But don't even think about such abortion before ;)

Well, after sleeping on it overnight, I decided that it was utterly foolish for me to waste so much of my processing time on such an inefficient app, so aborting those tasks was actually the only sensible thing to do. That was the first thing I've done here this morning, even before finishing my breakfast! I also aborted another AP v7 7.00 that was running on a different one of my hosts. That task was only 6.5% complete after limping along for more than 11 hours.

And if any more of those horrid apps come my way, I'll abandon them before they ever get started. It seems clear to me that the AP v7 7.00 app should never be sent to any host that has the capability of running one of the sse/sse2 or other app.

FYI >> Lunatics Windows Installer v0.43 Release Notes
Updated Installers, v0.43 for Windows Available now
LUNATICS team rocks ! :)

---

But don't even think about such abortion before ;)

Well, after sleeping on it overnight, I decided that it was utterly foolish for me to waste so much of my processing time on such an inefficient app, so aborting those tasks was actually the only sensible thing to do. That was the first thing I've done here this morning, even before finishing my breakfast! I also aborted another AP v7 7.00 that was running on a different one of my hosts. That task was only 6.5% complete after limping along for more than 11 hours.

And if any more of those horrid apps come my way, I'll abandon them before they ever get started. It seems clear to me that the AP v7 7.00 app should never be sent to any host that has the capability of running one of the sse/sse2 or other app.

The really sensible thing for all who looks on own BOINC setups more that one time per week is to install optimized apps. And of course not to abort tasks staying with stock apps.

My Mac is now averaging over 300 credits a task, http://setiathome.berkeley.edu/results.php?hostid=6796479&offset=0&show_names=0&state=4&appid=20
I'm still receiving 'normal' credits for APv6 tasks, http://setiathome.berkeley.edu/results.php?hostid=6796479&offset=0&show_names=0&state=4&appid=12

If we could get some more of those horrid tasks validated, perhaps APv7 will catch up to APv6.
perhaps...

My Mac is now averaging over 300 credits a task, http://setiathome.berkeley.edu/results.php?hostid=6796479&offset=0&show_names=0&state=4&appid=20
I'm still receiving 'normal' credits for APv6 tasks, http://setiathome.berkeley.edu/results.php?hostid=6796479&offset=0&show_names=0&state=4&appid=12

If we could get some more of those horrid tasks validated, perhaps APv7 will catch up to APv6.
perhaps...

I'm seeing the same, around 300 on my GTX770s.

And of course not to abort tasks staying with stock apps.

I didn't make that decision lightly. Out of nearly 400,000 tasks that my various hosts have received from S@H, those are the first 3 that I've ever aborted. I hated to do it, but that app should never have made it out of Beta with that kind of awful performance. Again, in my opinion CPU tasks for AP v7 7.00 should never be sent to machines that are capable of running any of the others. And anyone who gets one should abandon/abort it at the earliest opportunity.

If we could get some more of those horrid tasks validated, perhaps APv7 will catch up to APv6.
perhaps...

Seems to me that's a pretty poor reason for wasting other people's processing resources, just to artificially inflate some silly credit scores. That processing time can be better spent actually doing more work for the project.

If we could get some more of those horrid tasks validated, perhaps APv7 will catch up to APv6.
perhaps...

Seems to me that's a pretty poor reason for wasting other people's processing resources, just to artificially inflate some silly credit scores. That processing time can be better spent actually doing more work for the project.

Actually it works just the Opposite. People are leaving the tasks because they are not paying the same. Once the Credits have caught up, it's pretty much a done deal. They will catch up eventually anyway. This way they will catch up faster, more people will work the tasks sooner, everyone will be happy. People will not Abort tasks....all for a few days of a couple CPU cores. Seems a small price to pay, a few days of a couple CPU cores. It's not like asking for a couple GPUs that produce much more work, and everybody's GPUs would benefit greatly from the CPUs work. I just wish I had a couple Fast 8 core CPUs. It's not that much different from a few dozen people spending Months at Beta testing all this stuff...

What is all the fuss about, now that I am usually getting over 300 credits for a APV7 task in about 30 mins. They are paying more than MB tasks, and there is no indication that the increase in credits has stopped.

Stop whinging and get back to crunching.

It's not that much different from a few dozen people spending Months at Beta testing all this stuff...

+100.

If we could get some more of those horrid tasks validated, perhaps APv7 will catch up to APv6.
perhaps...

Seems to me that's a pretty poor reason for wasting other people's processing resources, just to artificially inflate some silly credit scores. That processing time can be better spent actually doing more work for the project.

Actually it works just the Opposite. People are leaving the tasks because they are not paying the same. Once the Credits have caught up, it's pretty much a done deal. They will catch up eventually anyway. This way they will catch up faster, more people will work the tasks sooner, everyone will be happy. People will not Abort tasks....all for a few days of a couple CPU cores. Seems a small price to pay, a few days of a couple CPU cores. It's not like asking for a couple GPUs that produce much more work, and everybody's GPUs would benefit greatly from the CPUs work. I just wish I had a couple Fast 8 core CPUs. It's not that much different from a few dozen people spending Months at Beta testing all this stuff...

Well, although I disagree with you wholeheartedly, I don't really want to get drawn into this whole credit system debate. That belongs in another thread anyway, and I don't go there.

And of course not to abort tasks staying with stock apps.

I didn't make that decision lightly. Out of nearly 400,000 tasks that my various hosts have received from S@H, those are the first 3 that I've ever aborted. I hated to do it, but that app should never have made it out of Beta with that kind of awful performance. Again, in my opinion CPU tasks for AP v7 7.00 should never be sent to machines that are capable of running any of the others. And anyone who gets one should abandon/abort it at the earliest opportunity.

