We've seen a ~17% decrease in credit awarded per work unit yet speedups have been 30%+ ??? Not that i mind the extra credit.
I don't mid extra credit at all too. :D
But why all this cross project credit calibrating stuff ???
Why can't we leave it the way it is ?
To avoid a credit war between projects, and the attendant flaming wars that would accompany it.
There's no credit war... if i really go for the credits i'd join a project that uses a quorum of 1 and use one of my previous released boinc clients (5.5.0, which are mirrored without my permission!!) which were only to be used with S@H to compensate the faster sience app.
Anyhow this whole credit leveling is not needed cause one will choose a project which fits the cpu beeing used ... except me .. got much more intels than amds working on E@H ... to bad my fault :P
P.S.
If a project "REALLY" intends to get participants with granting more credit...
Why worry !!! Were all in for the sience ! arn't we ?!?
Anyhow this whole credit leveling is not needed cause one will choose a project which fits the cpu beeing used ... except me .. got much more intels than amds working on E@H ... to bad my fault :P
P.S.
If a project "REALLY" intends to get participants with granting more credit...
Why worry !!! Were all in for the sience ! arn't we ?!?
Yes, all about the science Crunch3r -
However cross project credits are important, I know that Rosetta is revamping the way credits are granted as well - maybe someday it will be even between projects.
I hope I misunderstood what was posted before that credit per hour is flat between machines ?? I am assuming I am misunderstanding this at least. Faster machines should get higher credit per hour than slower ones.
Looks like I am going to give this new app a shot.
My credits too have dropped from 19.61 to 16.26 for a short WU (about 10 hours of CPU time) on my Pentium II running Linux, BOINC 5.4.9 and the 4.16 app. 4.17 was slower since I do not have any SSE, only MMX which is useless in scientific crunching. But I am still able to watch NASA WindowsMedia or Quicktime streams using Mplayer and Realplay streams using Realplayer.
Tullio
There's no credit war... if i really go for the credits i'd join a project that uses a quorum of 1 and use one of my previous released boinc clients (5.5.0, which are mirrored without my permission!!) which were only to be used with S@H to compensate the faster sience app.
Anyhow this whole credit leveling is not needed cause one will choose a project which fits the cpu beeing used ... except me .. got much more intels than amds working on E@H ... to bad my fault :P
P.S.
If a project "REALLY" intends to get participants with granting more credit...
Why worry !!! Were all in for the sience ! arn't we ?!?
G'day Cruncher
My comments were not directed at you personally, or any other participant. There is, however, a growing view that with the move to serverside/non benchmark related credits that the projects implementing these systems should do so responsibly.
There is no credit war yet, but it is easy to see how one could develop, therefore to avoid the situation from even arising in the first place there is a move to nip it in the bud.
Additionally, thank you for your comments regarding the responsible use of your 5.5.0 client.
Anyhow this whole credit leveling is not needed cause one will choose a project which fits the cpu beeing used ... except me .. got much more intels than amds working on E@H ... to bad my fault :P
P.S.
If a project "REALLY" intends to get participants with granting more credit...
Why worry !!! Were all in for the sience ! arn't we ?!?
Yes, all about the science Crunch3r -
However cross project credits are important, I know that Rosetta is revamping the way credits are granted as well - maybe someday it will be even between projects.
I hope I misunderstood what was posted before that credit per hour is flat between machines ?? I am assuming I am misunderstanding this at least. Faster machines should get higher credit per hour than slower ones.
Looks like I am going to give this new app a shot.
G'day Kevint
The idea is to keep credits/hour the same between projects on the same machine, not to keep credits/hour the same for every machine. A faster machine will get more credits/hour.
Well, Credits dropped a lot and speed remained the same on plain old Athlon (without SSE) - not very amusing...
I still remember the incredible speed increase from akosf for 3Dnow! - Improvement? - sad.
Well, Credits dropped a lot and speed remained the same on plain old Athlon (without SSE) - not very amusing...
I still remember the incredible speed increase from akosf for 3Dnow! - Improvement? - sad.
I have some non-SSE boxes too, although they run linux, so it will be interesting to see how the credit/hr pans out on these boxes.
Well, Credits dropped a lot and speed remained the same on plain old Athlon (without SSE) - not very amusing...
I still remember the incredible speed increase from akosf for 3Dnow! - Improvement? - sad.
