This is using the exact same parameter files and star files as one of the previously running searches that no one supposedly was having problems with. Let me know if you have any issues with these workunits (a link to failed tasks would be great as well).

Thanks!
--Travis

---

I just downloaded six work units.

ps_separation_10_2s_sample_2_1340672783_37277_0
ps_separation_10_2s_sample_2_1340672783_37278_0
ps_separation_10_2s_sample_2_1340672783_37279_0
ps_separation_10_2s_sample_2_1340672783_37280_0
ps_separation_10_2s_sample_2_1340672783_37281_0
ps_separation_10_2s_sample_2_1340672783_37286_0

All failed within 20 seconds of starting. If there's anything I can try, please let me know.

I just downloaded six work units.

ps_separation_10_2s_sample_2_1340672783_37277_0
ps_separation_10_2s_sample_2_1340672783_37278_0
ps_separation_10_2s_sample_2_1340672783_37279_0
ps_separation_10_2s_sample_2_1340672783_37280_0
ps_separation_10_2s_sample_2_1340672783_37281_0
ps_separation_10_2s_sample_2_1340672783_37286_0

All failed within 20 seconds of starting. If there's anything I can try, please let me know.

Try restarting your computer? It looks like your GPU is out of resources (maybe the GPU is also being used by something else?)

---

I tried rebooting, reinstalling video drivers (301.42,) and nothing is alleviating it. The only program running my GTX 560Ti is BOINC. This only started in the last couple of days, and I haven't made any PC changes.

Oh, and my PC picked up 25 new tasks at 0308 EDT, and then proceeded to have computation errors for each and every one of them within the next three minutes. They were all in this new batch of work units. The units that failed are as follows.

ps_separation_10_2s_sample_2_1340672783_133364_0
ps_separation_10_2s_sample_2_1340672783_133331_0
ps_separation_10_2s_sample_2_1340672783_133354_0
ps_separation_10_2s_sample_2_1340672783_127245_1
ps_separation_10_2s_sample_2_1340672783_133365_0
ps_separation_10_2s_sample_2_1340672783_133344_0
ps_separation_10_2s_sample_2_1340672783_133356_0
ps_separation_10_2s_sample_2_1340672783_133367_0
ps_separation_10_2s_sample_2_1340672783_133353_0
ps_separation_10_2s_sample_2_1340672783_133368_0
ps_separation_10_2s_sample_2_1340672783_133366_0
ps_separation_10_2s_sample_2_1340672783_133359_0
ps_separation_10_2s_sample_2_1340672783_133343_0
ps_separation_10_2s_sample_2_1340672783_127255_1
ps_separation_10_2s_sample_2_1340672783_133349_0
ps_separation_10_2s_sample_2_1340672783_133342_0
ps_separation_10_2s_sample_2_1340672783_133345_0
ps_separation_10_2s_sample_2_1340672783_133348_0
ps_separation_10_2s_sample_2_1340672783_133363_0
ps_separation_10_2s_sample_2_1340672783_133355_0
ps_separation_10_2s_sample_2_1340672783_127244_1
ps_separation_10_2s_sample_2_1340672783_133362_0
ps_separation_10_2s_sample_2_1340672783_127254_1
ps_separation_10_2s_sample_2_1340672783_133341_0
ps_separation_10_2s_sample_2_1340672783_133347_0

I'm just going to detach from the project and reattach at this point and see if that works at all.

Okay, I detached from the project and reattached to it. Once reattached, BOINC downloaded two work units.

ps_separation_10_2s_sample_2_1340672783_199705_0 - Milky Way 1.02 NVidia OpenCL - Failed in 1 second

ps_separation_10_2s_sample_2_1340672783_199706_0 - Milky Way 1.00 - Still running without an issue.

