Well, the benchmark does seem low- my Core Duo at 1.6 GHz can do better (just not atm since I'm running one of those ancient Linux core clients which always give poor benchmark results while running fine otherwise). Which, of course, leads to the question if maybe Windows 2k8 simply has the same "problem", namely benchmarking low without that affecting actual crunching speed (which wouldn't matter if you run fixed-credit projects only).
Then of course there has to be another reason for the poor Einstein performance. And yes, I would say the times mentioned by James are definitely way too long. Here, too, I get much better results from my Core Duo, which shouldn't really be the case... the Core architecture sure is good, but its not THAT good, especially since his machine is running at almost a GHz more than mine. Even before I started trying out the power user app I never got completion times of more than 80k seconds..
What it really comes down to is that we need results from other projects to compare with. James, I'm running SETI too, so if you could provide a link to a finished SETI WU from that box we would see if your box is just as slow there, or completely owns mine (which it should).
The reason I have not unmasked my computers is that this one is the only one that has no live data or development work on it while the others have a great deal of litigation material on them - I am an expert witness with lots of evidentiary material that MUST remain confidential.
The reason I have not unmasked my computers is that this one is the only one that has no live data or development work on it while the others have a great deal of litigation material on them - I am an expert witness with lots of evidentiary material that MUST remain confidential.
Well, I doubt that is really necessary, since all we would be able to see is your boxes performance data and tasks. We couldn't look at your hard drive or sth, so the data there would still be well protected. Still, hiding your boxes is of course your own decision which I thoroughly respect.
As for the SETI results- your box seems to do a good deal better than mine over there; about 14 creds/hour compared to 9-11 for mine. I don't know if the difference is quite big enough (maybe the Linux SETI app is a little better, or Longhorn hinders performance with the Windows app a bit) but it certainly is there. So, I would guess that your box/OS has trouble with the Einstein app...
There is only a single SETI completed result on display there at the moment.
It shows that a result with angle range .008595 reported taking 16186 CPU seconds.
For Angle Ranges in that neighborhood, my Q6600 requires about 5800 CPU seconds. Adjusting for our clock rate difference would make that equivalent to 7264 seconds at your clock rate. I'm running a "Rev: (R-2.4|xT|FFT:IPP_SSE3|Ben-Joe)" optimized ap, which may be more efficient than your "setiathome_enhanced 5.27 DevC++/MinGW", but not by that ratio.
I'd invite comment from others, especially those familiar with your version, but it appears SETI is also running too slow on this host, though by a smaller proportion on this single instance than your Einstein results.
Quote:
The reason I have not unmasked my computers is that this one is the only one that has no live data or development work on it while the others have a great deal of litigation material on them - I am an expert witness with lots of evidentiary material that MUST remain confidential.
Unmasking your BOINC results does not expose your IP to others, nor does it expose your host's network name. Those two things show up on the computer summary page only for the person logged in to your account.
You can check what is visible by looking at the result links you have provided on a machine where you are not logged in. I think your belief that you are providing protection to your evidentiary materials by masking your hosts is mistaken.
There's a current discussion thread at BOINC Dev, where Rom Walton (one of the main BOINC developers) has asked for a consultation by telephone with a user who's having problems with Windows Server 2008 "Longhorn" - most unusual. That's in connection with one of the later test releases of BOINC, whereas James is using the recommended v5.10.30, but it suggests to me that the OS may be the culprit.
Also, BOINCstats says that there are only 57 hosts running Windows Server 2008 "Longhorn" attached to Einstein at the moment, so there is very little prior experience to go on. Anyone know the host IDs for any of the other 56 boxes, so we can see how they're getting on?
There is only a single SETI completed result on display there at the moment.
It shows that a result with angle range .008595 reported taking 16186 CPU seconds.
For Angle Ranges in that neighborhood, my Q6600 requires about 5800 CPU seconds. Adjusting for our clock rate difference would make that equivalent to 7264 seconds at your clock rate. I'm running a "Rev: (R-2.4|xT|FFT:IPP_SSE3|Ben-Joe)" optimized ap, which may be more efficient than your "setiathome_enhanced 5.27 DevC++/MinGW", but not by that ratio.
I'd invite comment from others, especially those familiar with your version, but it appears SETI is also running too slow on this host, though by a smaller proportion on this single instance than your Einstein results.
I've got a lot of stored data on computers running the stock SETI applications. I would think that a reasonable estimate would be very close to 10,000 seconds (or just slightly below) for a Q6600 running a more established OS (XP or 32-bit Vista).
There is only a single SETI completed result on display there at the moment.
It shows that a result with angle range .008595 reported taking 16186 CPU seconds.
...
This is all very strange since when I display host performance I show that the SETI project has received over 400 credits on this system (that looks like about 5 completed work units, no?):
This is all very strange since when I display host performance I show that the SETI project has received over 400 credits on this system (that looks like about 5 completed work units, no?):
If you are asking why you don't see 5 tasks sitting in the list there, then that's easily explained. As soon as tasks have been returned by the multiple computers in the quorum and credit has been granted, their Task IDs are removed from the database.
The data is stored separately. It's no use to store the number of tasks you did as well as you got credit for them.
Thank you. A returning newbie shows his re-learning curve.
