Goodness, looks like I was late. I blame it on a rather hard first day back at university (I am so going to suck at the graphics stuff they're making us do this semester...) but well, it didn't really hurt to switch apps the complicated way. Hopefully this is going to help performance.
Given the fact that I was using 4.33 for the R2 result, I would guess that the 4.38/4.40 apps would've been slower than what 4.07 is on this system, so to me, the performance at this frequency seems to be even with S5R2 being processed with 4.33...
I don't know if somebody announced it already, but seems to me that 4.07 is now official app ... a couple of my crunchers got it and I for sure haven't installed it there manually ...
If you are currently beta-testing, it's best to keep the app_info.xml file because there will be a new beta app soon for Windows that will contain some features of the current Linux beta app.
If you are currently beta-testing, it's best to keep the app_info.xml file because there will be a new beta app soon for Windows that will contain some features of the current Linux beta app.
CU
Bikeman
Will the new Windows beta be up sometime this week do you think? Otherwise perhaps the 4.07 app could be removed from the Beta testing page since it is now no longer beta....
If you are currently beta-testing, it's best to keep the app_info.xml file because there will be a new beta app soon for Windows that will contain some features of the current Linux beta app.
Good thing Einstein thinks it's done enough work for now, so I can reasonably easy set it to NNT while waiting for the new beta to come along. Just so I don't have to restart my task all over again, because the switch to the new beta balked the running of the old app. ;-)
Or is the 4.07 performing more error checking hence the longer runtime?
As soon as I get to this client I would pull through some other 4.07-WUs to get a better base for comparisons.
Or is the 4.07 performing more error checking hence the longer runtime?
As soon as I get to this client I would pull through some other 4.07-WUs to get a better base for comparisons.
The natural variation in CPU time for results in the same sequence is much higher for S5R3 than we are used to.
RE: I am under the
)
My AMD's do better, but be sure to write the text statement EAH_NO_GRAPHICS into the BOINC directory, that made a noticeable difference.
And btw, the 4.07 is now the main Windows application as of today! ;)
Goodness, looks like I was
)
Goodness, looks like I was late. I blame it on a rather hard first day back at university (I am so going to suck at the graphics stuff they're making us do this semester...) but well, it didn't really hurt to switch apps the complicated way. Hopefully this is going to help performance.
RE: RE: I am under the
)
It looked like S5R2 work had mostly dried up, so I loaded up the 4.07 beta yesterday. I've submitted one result that was run completely with 4.07.
513.50 S5R2 result completed in 79579.125 seconds for credit of 449.04.
Caveat: This was with the 4.33 edited-for-AMD ("patched") app.
513.50 S5R3 result completed in 41054.671875 seconds for credit of 221.97.
79579.125 / 41054.671875 = 1.938
449.04 / 221.97 = 2.023
Given the fact that I was using 4.33 for the R2 result, I would guess that the 4.38/4.40 apps would've been slower than what 4.07 is on this system, so to me, the performance at this frequency seems to be even with S5R2 being processed with 4.33...
FWIW, YMMV, etc, etc, etc...
I don't know if somebody
)
I don't know if somebody announced it already, but seems to me that 4.07 is now official app ... a couple of my crunchers got it and I for sure haven't installed it there manually ...
Metod ...
RE: I don't know if
)
Correct! A new beta app may be up at some point in the not too distant(?) future, but for now the beta and the official application are identical.
RE: RE: I don't know if
)
See my message here.
If you are currently beta-testing, it's best to keep the app_info.xml file because there will be a new beta app soon for Windows that will contain some features of the current Linux beta app.
CU
Bikeman
RE: If you are currently
)
Will the new Windows beta be up sometime this week do you think? Otherwise perhaps the 4.07 app could be removed from the Beta testing page since it is now no longer beta....
RE: If you are currently
)
Good thing Einstein thinks it's done enough work for now, so I can reasonably easy set it to NNT while waiting for the new beta to come along. Just so I don't have to restart my task all over again, because the switch to the new beta balked the running of the old app. ;-)
Is it only with me, a
)
Is it only with me, a coincidence or is 4.07 slower than 4.01?
4.07: ~ 47250 s (2 WUs; 221.57 C; 47131 - 47347 s)
4.01: ~ 44920 s (11 WUs; 221.57 C; 44189 - 45630 s)
Or is the 4.07 performing more error checking hence the longer runtime?
As soon as I get to this client I would pull through some other 4.07-WUs to get a better base for comparisons.
RE: Is it only with me, a
)
The natural variation in CPU time for results in the same sequence is much higher for S5R3 than we are used to.
see:
message 75702
and
message 75697
for concrete examples.