AMD Athlon 64 running 36 h and 61-62 h to completion?

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494410
RAC: 0

Well, I agree about the "see

Well, I agree about the "see it as a challenge" thing, but then maybe I can't talk ;-) how many computation errors do you need to qualify? I'm not meaning this in an offensive way, and there is a bit of seriousness hidden underneath because I don't know if other people are really so much worse off or if it's based more on opinion/perception.
As for the theory that Einstein maybe doesn't have the resources to do a full-size beta test I can't say anything about that but it's certainly within the realm of possibility, so maybe one should give them the benefit of the doubt.
I'll stick with the project in any case, although I got a few computation errors and have pendings more than 10 times my current RAC, because the science is inspiring enough to motivate me and so is the great community...

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282700
RAC: 0

RE: so maybe one should

Message 63811 in response to message 63810

Quote:
so maybe one should give them the benefit of the doubt.

The problem that many seem to have when viewing posts of dissention, even if it is only mild dissention, is that the underlying motivation is missed in a furvor to defend the project and to put down or marginalize the dissenting view...

Nothing can be done about this situation now beyond putting the new bugfix versions into the beta area to make sure that they don't make things worse... The application is released and as the saying goes, "it is what it is." My sincere hope is that a "lessons learned" would be done and that one of those lessons learned will be that a better testing mechanism needs to be developed.

How many people will Kirsten talk to about EAH now? How will what she says impact the decision of people to join the project? Her primary complaint is that the app is slow. That could've been at least minimized, if not avoided, by more testing and getting code optimization in place. Odds are the Linux issues could've been avoided as well. The validator issues would've been brought to light. Fewer people would've been exposed to problems...

If a testing area is set up, you can still see your work done there "as a challenge", without having to take everyone along with you on that ride. I'm not belittling your idea of seeing it as a challenge, I'm just suggesting that you may wish to consider that not everyone likes things that way, just like not everyone will immediately sign up to bungee jump...

FWIW and YMMV...

Kirsten
Kirsten
Joined: 18 Jan 05
Posts: 16
Credit: 6780
RAC: 0

RE: RE: so maybe one

Message 63812 in response to message 63811

Quote:
Quote:
so maybe one should give them the benefit of the doubt.

How many people will Kirsten talk to about EAH now? How will what she says impact the decision of people to join the project? Her primary complaint is that the app is slow. That could've been at least minimized, if not avoided, by more testing and getting code optimization in place. Odds are the Linux issues could've been avoided as well. The validator issues would've been brought to light. Fewer people would've been exposed to problems...

If a testing area is set up, you can still see your work done there "as a challenge", without having to take everyone along with you on that ride. I'm not belittling your idea of seeing it as a challenge, I'm just suggesting that you may wish to consider that not everyone likes things that way, just like not everyone will immediately sign up to bungee jump...

FWIW and YMMV...

I have been asking the same questions about EAH in my Danish forum as I have here. I have of course not tried to talk any member into leaving the project, and no one seems to have responded in that way - on the contrary.

In our forum we have more or less concluded that S5R2 is not AMD Athlon friendly, though. One member says that according to Akos (who made the EAH optimizations in earlier days) the S5R2 compiler's output is rather messy and that Athlons are especially vulnerable to this. According to the same member several machines have crashed their WU's. These problems point IMO to creation of a testing project.

An Intel Core2 Quad CPU member is very surprised even of the Athlon 64 member's 30 hours crunching time. He is normally crunching a S5R2 in about 8 hours. (My own AMD Athlon 64 would - as earlier mentioned - supposedly run a S5R2 in 100 hours before completion.)

All in all our discussion has only concluded that our AMD Athlon members might wait for more Athlon friendly WU's, as they can use their machines more effectively in other projects.

I have often taken part in testing. As a matter of fact I tested the whole BOINC concept in very early days for SETI@home. I just want to *know* beforehand if a project is testing new applications, as I must expect to use time for reporting bugs and problems.

I do not understand any of the discussion here about Einstein not having resources for a separate testing area. Even a rather small project as ABC@home has had the resources to create a beta testing area.

