Well, I use the same driver under Vista too...
Maybe 180.80 ? IT's recommended for using with CUDA 2.1 ... And look this thread on beta:
http://setiathome.berkeley.edu/beta/forum_thread.php?id=1461
It's called "my lucky nvidia driver"....

Hm, it's some method but I don't think you like it. You need to check all incoming tasks headers and manually abort VeryLowAngleRange ones....
Too much manual work to be effective unless you write some script for it.
If you familiar with Perl it will be not so hard IMO.
...

Or check in client_state.xml, any VLAR workunit will have:

<rsc_fpops_est>80360000000000.000000</rsc_fpops_est>

That value will not be set for anything other than a VLAR WU, and the WU name is just three lines above. The corresponding task in BOINC Manager will have the same name with a _x appended where x is 0 to 9.

                                                             Joe

I'll try to get some new workunits, hopefully non-VLAR ones.

With ten new ones it worked well, apart from one which errored out with a -9 result_overflow.

I noticed that the CPU usage in Win Taskmanager was approx. 10% (ranging from 7 to 14 % or so), so there should be some slowdown of the CPU-bound tasks, since the CPU cores have to feed the GTX260. That makes sense. I was running two Docking@home task in parallel, which got the remaining 90% CPU usage.

AR : 0.636407 - time : 507 s - credits/day : 9000
AR : 2.720792 - time : 236 s - credits/day : 5304
AR : 6.139079 - time : 236 s - credits/day : 5136
AR : 2.715539 - time : 234 s - credits/day : 5352
AR : 2.715539 - time : 237 s - credits/day : 5280
AR : 2.715539 - time : 238 s - credits/day : 5256
AR : 2.720792 - time : 238 s - credits/day : 5256
AR : 2.715539 - time : 21 s, this one errored out with -9
AR : 2.715278 - time : 236 s - credits/day : 5304
AR : 0.636407 - time : 504 s - credits/day : 9048

The above given credits per day are based on the claimed credits (the results are still pending) and the above given wall-clock-times (not cpu times). Quite impressive.

A few days ago I tested four workunits with AR ~ 2.7 with one core available to SETI CUDA 6.05 alone and they were finishing in an average wall-clock-time of 248 seconds each, which resulted in 5112 credits per day. So with the modified application I'm getting a similar or perhaps even better performance for SETI CUDA plus 100% CPU usage on both cores. Great, thx again, Raistmer!

@ the project team: It would have been even better if we could have figured this out over on Beta first before releasing 6.05 on the main project. I know this has already been said by some of my fellow testers, just wanted to express that I'm feeling the same.

Last note: With the current configuration, that means BM 6.45 w/ cc_config tweak, I'm having some problems to get 2+1 (CPU+GPU) tasks running when doing GPUGRID calculations on the GPU. BM sometimes switches to 3+1, which means three CPU-bound tasks on the dualcore, and when I set processor settings in the local BM settings to 90% to reduce the tasks to 2+1 again, then BM might switch later to 1+1... well, remember Sisyphus? Hence I'll try to revert to 6.4.2 (as currently recommended by GPUGRID, without cc_config) later today, to see if it helps in this regard.

Alex

[OT]

...
AR : 2.715539 - time : 21 s, this one errored out with -9
...
Alex

As often explained, a

SETI@Home Informational message -9 result_overflow
NOTE: The number of results detected exceeds the storage space allocated.

isn't an error. The host correctly reports it as "Success" and if it is validated by another host the claimed credit is granted. I've often wondered if they simply removed the "-9" from that whether users would see and understand the rest of the surrounding text. What do you think?
[/OT]

                                                                    Joe

[OT]

As often explained, a

SETI@Home Informational message -9 result_overflow
NOTE: The number of results detected exceeds the storage space allocated.

isn't an error.

Yep, thanks for pointing out, my bad. However, my wingman for this workunit is crunching with BM 5.10.30 and app. version 6.03 up to now. I'm curious if he will also have an excessive number of results for this workunit. It has been pointed out elsewhere on the board that there are quite a lot mismatches between CUDA results and results calculated on the CPU.