PC specs as follows
AMD FX6100
16 GB RAM
Windows 7 Home Premium x64 SP1
BOINC v7.0.25
NVidia GTX560Ti - driver 301.42

This issue only started in the last couple of days, and I've changed nothing on my PC in that time, so I don't know what it is.

v0.82 ATI CAL application on Windows7 x64: not looking good. They error out the same way as the ps_separation_14 did (exit code -1073740940).


Here the stderr of the first task (with shotened chunk size list):

ps_separation_10_2s_sample_2_1340672783_134685_0

<core_client_version>6.12.34</core_client_version>
<![CDATA[
<message>
 - exit code -1073740940 (0xc0000374)
</message>
<stderr_txt>
Using SSE3 path
Found 1 CAL devices
Chose device 0

Device target:         CAL_TARGET_670
Revision:              41
CAL Version:           1.4.1546
Engine clock:          720 Mhz
Memory clock:          900 Mhz
GPU RAM:               512
Wavefront size:        64
Double precision:      CAL_TRUE
Compute shader:        CAL_FALSE
Number SIMD:           4
Number shader engines: 1
Pitch alignment:       256
Surface alignment:     4096
Max size 2D:           { 8192, 8192 }

Estimated iteration time 612.651910 ms
Target frequency 120.000000 Hz, polling mode 4
Dividing into 73 chunks, initially sleeping for 0 ms
Integration range: { nu_steps = 640, mu_steps = 1600, r_steps = 1400 }
Using 73 chunk(s) with sizes:   16   16   32   16   16   32   (...)
Integration time = 807.073512 s, average per iteration = 1261.052362 ms
Integral 0 time = 809.347605 s
Likelihood time = 4.666627 s
<background_integral> 0.000794396822494 </background_integral>
<stream_integral>  1900.813437891031500  1355.257476385759900 </stream_integral>
<background_likelihood> -7.521535247136720 </background_likelihood>
<stream_only_likelihood>  -3.448139555488806  -6.475711963571757 </stream_only_likelihood>
<search_likelihood> -3.447338749756394 </search_likelihood>

</stderr_txt>
]]>

For comparison the result of a wingman:

<background_integral> 0.000794396822494 </background_integral>
<stream_integral>  1900.813437891031800  1355.257476385759900 </stream_integral>
<background_likelihood> -7.521535247136718 </background_likelihood>
<stream_only_likelihood>  -3.448139555488806  -6.475711963571757 </stream_only_likelihood>
<search_likelihood> -3.447338749756394 </search_likelihood>

So it calculates right and "decides" than, that it's an error.


The other two tasks:
ps_separation_10_2s_sample_2_1340672783_240053_0
ps_separation_10_2s_sample_2_1340672783_240061_0

---

v0.82 ATI CAL application on Windows7 x64: not looking good. They error out the same way as the ps_separation_14 did (exit code -1073740940).


Here the stderr of the first task (with shotened chunk size list):

ps_separation_10_2s_sample_2_1340672783_134685_0

<core_client_version>6.12.34</core_client_version>
<![CDATA[
<message>
 - exit code -1073740940 (0xc0000374)
</message>
<stderr_txt>
Using SSE3 path
Found 1 CAL devices
Chose device 0

Device target:         CAL_TARGET_670
Revision:              41
CAL Version:           1.4.1546
Engine clock:          720 Mhz
Memory clock:          900 Mhz
GPU RAM:               512
Wavefront size:        64
Double precision:      CAL_TRUE
Compute shader:        CAL_FALSE
Number SIMD:           4
Number shader engines: 1
Pitch alignment:       256
Surface alignment:     4096
Max size 2D:           { 8192, 8192 }

Estimated iteration time 612.651910 ms
Target frequency 120.000000 Hz, polling mode 4
Dividing into 73 chunks, initially sleeping for 0 ms
Integration range: { nu_steps = 640, mu_steps = 1600, r_steps = 1400 }
Using 73 chunk(s) with sizes:   16   16   32   16   16   32   (...)
Integration time = 807.073512 s, average per iteration = 1261.052362 ms
Integral 0 time = 809.347605 s
Likelihood time = 4.666627 s
<background_integral> 0.000794396822494 </background_integral>
<stream_integral>  1900.813437891031500  1355.257476385759900 </stream_integral>
<background_likelihood> -7.521535247136720 </background_likelihood>
<stream_only_likelihood>  -3.448139555488806  -6.475711963571757 </stream_only_likelihood>
<search_likelihood> -3.447338749756394 </search_likelihood>