To nail it down: I am in BOINC for the science. I prefer SETI, but Einstein is normally my second choice - only not at the moment.

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282700
RAC: 0

RE: I have been asking the

Message 63813 in response to message 63812

Quote:

I have been asking the same questions about EAH in my Danish forum as I have here. I have of course not tried to talk other members into leaving the project, and no one seems to have responded in that way - on the contrary more members are active.

That's good to hear... This is a worthwhile project, even if I am a bit upset at the team at this point... :)

Quote:

In our forum we have more or less concluded that S5R2 is not AMD Athlon friendly.

More generally, and less abrasively, it is not modern cpu efficient. AMD does take more of a hit, which is strange to me considering the K8 FPU is better than what was available during Netburst times... :shrug:

Quote:
(My own AMD Athlon 64 would - as earlier mentioned - supposedly run a S5R2 in 100 hours before completion.)

I see one of those is a 530 credit unit. Extrapolating out based on my own experience and a bit of guessing, my overclocked 3700+ would've taken about 130000 seconds (+/- 3000 seconds), or roughly 35-37 hours. If you're not overclocked, and given that your 3200+ only has 512k of cache, I would guess it would've really taken you only about 60 hours, 70 at the absolute most... Even if you were to double my guessed-at times for my processor, you'd still only be at 74 hours. There is a Duration Correction Factor that assists in the projected runtimes. That wouldn't have been accurate at first...so it could've (and most likely did) overestimate...

Edit: bumped estimated times up by 10 hours

Edit #2: I noticed that the Core2 system that you mentioned only ran a 178 credit unit. You'd need to theoretically triple the time for a 530 credit unit. Also, your Athlon 64 system may be hindered by having less physical memory. Not sure...

Kirsten
Kirsten
Joined: 18 Jan 05
Posts: 16
Credit: 6780
RAC: 0

RE: RE: (My own AMD

Message 63814 in response to message 63813

Quote:
Quote:
(My own AMD Athlon 64 would - as earlier mentioned - supposedly run a S5R2 in 100 hours before completion.)

I see one of those is a 530 credit unit. Extrapolating out based on my own experience and a bit of guessing, my overclocked 3700+ would've taken about 130000 seconds (+/- 3000 seconds), or roughly 35-37 hours. If you're not overclocked, and given that your 3200+ only has 512k of cache, I would guess it would've really taken you only about 60 hours, 70 at the absolute most... Even if you were to double my guessed-at times for my processor, you'd still only be at 74 hours. There is a Duration Correction Factor that assists in the projected runtimes. That wouldn't have been accurate at first...so it could've (and most likely did) overestimate...

Edit: bumped estimated times up by 10 hours

I wondered, of course, whether BOINC made a correct estimation of time to completion or not, as the estimation at first was 67 hours. After 36 hours of crunching the estimation was only reduced to 62 hours, though. Therefore my question as I smelled a bug.

Off topic: It would have been nice to know that the estimated runtimes could be inaccurate *before* I detached the project and thereby wasted 36 hours of crunching. IMO I might have received a faster (and probably the quoted correct) answer to my original question if I have had the right to ask questions in the correct forum. To me a bug and problems forum is meant for newbies and/or beta testers.

Kirsten
Kirsten
Joined: 18 Jan 05
Posts: 16
Credit: 6780
RAC: 0

RE: RE: (My own AMD

Message 63815 in response to message 63813

Quote:

Quote:
(My own AMD Athlon 64 would - as earlier mentioned - supposedly run a S5R2 in 100 hours before completion.)

I see one of those is a 530 credit unit. Extrapolating out based on my own experience and a bit of guessing, my overclocked 3700+ would've taken about 130000 seconds (+/- 3000 seconds), or roughly 35-37 hours. If you're not overclocked, and given that your 3200+ only has 512k of cache, I would guess it would've really taken you only about 60 hours, 70 at the absolute most... Even if you were to double my guessed-at times for my processor, you'd still only be at 74 hours. There is a Duration Correction Factor that assists in the projected runtimes. That wouldn't have been accurate at first...so it could've (and most likely did) overestimate...