I've often wondered if they simply removed the "-9" from that whether users would see and understand the rest of the surrounding text. What do you think?

Hmm.. I'd say the second line is sufficient as an "informational message". One could add something like 'calculation aborted' for further explanation, but maybe that's already too verbose.

[/OT]

Regards
Alex

2) It consist of modified SETI MB CUDA and current SSE3 opt SETI AP binaries with corresponding app_info.xml file

Well, you never know when you need it, Astropulse, I mean. Just got one:


22.12.2008 05:36:46|SETI@home|Sending scheduler request: To fetch work. Requesting 645776 seconds of work, reporting 0 completed tasks
22.12.2008 05:36:56|SETI@home|Scheduler request completed: got 1 new tasks
22.12.2008 05:36:56|SETI@home|Message from server: No work can be sent for the applications you have selected
22.12.2008 05:36:56|SETI@home|Message from server: No work is available for SETI@home Enhanced
22.12.2008 05:36:56|SETI@home|Message from server: You have selected to receive work from other applications if no work is available for the applications you selected
22.12.2008 05:36:56|SETI@home|Message from server: Sending work from other applications
22.12.2008 05:36:58|SETI@home|Started download of ap_02dc08ai_B1_P0_00388_20081221_32379.wu


BM tells me it will take 9337 hours to process .. lol, I doubt it. That would make my CPDN workunits almost a blink of an eye.

If BM was in panic mode before - and it was - will it now going rampage?

Maybe 'berserkley mode'?

... sorry ... where's my hankie?

Don't worry, actual time will be MUCH shorter.

I receive MB workunits with estimated calculation time about 3h - but actually I need just a couple of minutes to finish them. The same with AP - initial estimated time is over 120h, but actual is much shorter.

Have been running Einstein, AP, and CUDA MB on my AMD. Have not noticed any slow down from previous tasks, and I'm able to process the Einstein tasks, but had to manually switch between AP and Einstein while running them.

The AP is also showing no slow down, so I expect it will also complete in the same amount of time as preivious ones done.

Very much like my experience with CUDA at Beta, I expect CUDA will continue on this pc :)

Edit: DCF at 0.227498 now after I reset it to 0.300000 a few days ago.

---

Hi Raistmer

Now running 1 cuda, 1 Einstein, 1 Astropulse.
The cuda task runs about twice as fast as the normal optimised version from lunatics. So the 1 cuda task is a bit faster than 2 optimised seti tasks.
CPU E6850 @ 3.00GHz [x86 Family 6 Model 15 Stepping 11] Number of CPUs 2
Coprocessors [CUDA|GeForce 8800 GTX|1|768MB]
Running fine on Vista, but is sometimes a bit sluggish I think the priority of the task is a bit to high or takes to much time.

Hi Raistmer


Running fine on Vista, but is sometimes a bit sluggish I think the priority of the task is a bit to high or takes to much time.

Well, probably you right. But its' BOINC's bug again ;) It runs CUDA tasks with normal (!) process priority instead of idle one.
So worker thread priority will be 7, not ~2-3 as intended.

Just installed your mod and now enjoy 10x increase of speed of computing. Thanks a lot %)

But it looks like now I still don't use all cores at 100% efficacy.

PC has 2 cores + GPU (8800).
I have AP and MB units running together. Judging by Task manager AP creates about 45-50% of CPU workload and MB (crunched mostly at GPU) - about 5-10%. So total CPU workload is about 60%. Are there any ways to utilize 100% by parallel computing of one more WU?

You need to read my first message completely.
Statement "7)"
7) For best CPU and GPU usage I recommend to set number of processors available for BOINC to real_number_of_cores+1. This will mitigate current BOINC bug with CPU+CUDA scheduling and will allow fully load CPU and GPU.
Did you make it?

i have tried this aswell (i didnt have a cc_config.xml file to start with) and i only have the CUDA application running, while the 2 cores of my CPU are sitting idle.

is there someone that can help me fix this? im not the best when it comes to programming/technical stuff

---

i have tried this aswell (i didnt have a cc_config.xml file to start with) and i only have the CUDA application running, while the 2 cores of my CPU are sitting idle.

is there someone that can help me fix this? im not the best when it comes to programming/technical stuff

I don't see that you have any Einstein tasks. Do you have any AP tasks? You would have to have either an AP task or Einstein task, (or other project if you decide to join another) before the cpu's have work.

