of run time. These 4 WU's are between 4% and 15% done.
Anyone with an SMP box give it a try and see if you come up with the same kind of results.
Actually, that's not too surprising. HyperThreading can help speed up server applications, but it can slow down most other applications. So, for desktop use, you're probably better off with HyperThreading turned off.
Deadlines aren't changing. I'm now getting a 149 hour item, still two weeks.
That’s another story, then. While it’s my impression that adjusting deadlines is a fairly simple tweak to the severs; perhaps they just haven’t got around to it yet. This project has a history of being pretty accommodating to slower systems (witness the tendency for them to be sent shorter tasks in the previous run), so I doubt that setting will remain miscalibrated for long.
My Celeron 433 got an extra long unit(243.10 Hz) that is supposed to take 9d 8h 47m 19s to complete. *lol*
And yes, I know there are much shorter units. But don't forget the people who crunch for more than one project. They may face serious problems with the current deadline.
Actually, that's not too surprising. HyperThreading can help speed up server applications, but it can slow down most other applications. So, for desktop use, you're probably better off with HyperThreading turned off.
This box was doing 18-18.5 S5R1's a day with HT on. It did about 1 less with HT off.
Ok, my result ran for 18
)
Ok, my result ran for 18 hours and a bit. So not bad on a P4 2.8GHz.
Well, my first R2 wu finished
)
Well, my first R2 wu finished on my 2.8GHz P4 in 82400 seconds, (22.8 hours), and has claimed 259 credit, (not granted yet). This gives 11.32 credits per hour, whereas this machine normally earns 15.3 credits per hour with R1 wu's.
From a purely runtime perspective, at least one of my machines will have to stop crunching Einstein.
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.
First 3 done, 1 granted
)
First 3 done, 1 granted 135.47 for 20.82 hours. The other 2 are pending. They claimed the same credit as the first one.
This machine was doing 18-18.5 WU's per day with HT on, and 17-17.5 with HT off. RAC was in the 950-980 range on R1's. If the R2's are all like the first 4, and I know they will not be, you are looking at about 625 a day for 4 WU's per day, plus a little on the next 4.
RE: From a purely runtime
)
As long as the deadlines remain roughly proportional to the crunching time, what difference does it make how long the individual tasks take? Taking arbitrary numbers out of the air, suppose an old and sporadically available system could only be sure of completing one ten-hour WU in two weeks. Shouldn’t it be equally capable of finishing a forty-hour run in two months?
RE: RE: From a purely
)
Deadlines aren't changing. I'm now getting a 149 hour item, still two weeks.
RE: RE: Thanks to Bernd &
)
I'm getting an estimate of 37+ hours to do the one I just received today. I think my average turnaround (distorted by some outages) was around 4 days for WU that were estimated to take 11-13 hours, so this should be interesting.
RE: I just tried something.
)
Actually, that's not too surprising. HyperThreading can help speed up server applications, but it can slow down most other applications. So, for desktop use, you're probably better off with HyperThreading turned off.
RE: Deadlines aren't
)
That’s another story, then. While it’s my impression that adjusting deadlines is a fairly simple tweak to the severs; perhaps they just haven’t got around to it yet. This project has a history of being pretty accommodating to slower systems (witness the tendency for them to be sent shorter tasks in the previous run), so I doubt that setting will remain miscalibrated for long.
My Celeron 433 got an extra
)
My Celeron 433 got an extra long unit(243.10 Hz) that is supposed to take 9d 8h 47m 19s to complete. *lol*
And yes, I know there are much shorter units. But don't forget the people who crunch for more than one project. They may face serious problems with the current deadline.
cu,
Michael
RE: Actually, that's not
)
This box was doing 18-18.5 S5R1's a day with HT on. It did about 1 less with HT off.