Edit: bumped estimated times up by 10 hours

Edit #2: I noticed that the Core2 system that you mentioned only ran a 178 credit unit. You'd need to theoretically triple the time for a 530 credit unit. Also, your Athlon 64 system may be hindered by having less physical memory. Not sure...

FYI I do not overclock and my AMD 64 Athlon 3200+ only has 512 Kb L2 cache. This particular machine only has 768 Mb RAM, but I fail to see any special EAH system requirements beyond the standard BOINC ones.

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282700
RAC: 0

RE: IMO I might have

Message 63816 in response to message 63814

Quote:
IMO I might have received a faster (and probably the quoted correct) answer to my original question if I have had the right to ask questions in the correct forum. To me a bug and problems forum is meant for newbies and/or beta testers.

I don't know for sure if the estimation was too much or not. My guess is that it was. The big wildcard is the amount of your physical memory and what impact it has on performance. You have 768MB, I believe... I have 2GB... I don't know how memory dependent the app is right now...

Anyway, the RAC policy is intended to keep malcontents at bay as much as possible. You know, the people who open up a new account so that they can come on and post inappropriate things... I suppose one could argue that if you have credits at all it should override the RAC policy. On the whole, I think the RAC policy is a good one. The difficulty is that, as you indicated, you're impatient ;-) You only waited 31 minutes before deciding to give up... A good friend of mine that lives in Latvia is highly impatient and impulsive too... I love her anyway... lol...

Also, if you detached, I'm curious why your results don't indicate that... That means that the other person who has submitted a result will have to wait the full deadline time before another unit is reissued... Not blaming you, just I would've thought that the BOINC CC should've sent something to the project to say "hey, this user detached, go ahead and close out any open work they have"...

Kirsten
Kirsten
Joined: 18 Jan 05
Posts: 16
Credit: 6780
RAC: 0

RE: RE: IMO I might have

Message 63817 in response to message 63816

Quote:
Quote:
IMO I might have received a faster (and probably the quoted correct) answer to my original question if I have had the right to ask questions in the correct forum. To me a bug and problems forum is meant for newbies and/or beta testers.

Anyway, the RAC policy is intended to keep malcontents at bay as much as possible. You know, the people who open up a new account so that they can come on and post inappropriate things... I suppose one could argue that if you have credits at all it should override the RAC policy. On the whole, I think the RAC policy is a good one. The difficulty is that, as you indicated, you're impatient ;-) You only waited 31 minutes before deciding to give up... A good friend of mine that lives in Latvia is highly impatient and impulsive too... I love her anyway... lol...

Also, if you detached, I'm curious why your results don't indicate that... That means that the other person who has submitted a result will have to wait the full deadline time before another unit is reissued... Not blaming you, just I would've thought that the BOINC CC should've sent something to the project to say "hey, this user detached, go ahead and close out any open work they have"...

Still off topic

To detach a project with unfinished tasks is to me a serious matter that normally does not come lightly. Then again I felt very unwelcome with the EAH message board rules, which brought my blood to a boiling point and made me leave in a - for me as a dedicated BOINCer and founder of teams in most projects - hitherto unseen haste.

To the best of my knowledge such rules do not exist in any other project.The SETI message boards for instance receive their share of inappropriate posts as well. We have a very effective team of moderators, though. I am one myself, BTW, and FYI moderation works!

Annika
Annika
Joined: 8 Aug 06
Posts: 720
Credit: 494410
RAC: 0

Okay Brian, point taken. Of

Okay Brian, point taken. Of course I also hope that the next run will start more smoothly.
Does your overclocked 3700+ really take 130k seconds for one of the big WUs? Wow. I guess you're running Windows on that? But even then I'm surprised the difference is that big, because as I said, my 3500+ (Venice, meaning 512 KB of cache, and not overclocked) rarely takes much more than 100k seconds...

Brian Silvers
Brian Silvers
Joined: 26 Aug 05
Posts: 772
Credit: 282700
RAC: 0

RE: Still off topic I

Message 63819 in response to message 63817

Quote:

Still off topic

I don't see where the discussion is "OT", nor do I see the necessity of bolding the text. This implies that you are, in essense, shouting (shooting?) at me...

Sorry you felt unwelcome...

Brian

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.