Do you mean this host, where the credit dropped from 176.26 to 146.17 (17%), but the time dropped from 107,449.56 to 77,044.77 (28.3%)?
Sounds like an improvement to me.
[Edit - I run a Celeron MMX and a Pentium II MMX, and I saw time reductions of 32% - 33% during the beta. So maybe the Athlon hasn't gained quite as much, but you're still in postive territory]
Well, Credits dropped a lot and speed remained the same on plain old Athlon (without SSE) - not very amusing...
I still remember the incredible speed increase from akosf for 3Dnow! - Improvement? - sad.
This incredible improvement is already incorporated in the steps from the old S4 App to the first generation of S5R1 Apps (speedup of 2,5) and in the latest switch to the 4.17/4.24 Apps (10-30%). Looking at the current code Akos and I agreed that the theoretical speedup from using 3DNow! would be completely eaten up by the time it takes to switch between FPU and 3DNow! mode, so we probably won't make a special 3DNow! version.
RE: RE: RE: We've seen
)
There's no credit war... if i really go for the credits i'd join a project that uses a quorum of 1 and use one of my previous released boinc clients (5.5.0, which are mirrored without my permission!!) which were only to be used with S@H to compensate the faster sience app.
Anyhow this whole credit leveling is not needed cause one will choose a project which fits the cpu beeing used ... except me .. got much more intels than amds working on E@H ... to bad my fault :P
P.S.
If a project "REALLY" intends to get participants with granting more credit...
Why worry !!! Were all in for the sience ! arn't we ?!?
RE: Anyhow this whole
)
Yes, all about the science Crunch3r -
However cross project credits are important, I know that Rosetta is revamping the way credits are granted as well - maybe someday it will be even between projects.
I hope I misunderstood what was posted before that credit per hour is flat between machines ?? I am assuming I am misunderstanding this at least. Faster machines should get higher credit per hour than slower ones.
Looks like I am going to give this new app a shot.
Join SETI.USA
My credits too have dropped
)
My credits too have dropped from 19.61 to 16.26 for a short WU (about 10 hours of CPU time) on my Pentium II running Linux, BOINC 5.4.9 and the 4.16 app. 4.17 was slower since I do not have any SSE, only MMX which is useless in scientific crunching. But I am still able to watch NASA WindowsMedia or Quicktime streams using Mplayer and Realplay streams using Realplayer.
Tullio
RE: There's no credit
)
G'day Cruncher
My comments were not directed at you personally, or any other participant. There is, however, a growing view that with the move to serverside/non benchmark related credits that the projects implementing these systems should do so responsibly.
There is no credit war yet, but it is easy to see how one could develop, therefore to avoid the situation from even arising in the first place there is a move to nip it in the bud.
Additionally, thank you for your comments regarding the responsible use of your 5.5.0 client.
RE: RE: Anyhow this
)
G'day Kevint
The idea is to keep credits/hour the same between projects on the same machine, not to keep credits/hour the same for every machine. A faster machine will get more credits/hour.
Well, Credits dropped a lot
)
Well, Credits dropped a lot and speed remained the same on plain old Athlon (without SSE) - not very amusing...
I still remember the incredible speed increase from akosf for 3Dnow! - Improvement? - sad.
Aloha, Uli
RE: Well, Credits dropped a
)
I have some non-SSE boxes too, although they run linux, so it will be interesting to see how the credit/hr pans out on these boxes.
RE: Well, Credits dropped a
)
Do you mean this host, where the credit dropped from 176.26 to 146.17 (17%), but the time dropped from 107,449.56 to 77,044.77 (28.3%)?
Sounds like an improvement to me.
[Edit - I run a Celeron MMX and a Pentium II MMX, and I saw time reductions of 32% - 33% during the beta. So maybe the Athlon hasn't gained quite as much, but you're still in postive territory]
RE: Do you mean this host,
)
Yes, you are right. It is indeed an improvement - maybe i simply expected too much. ;)
Aloha, Uli
RE: Well, Credits dropped a
)
This incredible improvement is already incorporated in the steps from the old S4 App to the first generation of S5R1 Apps (speedup of 2,5) and in the latest switch to the 4.17/4.24 Apps (10-30%). Looking at the current code Akos and I agreed that the theoretical speedup from using 3DNow! would be completely eaten up by the time it takes to switch between FPU and 3DNow! mode, so we probably won't make a special 3DNow! version.
BM
BM