The informations given to the users at this project is far too low for me to accept. I see no rational reason to hold off one of the project members to give daily(or nearby) updates on what's going on. That's only 5 min. of time which the volunteers imho realy deserve.
FACK
solaris is like a wigwam: no windows, no gates and a apache inside !
The informations given to the users at this project is far too low for me to accept. I see no rational reason to hold off one of the project members to give daily(or nearby) updates on what's going on. That's only 5 min. of time which the volunteers imho realy deserve.
cu,
Michael
Yeah, i agree with that....
As i remember this was way better in the past, einstein was maybe the project with most information directly from the project leaders and i really liked that, but it has changed as time gone by.....
You do realize the drop from 100% in the middle of that graph signified the end of the old run?
I mean... wow.
I think what he means is the curving down of the second part of the graph (after the middle of April), it is not a straight line like one would expect. We should be closer to 50% done, rather than 45%.
There are 10^11 stars in the galaxy. That used to be a huge number. But it's only a hundred billion. It's less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers. - Richard Feynman
You do realize the drop from 100% in the middle of that graph signified the end of the old run?
I mean... wow.
I think what he means is the curving down of the second part of the graph (after the middle of April), it is not a straight line like one would expect. We should be closer to 50% done, rather than 45%.
It is summer in the Northern Hemisphere. There is often a reduction during this time of year due to students not being at school, people trying to conserve power, and people that don't have adequate enough cooling to be able to run as much...
The general topic is about poor performance caused by compiler switches and the
"dubious" method the run time lib (Microsoft) used to determine CPU properties.
Part 1 Question (General)
Should this be a separate thread for discussion, or does it only apply
to the S5R2 applications and work?
Part 2 Question (Specific)
I'm running windows 200 on a laptop - that shows:
6/16/2007 2:36:23 AM||Starting BOINC client version 5.8.16 for windows_intelx86
6/16/2007 2:36:23 AM||log flags: task, file_xfer, sched_ops
6/16/2007 2:36:23 AM||Libraries: libcurl/7.16.0 OpenSSL/0.9.8a zlib/1.2.3
6/16/2007 2:36:23 AM||Data directory: C:\\Program Files\\BOINC
6/16/2007 2:36:23 AM||Processor: 1 GenuineIntel Intel(R) Pentium(R) 4 Mobile CPU 1.70GHz [x86 Family 15 Model 2 Stepping 4] [fpu tsc sse mmx]
6/16/2007 2:36:23 AM||Memory: 254.98 MB physical, 981.45 MB virtual
((Yes that's not much memory - There is a problem with the motherboard detecting the second mem card..))
I have noticed that it takes me a LONG tome to do a WU:
137,895 second or 18.3 hours - This is cpu time elapsed time can be 2 to 3 days.
Is this part of the sse detection problem?
If so, is there a way to recompile and force a correction???
I use CYGWIN and use the GCC compiler for running other apps on w2k.
To the first part of your question: Yes, these problems afflict only the current S5R2 science app.
Second: Depending on the size of the WUs you get, this completion time does not seem particularly long to me. My Athlon 64 3500+ takes a little under 30 hours for a really big (530 credit) WU (18.3 creds/hour I think was the average) running Linux, so no penalty involved. My Core Duo (clock speed 1.6 GHz, but should be a bit more efficient per clock than yours) gives me about 15 creds/hour per core. You might want to calculate that for your box and compare.
So far there is no evidence of P4-type computers getting a significant penalty under Windows. You see it happen with newer AMD machines, and occasionally with P3s, Intel Mobiles and Cores, but never with a P4 so far. So don't worry too much.
Hope I was able to help.
Annika
The general topic is about poor performance caused by compiler switches and the
"dubious" method the run time lib (Microsoft) used to determine CPU properties.
Part 1 Question (General)
Should this be a separate thread for discussion, or does it only apply
to the S5R2 applications and work?
Well, it's probably better covered in the "Problems" section than in Crunchers' Corner, but .... nevertheless...
Quote:
Part 2 Question (Specific)
I'm running windows 200 on a laptop - that shows:
6/16/2007 2:36:23 AM||Starting BOINC client version 5.8.16 for windows_intelx86
6/16/2007 2:36:23 AM||log flags: task, file_xfer, sched_ops
6/16/2007 2:36:23 AM||Libraries: libcurl/7.16.0 OpenSSL/0.9.8a zlib/1.2.3
6/16/2007 2:36:23 AM||Data directory: C:\\Program Files\\BOINC
6/16/2007 2:36:23 AM||Processor: 1 GenuineIntel Intel(R) Pentium(R) 4 Mobile CPU 1.70GHz [x86 Family 15 Model 2 Stepping 4] [fpu tsc sse mmx]
6/16/2007 2:36:23 AM||Memory: 254.98 MB physical, 981.45 MB virtual
((Yes that's not much memory - There is a problem with the motherboard detecting the second mem card..))
I have noticed that it takes me a LONG tome to do a WU:
137,895 second or 18.3 hours - This is cpu time elapsed time can be 2 to 3 days.
Is this part of the sse detection problem?
If so, is there a way to recompile and force a correction???
I use CYGWIN and use the GCC compiler for running other apps on w2k.
THANKS IN ADVANCE,
Jay
This is kind of strange, it seems that your CPU is detected by BOINC (not the science app) as NOT being SSE2 capable (only SSE seems to be supported). AFAIK, all Pentium 4s are SSE2 capable, even the "mobile" variants.
If your CPU really does not support the SSE2 instruction set, the Microsoft Runtime lib inside the Windows science app will switch to a slower code path which cannnot be helped at the moment, except maybe by switching to Linux :-).
