Maybe I should of stuck to the U.41.02/03 version, my second result of the U.41.04 version has gone waacky, it was at a hundred percent completion, and BOINC stopped to resume an xtremlabs WU, upon resuming the Einstein WU still showing CPU time of 1 hr 7 minutes went to 0% completion, and time left went to 3 more hours...I double checked, its the same work unit, since boinc hasn't sent off any completed workunits within the past hour.
The application needs some seconds to initialize some dataareas when it restarts a stopped wu, but not more...
Maybe I should of stuck to the U.41.02/03 version, my second result of the U.41.04 version has gone waacky, it was at a hundred percent completion, and BOINC stopped to resume an xtremlabs WU, upon resuming the Einstein WU still showing CPU time of 1 hr 7 minutes went to 0% completion, and time left went to 3 more hours...I double checked, its the same work unit, since boinc hasn't sent off any completed workunits within the past hour.
The application needs some seconds to initialize some dataareas when it restarts a stopped wu, but not more...
Well, its been 4 minutes since I've resumed it and it seems to have started over again, its showing 7% completion now...despite the fact that it was completed...
My HT Prescott has crunched 10 WUs with U41.04, 8 of them short and 2 of them long. All have validated and received credit. However, the performance on this chip lags behind that obtained with either S39L or S40.04. Here are the numbers:
P-4 Model 540 Prescott, 3.20 GHz, HT enabled
Akosf-optimised client version S39L
Short WUs, sample of 18 processed, avg time 2086 sec (34m 46s)
Long WUs, sample of 15 processed, avg time 6949 sec (1h 55m 49s)
Akosf-optimised client version S40.04
Short WUs, sample of 29 processed, avg time 2103 sec (35m 3s)
Long WUs, sample of 16 processed, avg time 6899 sec (1h 54m 59s)
Akosf-optimised client version U41.04
Short WUs, sample of 8 processed, avg time 2501 sec (41m 41s)
Long WUs, sample of 2 processed, avg time 8091 sec (2h 14m 51s)
U41.04 faulted 61 results and put my AMD 64 3800+ (id 427427) out of business for 12 hours.
5.2.13 BoincStudio 0.4b
- exit code -1073741795 (0xc000001d)
2006-05-05 07:19:25.1519 [normal]: U41.04 (akosf) -> 'projects/einstein.phys.uwm.edu/albert_4.37_windows_intelx86.exe'.
2006-05-05 07:19:25.1519 [normal]: Started search at lalDebugLevel = 0
2006-05-05 07:19:26.0426 [normal]: Checkpoint-file 'Fstat.out.ckp' not found.
2006-05-05 07:19:26.0426 [normal]: No usable checkpoint found, starting from beginning.
***UNHANDLED EXCEPTION****
Reason: Illegal Instruction (0xc000001d) at address 0x0040A8BA
It seems like your CPU doesn't have SSE3. It's either a winchester core (D-stepping) or a 130nm C-stepping, both of which have *only* SSE2. Only E-steppings have SSE3. They're being sold since ~1 year now.
My HT Prescott has crunched 10 WUs with U41.04, 8 of them short and 2 of them long. All have validated and received credit. However, the performance on this chip lags behind that obtained with either S39L or S40.04.
...
It is already known, that HT causes processors being not fully exploited with the newest Akos' optimizations. If you want the performance to be improved, turn OFF the hyperthreading.
I have p4 Prescott 3.4 GHz. I used S39L since it was faster than S41.07. However, after turning off HT, S41.07 is approximately 10% faster than S39L with HT!
S39L with HT on : 6100 seconds for 2 WUs (thus 3050 s for one WU).
S41.07 without HT: 2700 seconds for one WU. (all results for long WUs)
I haven't tried S40.06 nor U41.07, because I had had not enough time...
D920@4Ghz U41.04 vs.
)
D920@4Ghz
U41.04 vs. S40.04
38 minutes vs. 52 minutes
Celeron Prescott
47 minutes vs. 1h 07 minutes
or 27%-29% faster
RE: Maybe I should of stuck
)
The application needs some seconds to initialize some dataareas when it restarts a stopped wu, but not more...
RE: RE: Maybe I should of
)
Well, its been 4 minutes since I've resumed it and it seems to have started over again, its showing 7% completion now...despite the fact that it was completed...
Human Stupidity Is Infinite...
My HT Prescott has crunched
)
My HT Prescott has crunched 10 WUs with U41.04, 8 of them short and 2 of them long. All have validated and received credit. However, the performance on this chip lags behind that obtained with either S39L or S40.04. Here are the numbers:
P-4 Model 540 Prescott, 3.20 GHz, HT enabled
Akosf-optimised client version S39L
Short WUs, sample of 18 processed, avg time 2086 sec (34m 46s)
Long WUs, sample of 15 processed, avg time 6949 sec (1h 55m 49s)
Akosf-optimised client version S40.04
Short WUs, sample of 29 processed, avg time 2103 sec (35m 3s)
Long WUs, sample of 16 processed, avg time 6899 sec (1h 54m 59s)
Akosf-optimised client version U41.04
Short WUs, sample of 8 processed, avg time 2501 sec (41m 41s)
Long WUs, sample of 2 processed, avg time 8091 sec (2h 14m 51s)
That box is now back on S40.04. ;-)
U41.04 faulted 61 results and
)
U41.04 faulted 61 results and put my AMD 64 3800+ (id 427427) out of business for 12 hours.
5.2.13 BoincStudio 0.4b
- exit code -1073741795 (0xc000001d)
2006-05-05 07:19:25.1519 [normal]: U41.04 (akosf) -> 'projects/einstein.phys.uwm.edu/albert_4.37_windows_intelx86.exe'.
2006-05-05 07:19:25.1519 [normal]: Started search at lalDebugLevel = 0
2006-05-05 07:19:26.0426 [normal]: Checkpoint-file 'Fstat.out.ckp' not found.
2006-05-05 07:19:26.0426 [normal]: No usable checkpoint found, starting from beginning.
***UNHANDLED EXCEPTION****
Reason: Illegal Instruction (0xc000001d) at address 0x0040A8BA
1: 05/05/06 07:19:26
1: e:\\einsteinathome\\cfs\\windows_build\\albert4.37\\cfslaldemod.c(850) +0 bytes (TestLALDemod)
1: SymGetLineFromAddr(): GetLastError = 126
It seems like your CPU
)
It seems like your CPU doesn't have SSE3. It's either a winchester core (D-stepping) or a 130nm C-stepping, both of which have *only* SSE2. Only E-steppings have SSE3. They're being sold since ~1 year now.
MrS
Scanning for our furry friends since Jan 2002
I guess, I read a post that
)
I guess, I read a post that it should work. Guess I shouldn't have believed it.
RE: I guess, I read a post
)
When in doubt, get CPU-Z. It tells you which instruction sets your chip supports.
AMD Athlon 64 X2 3800+
)
AMD Athlon 64 X2 3800+ (stock)
S39L -> U41.04 ~42% improvement
:-) well done again Akos
RE: My HT Prescott has
)
It is already known, that HT causes processors being not fully exploited with the newest Akos' optimizations. If you want the performance to be improved, turn OFF the hyperthreading.
I have p4 Prescott 3.4 GHz. I used S39L since it was faster than S41.07. However, after turning off HT, S41.07 is approximately 10% faster than S39L with HT!
S39L with HT on : 6100 seconds for 2 WUs (thus 3050 s for one WU).
S41.07 without HT: 2700 seconds for one WU. (all results for long WUs)
I haven't tried S40.06 nor U41.07, because I had had not enough time...