---

i dont have any AP tasks because i havent been assigned any, so ho do i go about getting some assigned? its all setup in the user settings to allow me to get astro pulse

---

it seemed to be working fine for a day or two but now it seems i'm getting a slew of computational errors:

http://setiathome.berkeley.edu/result.php?resultid=1098999346

Cuda error 'find_pulse_kernel2<5, true>' in file 'd:/BTR/seticuda/seti_boinc/client/cuda/cudaAcc_pulsefind.cu' in line 1178 : unknown error.

this only started to happen recently (in the last day). perhaps something different with the work units that are being sent?

i have an 8800GTX stock. i updated the nvidia drivers to latest (180.48) and also installed the latest BOINC client (6.5.0) but i'm still getting the error.

any thoughts?? :)

thanks again for your hard work and port of this application! :)

---

it seemed to be working fine for a day or two but now it seems i'm getting a slew of computational errors:

http://setiathome.berkeley.edu/result.php?resultid=1098999346

Cuda error 'find_pulse_kernel2<5, true>' in file 'd:/BTR/seticuda/seti_boinc/client/cuda/cudaAcc_pulsefind.cu' in line 1178 : unknown error.

This only started to happen recently (in the last day). Perhaps something different with the work units that are being sent?

I have an 8800GTX stock. I updated the nvidia drivers to latest (180.48) and also installed the latest BOINC client (6.5.0) but I'm still getting the error.

Any thoughts?? :)

Thanks again for your hard work and port of this application! :)

Try the 180.84 driver maybe?

---

Savoir-Faire is everywhere!
The T1 Trust, T1 Class 4-4-4-4 #5550, America's First HST

i dont have any AP tasks because i havent been assigned any, so ho do i go about getting some assigned? its all setup in the user settings to allow me to get astro pulse

About the only way I have found to be mostly guaranteed to get some AP tasks is to adjust your cache to 4+ days. Of course this isn't 100%, but I have pretty good luck with that.

But as far as I know, at the moment, the splitters are trying their best, but the number of AP tasks ready to send are um.. zero. APs likely won't be sent out for a short while until they start splitting faster. Plenty of MBs, but no APs.

---

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

Copy this in e.g. notebook

<cc_config>
<options>
<ncpus>3</ncpus>
</options>
</cc_config>

and save this as a plain text file with the name: cc_config.xml
Move this file into the BOINC data directory.

i dont have any AP tasks because i havent been assigned any, so ho do i go about getting some assigned? its all setup in the user settings to allow me to get astro pulse

The easiest way I've found to get AP assigned, would be to go into your SETI@home preferences and adjust the settings so that AP is the only thing you can get. Usually once you do this it doesn't take very long to be assigned an AP, and once you get however many you want, you then can change the settings back to be assigned MB as well as AP.

---

i dont have any AP tasks because i havent been assigned any, so ho do i go about getting some assigned? its all setup in the user settings to allow me to get astro pulse

The easiest way I've found to get AP assigned, would be to go into your SETI@home preferences and adjust the settings so that AP is the only thing you can get. Usually once you do this it doesn't take very long to be assigned an AP, and once you get however many you want, you then can change the settings back to be assigned MB as well as AP.

Hm.. I suppose that method works as well. Just a little extra work to have to go and modify preferences and all of that. As I said, I found that setting a larger cache when your list is relatively low on work typically does the trick, but usually only if there are AP tasks available for assignment.

---

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

Well, maybe just to sign up with Eisstein@home? They do very interesting work too BTW.
These tricks to get some AP tasks looking very boring indeed.

About different work - sure, tasks with different ARs can be treated as different work in case of app stability. And it's known bug already - CUDA MB can't handle very low AR tasks. And, maybe, not only that.