</stderr_txt>
]]>

For comparison the result of a wingman:

<background_integral> 0.000794396822494 </background_integral>
<stream_integral>  1900.813437891031800  1355.257476385759900 </stream_integral>
<background_likelihood> -7.521535247136718 </background_likelihood>
<stream_only_likelihood>  -3.448139555488806  -6.475711963571757 </stream_only_likelihood>
<search_likelihood> -3.447338749756394 </search_likelihood>

So it calculates right and "decides" than, that it's an error.


The other two tasks:
ps_separation_10_2s_sample_2_1340672783_240053_0
ps_separation_10_2s_sample_2_1340672783_240061_0

So Matt Arsenault is telling me that the ATI 0.82 is deprecated and you should be using the newer OpenCL version.

---

So Matt Arsenault is telling me that the ATI 0.82 is deprecated and you should be using the newer OpenCL version.

Not possible on an ATI HD3850, it has no OpenCL support.

Well, I guess I'll have to live with Collatz as project for my GPU...

---

Link, don't worry. For example me too. :(

Sadly, my "Ultra-super-GOD-Like" Mobile Radeon HD 5870 1GB GDDR5 does NOT have double precision! (needed for MilkyWay)

I need NBody work units for my 8-threaded CPU, please !
I am not willing to waste CPU time to calculate separation WHICH double-precision GPUs can do 100 or 800 times faster! :)

Sadly, my "Ultra-super-GOD-Like" Mobile Radeon HD 5870 1GB GDDR5 does NOT have double precision!

HD 5870 definitely should support DP (maybe not the mobile version), however I don't see any in your copmuter list. Only one 5700 series GPU and one HD 6250/6310, which indeed have no DP. And than there's one not properly recognized GPU with Cedar core, which makes it 5400 series, so no DP again.

All HD 58x0 series GPUs have the "Cypress" core, which supports DP, so either your card was not properly detected by the drivers or you don't have a HD 5870. See also Wikipedia/Evergreen.


EDIT: now I see the word "Mobile"... yeah, they are actually not HD58x0 but HD57x0, as BOINC correctly detects.

---

I have had no troubles at all with the new separation units, unlike the old ones, using a Radeon 6950 OC. Thanks and happy crunching! :)

So Matt Arsenault is telling me that the ATI 0.82 is deprecated and you should be using the newer OpenCL version.

Not possible on an ATI HD3850, it has no OpenCL support.

Just processed 6 'ps_separation_10_2s_sample_2' tasks quite happily on my HD3850.

O/S is XP SP3 (32-bit).

I wonder if my earlier message on a different thread simply didn't get through. I'll try again here:-

HD3850 XP SP3 (32-bit) works.
HD3850 W7 SP1 (64-bit) failed.

I wonder if my earlier message on a different thread simply didn't get through. I'll try again here:-

HD3850 XP SP3 (32-bit) works.
HD3850 W7 SP1 (64-bit) failed.

IIRC, we can make

ATI v0.82 CAL app XP SP3 (32-bit) works.
ATI v0.82 CAL app W7 SP1 (64-bit) failed.

out of it. The issues were not limited to the HD3850 GPUs.

---

we can make

ATI v0.82 CAL app XP SP3 (32-bit) works.
ATI v0.82 CAL app W7 SP1 (64-bit) failed.

out of it. The issues were not limited to the HD3850 GPUs.

I kinda guessed because so many people said (or implied) that the "latest" WUs caused problems on their various GPUs runing W7 64 bit.