I've only aborted tasks when there was very good reason; once when a quirk of BOINC delivered far more work than the host could do within deadline, and a case where a batch of tasks were corrupted in a fashion that they were guaranteed to eventually error (in that case I requested and aborted tasks until the Scheduler would not send any more, just to reduce the number of those tasks which would be sent to hosts of users who weren't aware of the problem).

There's an alternative to aborting tasks which I prefer, though it takes more effort. Exiting BOINC and deleting <file> and <workunit> sections for the task plus <file> and <result> sections for the expected result file from the client_state.xml file make the task become "lost". When BOINC is restarted it has less work so should ask for more fairly soon, and the Scheduler will resend the task. At that time it could be assigned to any app version which can do the work.

As my Core 2 Duo host is only doing CPU tasks and running 32 bit XP, at this point there's a 50% chance the resend will be for the SSE app version. It did finish 2 low blanked tasks with 7.00, and I feel that fulfilled my duty as far as contributing to the average for that app version is concerned. Subsequently I've forced 1 into that "lost" state to get it resent for SSE. A second try was needed before it actually worked. That task actually has over 47% blanking, so will be a runtime outlier anyhow. Since then there have been 2 more AP v7 tasks sent, both of which were assigned to the SSE app without any special treatment. They aren't started yet, I'm hoping for low blanking so the host can make some advance toward having 11 SSE tasks completed.

There has been the feeling that the slowest CPU app version has a major effect on the credit scaling. I don't think so, but there is some evidence that Eric may. That's another possible reason for releasing the 7.00 CPU app for Windows without attempting to improve it.

                                                                  Joe

My Mac is now averaging over 300 credits a task, http://setiathome.berkeley.edu/results.php?hostid=6796479&offset=0&show_names=0&state=4&appid=20
I'm still receiving 'normal' credits for APv6 tasks, http://setiathome.berkeley.edu/results.php?hostid=6796479&offset=0&show_names=0&state=4&appid=12

If we could get some more of those horrid tasks validated, perhaps APv7 will catch up to APv6.
perhaps...

@TBar: I just checked your links and you can see an increase over time which is good. The approximate breakeven point between v6 and v7 was circa 484 when I reviewed my data/results which would tend to suggest that there is a fair way to go yet, but it is heading in the right direction.

---

What is all the fuss about, now that I am usually getting over 300 credits for a APV7 task in about 30 mins. They are paying more than MB tasks, and there is no indication that the increase in credits has stopped.

Stop whinging and get back to crunching.

@WinterKnight: I had a look at your results. Some of them are over 300 but the majority of your last 40 results on each machine are under 300. If I was to take out the outliers, your average would appear to be under 300 in just looking at the numbers. It will be interesting to see how your results change over the next few days.

As to MB, when I looked at my data the average for APv7 was 265 with credit per second per WU at 0.93654, for MBv6 the credit per second per WU was 0.1303, implying a breakeven of circa 369 for APv7.

cheers

L.

---

I just looked at some of my validated results and things may actually be looking up. I've had a series of 300+ credit results and actually had one over 700! I'll keep watching.

---

There's an alternative to aborting tasks which I prefer, though it takes more effort. Exiting BOINC and deleting <file> and <workunit> sections for the task plus <file> and <result> sections for the expected result file from the client_state.xml file make the task become "lost". When BOINC is restarted it has less work so should ask for more fairly soon, and the Scheduler will resend the task. At that time it could be assigned to any app version which can do the work.

Okay, thanks Joe. Perhaps I'll try that technique if I get another one. I don't particularly like seeing aborted or abandoned tasks in my database.

...but there is some evidence that Eric may. That's another possible reason for releasing the 7.00 CPU app for Windows without attempting to improve it.

Rolls eyes and shakes head disgustedly.

I just looked at some of my validated results and things may actually be looking up. I've had a series of 300+ credit results and actually had one over 700! I'll keep watching.

@ExchangeMan: I just had a look at your results and they have definitely trended up. At the moment I would say that you seem to be landing around the breakeven for MB that I calculated.

cheers

---

The average of the results here, since midnight, on my computer is about 700/hr on V7. The average for the V6's reported in the 5 hrs before midnight was ~900/hr.

Those seven V6's only had one high blanking task in the mix, where as at the moment I have been getting about 20% of high blanked V7's.

So for me I think we are already on even parity.

edit] Last batch of MB's done on the GPU was ~440/hr

Further to my last, after doing a few more AP V6's and a bunch of V7's. and putting the results into ye old spreadsheet.

If you compare doing V6 with two high blanking tasks in 24 hours against V7 with credits as they are now, mine are between 312 and 370, then the daily totals would be ~17,600 for V6 and ~17,400 for V7 per computer running 24/7.

Hi Folks,
Well it looks as if AP7 is a tad iffy right now, so I've just changed my setting to remove it from the types I'm wanting to do.

Frankly I've already had a problem with long running WU that accrue little credit, while others that take only a 10th of the time to process garner a lot more.

Seems to me that whomever is responsible for sorting out credit etc should decide how many punters are doing projects for the science alone and how many are doing projects for the science AND credits they can accrue. Then decide if punters that are involved as much for the bragging rights of high RAC are of less importance to the project than the pure science folks.

As it stands some projects allocate 3k3 credits for 1 type of WU and 1k for another, while S@H allocates 60 or 100 for WU that take the same amount of time and electricity for users to process..

I guess in the end it will all sort itself out:-)

When it does I'll select APv7 again

Regards,

---

Cliff,
Been there, Done that, Still no damm T shirt!