I hope you will consider re-joining Einstein@Home once the AMD/Windows performance issue is resolved?
CU
BRM
There's nothing to reconsider. As soon as the problems with the AMD's, etc are cleared up I WILL be back. I have no intention of abandoning my favorite project.
During the last month (invalid results) / (valid results within past 7 days / 7) was increasing exponentialy (linear in log scale) and is at the monemt 26%. That is an increase of more than 2 decades within the last 6 weeks! There must be some applications (LINUX ?), that never get a validated result. What a luck, I´m using Win2K and had up to now no not validated file at all during the last 2 years.
The Linux app validates fine most of the time, just like the Windows app. In a few cases, results returned from the two initially tasked hosts do not agree and a third host is asked to compute the result for that workunit. In these cases, most of the time the OS in the majority wins, which is most likely Windows. But it's far from true that the Linux app. never validates (see my hosts, for example). Validation problems do happen (as in most other BOINC projects) but are kind of rare.
...
But looking at the posts it does seems that there is an issue with linux applications on AMD Athlon processors. It's hard to see from the few that actually posts, but say looking at: this thread I can see that he has the same problems I have with one of my computers. Some results end with an error while other just don't get validated. I'm running MEMTEST86 now to make sure that the computer is actually fine, but it did use to run flawlessly before.
During the last month (invalid results) / (valid results within past 7 days / 7) was increasing exponentialy (linear in log scale) and is at the monemt 26%. That is an increase of more than 2 decades within the last 6 weeks! There must be some applications (LINUX ?), that never get a validated result. What a luck, I´m using Win2K and had up to now no not validated file at all during the last 2 years.
The Linux app validates fine most of the time, just like the Windows app. In a few cases, results returned from the two initially tasked hosts do not agree and a third host is asked to compute the result for that workunit. In these cases, most of the time the OS in the majority wins, which is most likely Windows. But it's far from true that the Linux app. never validates (see my hosts, for example). Validation problems do happen (as in most other BOINC projects) but are kind of rare.
...
But looking at the posts it does seems that there is an issue with linux applications on AMD Athlon processors. It's hard to see from the few that actually posts, but say looking at: this thread I can see that he has the same problems I have with one of my computers. Some results end with an error while other just don't get validated. I'm running MEMTEST86 now to make sure that the computer is actually fine, but it did use to run flawlessly before.
Hi !
The two client errors I could see ion your results look similar, for both the debugging output contains lines like:
2007-05-27 15:36:55.7653 [debug]: Read checkpoint - reading previous output...
2007-05-27 15:36:55.9789 [CRITICAL]: Line 17485 is empty.
2007-05-27 15:36:55.9790 [debug]: DEBUG: read_fstat_toplist_from_fp() returned -1
2007-05-27 15:36:55.9790 [CRITICAL]: ERROR: format error in toplist
2007-05-27 15:36:55.9790 [debug]: ERROR reading previous output
To me it looks like there's a problem when shuting down BOINC. Einstein can't continue the work at the point where it last stopped because the file used to store the computing results is somehow corrupted. Seems to be more a disk IO issue than a RAM or CPU issue.
During the last month (invalid results) / (valid results within past 7 days / 7) was increasing exponentialy (linear in log scale) and is at the monemt 26%. That is an increase of more than 2 decades within the last 6 weeks! There must be some applications (LINUX ?), that never get a validated result. What a luck, I´m using Win2K and had up to now no not validated file at all during the last 2 years.
The Linux app validates fine most of the time, just like the Windows app. In a few cases, results returned from the two initially tasked hosts do not agree and a third host is asked to compute the result for that workunit. In these cases, most of the time the OS in the majority wins, which is most likely Windows. But it's far from true that the Linux app. never validates (see my hosts, for example). Validation problems do happen (as in most other BOINC projects) but are kind of rare.
Yup, I guess more frequent official status reports would be nice. All I know is that the known problems are worked on. I had some correspondence with Bernd about the AMD performance problem under Windows and judging from the time some of the e-mails were sent, one would even think that the staff is working 24/7 :-).
CU
BRM
Well, this is the first that anyone without direct correspondence with the staff of the project has heard on the subject. Most of us rely on getting updates from the website, while a few get the updates from the forums. It is nice to know that someone is feverishly working on solving the issues. Since you are in communication with Bernd, have you been privy to a timeframe for solving our validation and performance issues?
I hope you will consider
)
I hope you will consider re-joining Einstein@Home once the AMD/Windows performance issue is resolved?
CU
BRM
RE: I hope you will
)
There's nothing to reconsider. As soon as the problems with the AMD's, etc are cleared up I WILL be back. I have no intention of abandoning my favorite project.
RE: RE: During the last
)
But looking at the posts it does seems that there is an issue with linux applications on AMD Athlon processors. It's hard to see from the few that actually posts, but say looking at: this thread I can see that he has the same problems I have with one of my computers. Some results end with an error while other just don't get validated. I'm running MEMTEST86 now to make sure that the computer is actually fine, but it did use to run flawlessly before.
RE: RE: RE: During the
)
Hi !
The two client errors I could see ion your results look similar, for both the debugging output contains lines like:
2007-05-27 15:36:55.7653 [debug]: Read checkpoint - reading previous output...
2007-05-27 15:36:55.9789 [CRITICAL]: Line 17485 is empty.
2007-05-27 15:36:55.9790 [debug]: DEBUG: read_fstat_toplist_from_fp() returned -1
2007-05-27 15:36:55.9790 [CRITICAL]: ERROR: format error in toplist
2007-05-27 15:36:55.9790 [debug]: ERROR reading previous output
To me it looks like there's a problem when shuting down BOINC. Einstein can't continue the work at the point where it last stopped because the file used to store the computing results is somehow corrupted. Seems to be more a disk IO issue than a RAM or CPU issue.
CU
BRM
RE: RE: During the last
)
Well, this is the first that anyone without direct correspondence with the staff of the project has heard on the subject. Most of us rely on getting updates from the website, while a few get the updates from the forums. It is nice to know that someone is feverishly working on solving the issues. Since you are in communication with Bernd, have you been privy to a timeframe for solving our validation and performance issues?
RE: Since you are in
)
We only exchanged a few emails, it's not that I'm in constant contact with Bernd :-). No, I have no idea about a timeframe, sorry.
CU
H-B
RE: RE: Since you are in
)
The Einstein staff works on the client errors.
But they could not localize the problems exactly.