BTW, That host is about to report a completed SETI work unit (5.27) and it took 142 hours, 50 minutes and 39 seconds of CPU time to complete. Does that sound reasonable for this system? If 'Longhorn' is the problem I really have no alternative - I must use that OS on that system (it's the only 64 bit OS I have) in order to access the full 5 gig of ram to, in turn, run a very large collision reconstruction model I built for my work. So, I may just quit throwing cpu cycles at boinc projects on that system instead of burning up the electricity.
Well, the benchmark does seem
)
Well, the benchmark does seem low- my Core Duo at 1.6 GHz can do better (just not atm since I'm running one of those ancient Linux core clients which always give poor benchmark results while running fine otherwise). Which, of course, leads to the question if maybe Windows 2k8 simply has the same "problem", namely benchmarking low without that affecting actual crunching speed (which wouldn't matter if you run fixed-credit projects only).
Then of course there has to be another reason for the poor Einstein performance. And yes, I would say the times mentioned by James are definitely way too long. Here, too, I get much better results from my Core Duo, which shouldn't really be the case... the Core architecture sure is good, but its not THAT good, especially since his machine is running at almost a GHz more than mine. Even before I started trying out the power user app I never got completion times of more than 80k seconds..
What it really comes down to is that we need results from other projects to compare with. James, I'm running SETI too, so if you could provide a link to a finished SETI WU from that box we would see if your box is just as slow there, or completely owns mine (which it should).
Greetings, Annika
Here is a SETI page for this
)
Here is a SETI page for this computer as requested: http://setiathome.berkeley.edu/show_host_detail.php?hostid=4161455
The reason I have not unmasked my computers is that this one is the only one that has no live data or development work on it while the others have a great deal of litigation material on them - I am an expert witness with lots of evidentiary material that MUST remain confidential.
RE: Here is a SETI page for
)
Well, I doubt that is really necessary, since all we would be able to see is your boxes performance data and tasks. We couldn't look at your hard drive or sth, so the data there would still be well protected. Still, hiding your boxes is of course your own decision which I thoroughly respect.
As for the SETI results- your box seems to do a good deal better than mine over there; about 14 creds/hour compared to 9-11 for mine. I don't know if the difference is quite big enough (maybe the Linux SETI app is a little better, or Longhorn hinders performance with the Windows app a bit) but it certainly is there. So, I would guess that your box/OS has trouble with the Einstein app...
RE: Here is a SETI page for
)
There is only a single SETI completed result on display there at the moment.
It shows that a result with angle range .008595 reported taking 16186 CPU seconds.
For Angle Ranges in that neighborhood, my Q6600 requires about 5800 CPU seconds. Adjusting for our clock rate difference would make that equivalent to 7264 seconds at your clock rate. I'm running a "Rev: (R-2.4|xT|FFT:IPP_SSE3|Ben-Joe)" optimized ap, which may be more efficient than your "setiathome_enhanced 5.27 DevC++/MinGW", but not by that ratio.
I'd invite comment from others, especially those familiar with your version, but it appears SETI is also running too slow on this host, though by a smaller proportion on this single instance than your Einstein results.
Unmasking your BOINC results does not expose your IP to others, nor does it expose your host's network name. Those two things show up on the computer summary page only for the person logged in to your account.
You can check what is visible by looking at the result links you have provided on a machine where you are not logged in. I think your belief that you are providing protection to your evidentiary materials by masking your hosts is mistaken.
RE: Here is a SETI page for
)
Hm...14.2 credits/CPU hour isn't really that hot for a C2Q either. This PC seems to be running with the hand-brake engaged.
CU
Bikeman
There's a current discussion
)
There's a current discussion thread at BOINC Dev, where Rom Walton (one of the main BOINC developers) has asked for a consultation by telephone with a user who's having problems with Windows Server 2008 "Longhorn" - most unusual. That's in connection with one of the later test releases of BOINC, whereas James is using the recommended v5.10.30, but it suggests to me that the OS may be the culprit.
Also, BOINCstats says that there are only 57 hosts running Windows Server 2008 "Longhorn" attached to Einstein at the moment, so there is very little prior experience to go on. Anyone know the host IDs for any of the other 56 boxes, so we can see how they're getting on?
RE: RE: Here is a SETI
)
I've got a lot of stored data on computers running the stock SETI applications. I would think that a reasonable estimate would be very close to 10,000 seconds (or just slightly below) for a Q6600 running a more established OS (XP or 32-bit Vista).
RE: RE: RE: Here is a
)
RE: RE: http://setiathome
)
If you are asking why you don't see 5 tasks sitting in the list there, then that's easily explained. As soon as tasks have been returned by the multiple computers in the quorum and credit has been granted, their Task IDs are removed from the database.
The data is stored separately. It's no use to store the number of tasks you did as well as you got credit for them.
Thank you. A returning
)
Thank you. A returning newbie shows his re-learning curve.
BTW, That host is about to report a completed SETI work unit (5.27) and it took 142 hours, 50 minutes and 39 seconds of CPU time to complete. Does that sound reasonable for this system? If 'Longhorn' is the problem I really have no alternative - I must use that OS on that system (it's the only 64 bit OS I have) in order to access the full 5 gig of ram to, in turn, run a very large collision reconstruction model I built for my work. So, I may just quit throwing cpu cycles at boinc projects on that system instead of burning up the electricity.