Technically it is worse than that. ONe study I had in college days indicated that the "real" number of bugs actually stays constant. Each bug removed installs a new one that is more subtle and or less likely to cause problems.

The one in ten number is actually for each 10 lines of code the AVERAGE programmer makes one error. This goes down with skill level to rise to higher LOC counts with very skilled programmers only making a mistake about one in 200 LOC.

It probably depends whose statistics you use, and how you measure.

Besides, 87.4% of all statistics are made up.

Hehe.. It's like the DirecTV commercial a few months ago.. "95% of all statistics can be made to say anything... 50% of the time."

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

I am running the CUDA version and the last 20 workunits have finished in an average time of 3 minutes only. My system has two NVidia 260 OCs installed in SLI mode. Does CUDA take advantage of both GPUs? Also, since the NVidia 260 is one of the first GPUs to perform double precision floating point instructions, will this provide a significant edge. Does the SETI client use double precision or single precision floating point in its fourier transform module?

1) 3 minutes of CPU time (BOINC shows CPU time!) or wall clock time?
2) SLI not supported, you need 2 GPU w/o SLI mode to use them both.
(look nVidia site)
3) float in most places, only small part does in double.

I see it dose not pay to get winged 5.28 and 6.05.

1091890751 4687023 17 Dec 2008 21:38:45 UTC 18 Dec 2008 14:24:29 UTC Over Success Done 18,773.01 46.33 46.33
1091890752 4491788 17 Dec 2008 21:38:46 UTC 19 Dec 2008 10:18:12 UTC Over Success Done 253.20 59.77 46.33

Michael

Three of the 11 tasks have validated correctly.

---

I see it dose not pay to get winged 5.28 and 6.05.

1091890751 4687023 17 Dec 2008 21:38:45 UTC 18 Dec 2008 14:24:29 UTC Over Success Done 18,773.01 46.33 46.33
1091890752 4491788 17 Dec 2008 21:38:46 UTC 19 Dec 2008 10:18:12 UTC Over Success Done 253.20 59.77 46.33

Michael

It's hard to say from your example, since the wingman used <core_client_version>6.2.18</core_client_version>

---

Agreed but just a comment on different credit granting within projects.



Michael

I see it dose not pay to get winged 5.28 and 6.05.

1091890751 4687023 17 Dec 2008 21:38:45 UTC 18 Dec 2008 14:24:29 UTC Over Success Done 18,773.01 46.33 46.33
1091890752 4491788 17 Dec 2008 21:38:46 UTC 19 Dec 2008 10:18:12 UTC Over Success Done 253.20 59.77 46.33

Michael

4687023: gnnome's Power Mac seems to be claiming fairly.
4491788: Michael's (different Michael, not wandering willie) Q6600/CUDA seems to be consistently overclaiming.

Hi richard

I amazed that this was never Picked up over on Beta in what way is it over claiming
I thought that CUDA was worked out on flops'

Michael (wandering willie.)

Hi richard

I amazed that this was never Picked up over on Beta in what way is it over claiming
I thought that CUDA was worked out on flops'

Michael (wandering willie.)

I'm not surprised it wasn't, since probably two thirds of my tasks didn't have a wingman, and a majority of the tasks pending are still pending. I'm guesing the same was true for you.

---

Hi richard

I amazed that this was never Picked up over on Beta in what way is it over claiming
I thought that CUDA was worked out on flops'

Michael (wandering willie.)

Hang on, let's check all this out before jumping to conclusions.

This is the first report I've seen. Does it apply to all v6.4.5/v6.05 CUDA installations? or does it just apply to host 4491788? More data, please.

I looked at the first two tasks you posted: the flopcounts were different and roughly (by eye) in proportion to the credit inflation. I've got to go out soon, so someone else will have to do the detailed arithmetic.