However, this does not explain fully the far from spectacular performance of this host ;-). The benchmarks don't look that hot either, so I'd guess that the notebook might be running in a power saving mode with reduced clock freq., or maybe your motherboard has some problems in addition to detecting the additional RAM. If you have other applications running at the same time, swapping might be an issue.
EDIT: Annika, you beat me.... I must type a bit faster, I guess..
CU
All the P4s under Win 2k are lacking the sse2 flag !
Very strange. The OS has to support SSE2 (AFAIK SSE2 is disabled by default and is switched on by the OS when booting, and the OS has to be aware of the SSE2 registers so they are saved and restored on task-switches). But Win 2k should support SSE2! Strange!
RE: The informations given
)
FACK
solaris is like a wigwam: no windows, no gates and a apache inside !
RE: The informations given
)
Yeah, i agree with that....
As i remember this was way better in the past, einstein was maybe the project with most information directly from the project leaders and i really liked that, but it has changed as time gone by.....
And the curve keeps
)
And the curve keeps flattening more and more....
Besides summer breaks... looks like more and more people is leaving....
Are we looking at teh same
)
Are we looking at teh same graph?
You do realize the drop from 100% in the middle of that graph signified the end of the old run?
I mean... wow.
RE: Are we looking at teh
)
I think what he means is the curving down of the second part of the graph (after the middle of April), it is not a straight line like one would expect. We should be closer to 50% done, rather than 45%.
There are 10^11 stars in the galaxy. That used to be a huge number. But it's only a hundred billion. It's less than the national deficit! We used to call them astronomical numbers. Now we should call them economical numbers. - Richard Feynman
RE: RE: Are we looking at
)
It is summer in the Northern Hemisphere. There is often a reduction during this time of year due to students not being at school, people trying to conserve power, and people that don't have adequate enough cooling to be able to run as much...
Hi, Here is a two part
)
Hi,
Here is a two part question:
The general topic is about poor performance caused by compiler switches and the
"dubious" method the run time lib (Microsoft) used to determine CPU properties.
Part 1 Question (General)
Should this be a separate thread for discussion, or does it only apply
to the S5R2 applications and work?
Part 2 Question (Specific)
I'm running windows 200 on a laptop - that shows:
6/16/2007 2:36:23 AM||Starting BOINC client version 5.8.16 for windows_intelx86
6/16/2007 2:36:23 AM||log flags: task, file_xfer, sched_ops
6/16/2007 2:36:23 AM||Libraries: libcurl/7.16.0 OpenSSL/0.9.8a zlib/1.2.3
6/16/2007 2:36:23 AM||Data directory: C:\\Program Files\\BOINC
6/16/2007 2:36:23 AM||Processor: 1 GenuineIntel Intel(R) Pentium(R) 4 Mobile CPU 1.70GHz [x86 Family 15 Model 2 Stepping 4] [fpu tsc sse mmx]
6/16/2007 2:36:23 AM||Memory: 254.98 MB physical, 981.45 MB virtual
((Yes that's not much memory - There is a problem with the motherboard detecting the second mem card..))
I have noticed that it takes me a LONG tome to do a WU:
137,895 second or 18.3 hours - This is cpu time elapsed time can be 2 to 3 days.
Is this part of the sse detection problem?
If so, is there a way to recompile and force a correction???
I use CYGWIN and use the GCC compiler for running other apps on w2k.
THANKS IN ADVANCE,
Jay
To the first part of your
)
To the first part of your question: Yes, these problems afflict only the current S5R2 science app.
Second: Depending on the size of the WUs you get, this completion time does not seem particularly long to me. My Athlon 64 3500+ takes a little under 30 hours for a really big (530 credit) WU (18.3 creds/hour I think was the average) running Linux, so no penalty involved. My Core Duo (clock speed 1.6 GHz, but should be a bit more efficient per clock than yours) gives me about 15 creds/hour per core. You might want to calculate that for your box and compare.
So far there is no evidence of P4-type computers getting a significant penalty under Windows. You see it happen with newer AMD machines, and occasionally with P3s, Intel Mobiles and Cores, but never with a P4 so far. So don't worry too much.
Hope I was able to help.
Annika
RE: Hi, Here is a two part
)
Well, it's probably better covered in the "Problems" section than in Crunchers' Corner, but .... nevertheless...
This is kind of strange, it seems that your CPU is detected by BOINC (not the science app) as NOT being SSE2 capable (only SSE seems to be supported). AFAIK, all Pentium 4s are SSE2 capable, even the "mobile" variants.
If your CPU really does not support the SSE2 instruction set, the Microsoft Runtime lib inside the Windows science app will switch to a slower code path which cannnot be helped at the moment, except maybe by switching to Linux :-).
However, this does not explain fully the far from spectacular performance of this host ;-). The benchmarks don't look that hot either, so I'd guess that the notebook might be running in a power saving mode with reduced clock freq., or maybe your motherboard has some problems in addition to detecting the additional RAM. If you have other applications running at the same time, swapping might be an issue.
EDIT: Annika, you beat me.... I must type a bit faster, I guess..
CU
BRM
Hmm.. there seems to be a
)
Hmm.. there seems to be a general issue concerning SSE2 and Windows 2000 ???
See this user on Seti, for example, he/she has a good mix of Win variants and CPUs:
http://setiathome.berkeley.edu/hosts_user.php?userid=7801669
All the P4s under Win 2k are lacking the sse2 flag !
Very strange. The OS has to support SSE2 (AFAIK SSE2 is disabled by default and is switched on by the OS when booting, and the OS has to be aware of the SSE2 registers so they are saved and restored on task-switches). But Win 2k should support SSE2! Strange!
CU
BRM