Reminder:- I rebuilt my W7 SP1 (64-bit) PC back to XP SP3 (32-bit) to make the "latest" WUs work on my HD3850 AGP hardware.

Kindly, I ask ... what's the answer?

we can make

ATI v0.82 CAL app XP SP3 (32-bit) works.
ATI v0.82 CAL app W7 SP1 (64-bit) failed.

out of it. The issues were not limited to the HD3850 GPUs.

I kinda guessed because so many people said (or implied) that the "latest" WUs caused problems on their various GPUs runing W7 64 bit.

Reminder:- I rebuilt my W7 SP1 (64-bit) PC back to XP SP3 (32-bit) to make the "latest" WUs work on my HD3850 AGP hardware.

Kindly, I ask ... what's the answer?

I'm not quite sure what the issue is with that old ATI code that Matt A. built. He seems pretty adamant about not supporting it anymore in favor of OpenCL (which I can understand because it makes the code base much easier to maintain).

Our code IS open source however, so if someone really wanted to they could grab the code from github and update the old ATI version to play nice:

https://github.com/Milkyway-at-home/milkywayathome_client/

---

Our code IS open source however, so if someone really wanted to they could grab the code from github and update the old ATI version to play nice:

https://github.com/Milkyway-at-home/milkywayathome_client/

I can't update the code, but if someone does and needs a tester, I'd be happy to test.

---

I tried running more tasks with the same results - compute fails.

http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=243957513
http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=243957512
http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=243957511

I just put up the three latest links. Any help in discovering why this is all of a sudden not working properly would be appreciated. Until then, I just suspended work on NVidia OpenCL tasks since the one CPU task I downloaded recently worked fine.

I tried running more tasks with the same results - compute fails.

http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=243957513
http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=243957512
http://milkyway.cs.rpi.edu/milkyway/result.php?resultid=243957511

I just put up the three latest links. Any help in discovering why this is all of a sudden not working properly would be appreciated. Until then, I just suspended work on NVidia OpenCL tasks since the one CPU task I downloaded recently worked fine.

Fowarded it on, hopefully we'll figure something out. :)

---

ATI v0.82 CAL app XP SP3 (32-bit) works.
ATI v0.82 CAL app W7 SP1 (64-bit) failed.

OK, I had an idea and eventually found the solution: use the x86 (32-bit) application.

Eventually because all 4 WUs I tested went into "validation inconclusive" status, see here. Maybe the server doesn't trust my computer after all the errors... I wouldn't at least.

Anyone who wants to test that approach has to do two things:

1. Download milkyway_separation_0.82_windows_intelx86__ati14.exe

2. Use this app_info (eventually modify or delete the cmd parameters or change to 1 WU/GPU if you want to, I just post my here as something to start with):

<app_info>
 <app>
  <name>milkyway</name>
  <user_friendly_name>MilkyWay@Home</user_friendly_name>
 </app>
 <file_info>
  <name>milkyway_separation_0.82_windows_intelx86__ati14.exe</name>
  <executable/>
 </file_info>
 <app_version>
  <app_name>milkyway</app_name>
  <version_num>82</version_num>
  <avg_ncpus>0.05</avg_ncpus>
  <max_ncpus>0.05</max_ncpus>
  <plan_class>ati14ati</plan_class>
  <coproc>
   <type>ATI</type>
   <count>0.50</count>
  </coproc>
  <cmdline>--gpu-target-frequency 120 --gpu-polling-mode 4 --gpu-disable-checkpointing</cmdline>
  <file_ref>
   <file_name>milkyway_separation_0.82_windows_intelx86__ati14.exe</file_name>
   <main_program/>
  </file_ref>
 </app_version>
</app_info>

So, now I will wait for those WUs to validate (or not), I have 3.5 days cache from Collatz to run down anyway.


EDIT: OK, 3 of them have validated so far, the last one was send to a CPU, so it's going to take a little longer, but in general it looks good.

---

I guess it got fixed. I haven't had a failed work unit since 6/27. Awesome job guys!