WRT Beta: the earliest CUDA tasks back were re-issues against work done with the stock v6.03 application (edit - reference), which we know claims the same as v5.28. I didn't notice any credit inflation for those early results. Subsequently, newly issued Beta tasks have got caught up in the 'single redundancy mode' test and the upload server failure, so no new data have been available for comparative scrutiny.

Good morning again Richard

Thank you once again for your reply I will wait and see if any other reports come in on this subject.

It appears over here unless there is something like 0.0 credit granted no host report any thing.

Michael.

Just recieved a 0 credt task and a third wingman still to be sent out. I see the result is different from my wingman using 5.4.9 so I understand why the 3rd wingman, just wanted to report since it was with me using 6.05 CUDA

---

Just recieved a 0 credt task and a third wingman still to be sent out. I see the result is different from my wingman using 5.4.9 so I understand why the 3rd wingman, just wanted to report since it was with me using 6.05 CUDA

That's not a zero task - that's the AP validator bug which has now infected the MB validator.

Unusual to see tie-breaker tasks remaining unsent for so long, though (especially while the ready-to-send queue is so low). Did somebody mention a new scheduler a while back?

That's not a zero task - that's the AP validator bug which has now infected the MB validator.

Great, I keep getting all the good stuff. Maybe if I'm real good, all the remaining people sent out on it will be no reply too. ;P

And here I thought I'd just be crunching

---

Over here at the moment we are running MB from 5.28 upto 6.03.

Astropulse AP 5.00

Also MB 6.05 CUDA.

Which trying to compare 5.28 along side 6.03 I have had possible a 2 or 3 drop in credits.

The drop of almost 14 does ask the question. Why the difference in claimed?
As the WUs concerned were validated against each other on all WU concerned they must have been similar.Total of 4WUs.

So is any one else seeing this or is it just this host.

The origional querry was the difference between 5.28 and 6.05 I'm sure that I am not the only host to see this difference. In reported /claimed credits.

I am still waiting for my other host to complete which is also a Q6600.

Post 841931

Michael.*Wandering willie

Another example here (381743057). 6.4.5 CUDA claims 79.11, but 5.4.9 BOINC w/MB 6.03 claims 64.71 (granted 64.71).

CUDA flops: 27527396877983.301000
MB flops: 22515640652655.125000

AR is 0.386214

Hi richard

I amazed that this was never Picked up over on Beta in what way is it over claiming
I thought that CUDA was worked out on flops'

Michael (wandering willie.)

Hang on, let's check all this out before jumping to conclusions.

This is the first report I've seen. Does it apply to all v6.4.5/v6.05 CUDA installations? or does it just apply to host 4491788? More data, please.

I looked at the first two tasks you posted: the flopcounts were different and roughly (by eye) in proportion to the credit inflation. I've got to go out soon, so someone else will have to do the detailed arithmetic.

WRT Beta: the earliest CUDA tasks back were re-issues against work done with the stock v6.03 application (edit - reference), which we know claims the same as v5.28. I didn't notice any credit inflation for those early results. Subsequently, newly issued Beta tasks have got caught up in the 'single redundancy mode' test and the upload server failure, so no new data have been available for comparative scrutiny.

Thanks Randy for your input was beginning to think it was just me

Michael

Two questions:

1) My mac just gets "can't load library libcudart", so I'm assuming either the driver isn't compatible or nobody has ported the library to intel macs yet.

2) Given the normal mechanism for credit, it doesn't seem reasonable on its face to assign credit for a WU based on a cuda result to a non-cuda machine - the actual cobblestones expended are different.

---

--

Sorry for spamming this thread, but I have 11 workstations with 8 cores, and Quadro FX 4600 cards. None of them uses the Quadro card, is this because of some test in the beginning that estimates that crunching without GPU will be more efficient?

---

Checkout http://seti.smlug.net for optimized Sun binaries.

That estimation would probably be correct, but didn't heard of such feature.
EDIT: ah, those 11 hosts running Linux. Enchanced 6.05 is yet for Windows only.

---