First complete WU completed with the S39 on my P4 2.8 ghz (512kb L2) at 1:14:04. This is down from an average of 5 hours on this size unit on the basic albert 4.37.
Kudos to AKOSF for his efforts and giving us such a huge step toward closing the data collection vs. crunching gap.
Dang..I could just SWEAR I've read past posts assuring us this client was ALREADY optimized.
As a worldwide average price per kWh at retail USD 0.20 is probably pretty close. Never mind CA. You have Australia, France, Germany, Hungary, ... , and yes the US in that mix. But who cares, if $0.20 is a little high. Just over 300W per host is almost certainly low, when you take all inefficiencies and components into account. 200 days for S4 before optimization is low as you can see from the stats. The way I calculated active hosts probably tends toward low.
In essence: the energy cost of doing the S4 analysis goes from roughly $10M to roughly $2.5M after optimization.
That's just the economic side to it, there are pretty obvious ecological benefits too.
Code review and optimization should become a mandatory step for all BOINC apps that want to be considered "production apps". The wastefulness of not running optimized code on hundreds of thousands of machines is plain ugly.
The costs per WU is a function of the duration time per WU.
On my PC Pentium4 2,6GHz means that: 10h to 1,5h with the new S39.
The costs are equivalent 10 to 1,5 independent from currancy or price of KWh.
We can now produce 6,6 WU in the same time as formerly 1 WU. That means a enourmous increasing productivity. (seti_britta)
As far as 20c/kwh goes I'll file that as reason not to move to CA #392. I'm only paying 8. :)
0.20$/kWh as "average" isn't that unrealistic I guess, since most of the participants are North Americans or Europeans ... never mind. We keep on crunching what ever it takes :)
As far as 20c/kwh goes I'll file that as reason not to move to CA #392. I'm only paying 8. :)
0.20$/kWh as "average" isn't that unrealistic I guess, since most of the participants are North Americans or Europeans ... never mind. We keep on crunching what ever it takes :)
Whatever.. I guess it's one way to look at it. Another way to look at it is my computer, with only neccessary and short breaks, has been running 24/7 since may of 99 when I joined the original SAH.
More profitable and realistic I think to recognize the science is getting done in a more timely manner and that's the important thing.
As to the electricity..our town electric company gave us a 1/3 rate hike last fall. Yep. I'm now paying a full 4 US cents/KWh. ouch. ;=)
First wu with S39 on a Barton 2500+ :
"before", between 5,335.00 and 5,024.05 sec
now : for the first time, under the 5,000 sec : 4,720.00 sec.
This ist quite funny. On my Barton 2500+ I don't see any increase of speed comparing alike WUs nor is S39 slower either.
Just stagnany in completion time.
AMD XP 2600+ S38->S39 ~6.7% reduction in CPU time / wu
Result duration correction factor 0.281227
AMD X2 3800+ S38->S39 ~9.6% reduction in CPU time / wu
Result duration correction factor 0.258055 (hair's breath from 75% improvement on dist and you can bet there were no slouchers coding the dist)
AMD XP 2600+ S38->S39 ~6.7% reduction in CPU time / wu
AMD X2 3800+ S38->S39 ~9.6% reduction in CPU time / wu
Result duration correction factor 0.258055 (hair's breath from 75% improvement on dist and you can bet there were no slouchers coding the dist)
Interesting... My AMD XP 2600+ (Thoroughbred) went from ~6400 to ~4650 secs. About a 20-25% improvement.
First complete WU completed
)
First complete WU completed with the S39 on my P4 2.8 ghz (512kb L2) at 1:14:04. This is down from an average of 5 hours on this size unit on the basic albert 4.37.
Kudos to AKOSF for his efforts and giving us such a huge step toward closing the data collection vs. crunching gap.
Dang..I could just SWEAR I've read past posts assuring us this client was ALREADY optimized.
I found little or no increase
)
I found little or no increase in speed using an Athlon XP2800+
Mind you I have been using the PC whilst crunching. I'll leave it going overnight and see what speeds I get
RE: As a worldwide average
)
The costs per WU is a function of the duration time per WU.
On my PC Pentium4 2,6GHz means that: 10h to 1,5h with the new S39.
The costs are equivalent 10 to 1,5 independent from currancy or price of KWh.
We can now produce 6,6 WU in the same time as formerly 1 WU. That means a enourmous increasing productivity. (seti_britta)
RE: As far as 20c/kwh goes
)
0.20$/kWh as "average" isn't that unrealistic I guess, since most of the participants are North Americans or Europeans ... never mind. We keep on crunching what ever it takes :)
Greetings, Santas little helper
RE: RE: As far as 20c/kwh
)
Whatever.. I guess it's one way to look at it. Another way to look at it is my computer, with only neccessary and short breaks, has been running 24/7 since may of 99 when I joined the original SAH.
More profitable and realistic I think to recognize the science is getting done in a more timely manner and that's the important thing.
As to the electricity..our town electric company gave us a 1/3 rate hike last fall. Yep. I'm now paying a full 4 US cents/KWh. ouch. ;=)
RE: I found little or no
)
On my XP2000+, I went from the 59xx,xx range to the 54xx,xx range, so defo an improvement, but not big. Perhaps other stuff are hogging your pc ??
RE: First wu with S39 on a
)
This ist quite funny. On my Barton 2500+ I don't see any increase of speed comparing alike WUs nor is S39 slower either.
Just stagnany in completion time.
S39 dropped another 1600
)
S39 dropped another 1600 seconds/wu over S38 on my 3.4 Prescott (HT enabled).
me-[at]-rescam.org
AMD XP 2600+ S38->S39 ~6.7%
)
AMD XP 2600+ S38->S39 ~6.7% reduction in CPU time / wu
Result duration correction factor 0.281227
AMD X2 3800+ S38->S39 ~9.6% reduction in CPU time / wu
Result duration correction factor 0.258055 (hair's breath from 75% improvement on dist and you can bet there were no slouchers coding the dist)
RE: AMD XP 2600+ S38->S39
)
Interesting... My AMD XP 2600+ (Thoroughbred) went from ~6400 to ~4650 secs. About a 20-25% improvement.
Seti Classic Final Total: 11446 WU.