We're working on the next GW search, "O3AS1HF" (O3 data, All-Sky, High Frequency). There's still a change to be implemented and tested in the application. We hope to get the search started next week.
However, a few workunits (~100k) of the old O3MD1 search will need to be re-issued.
The last batch of data for O3MD1 I got was by far too big. I still have 131 tasks in spare, but do crunch only 11,5 tasks/day, and the date of expire is tomorrow about 15:55. So about 118 tasks will expire. Is it possible to shift the date of expire for the remaining tasks. I believe I'm not the only one.
The last batch of data for O3MD1 I got was by far too big. I still have 131 tasks in spare, but do crunch only 11,5 tasks/day, and the date of expire is tomorrow about 15:55. So about 118 tasks will expire. Is it possible to shift the date of expire for the remaining tasks. I believe I'm not the only one.
No, deadlines are set serverside the second a tasks is assigned to a host which requested more work. The best you can do now: Estimate which tasks you will finish until their expiry date (and time of day). Abort all other O3MD1 tasks manually in the BOINC manager. This way they will be scheduled to other users ASAP. Einstein implemented some clever task locality scheme into BOINC's server software. Tasks aren't assigned randomly to users, but if possible to hosts which already downloaded the corresponding raw data files. The same set of raw data files is analyzed by many (thousands?) of different tasks applying different science parameters. Therefore, I assume there will be few additional load on download servers, due to aborting such O3MD1 tasks (different story for e.g. BRP4). This way your "lost tasks" will be drained through the einstein pipeline ASAP.
Important: Please check your account preferences (einstein website). Check "network usage" --> "store at least" [days] of work, "store up to an additional" [days] of work.
The estimated task duration logic for einstein tasks is not perfect. There are many threads discussing the reasons (quite complex) somewhere in the forums. So, if your work load changes, new science runs (like current O3MD1) starts, the estimated runtimes are often far from reality. [EDIT:] If they are way too short, your hosts requests way to much work. It isn't the einstein server assigning too much work to your computer, but your computer is requesting too much (see log of last server request at your host's page on einstein website).[/EDIT:] We (users) have to adapt. Einstein works best with very small task caches. It's not necessary to store many days as servers are very reliable, rarely offline for maintenance.
[EDIT:]I haven't seen any of the current O3MD1 tasks. Are there CPU tasks or only GPU? Tech news here, not Einstein Café.
This old man needs some help. Right now, I'm running twin Nvidia GTX 1070s and only one is running the gavitational-wave-search data.
08/18/2023 21:31:35 | | Starting BOINC client version 7.24.1 for windows_x86_64
08/18/2023 21:31:35 | | log flags: file_xfer, sched_ops, task, cpu_sched, sched_op_debug
08/18/2023 21:31:35 | | Libraries: libcurl/8.2.1-DEV Schannel zlib/1.2.13
08/18/2023 21:31:35 | | Data directory: D:\BOINC
08/18/2023 21:31:35 | | Running under account clhar
08/18/2023 21:31:36 | | CUDA: NVIDIA GPU 0: NVIDIA GeForce GTX 1070 (driver version 531.29, CUDA version 12.1, compute capability 6.1, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | CUDA: NVIDIA GPU 1: NVIDIA GeForce GTX 1070 (driver version 531.29, CUDA version 12.1, compute capability 6.1, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | OpenCL: NVIDIA GPU 0: NVIDIA GeForce GTX 1070 (driver version 531.29, device version OpenCL 3.0 CUDA, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | OpenCL: NVIDIA GPU 1: NVIDIA GeForce GTX 1070 (driver version 531.29, device version OpenCL 3.0 CUDA, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | OpenCL: Intel GPU 0 (ignored by config): Intel(R) UHD Graphics 630 (driver version 31.0.101.2125, device version OpenCL 3.0 NEO, 6396MB, 6396MB available, 230 GFLOPS peak)
08/18/2023 21:31:36 | | Windows processor group 0: 20 processors
08/18/2023 21:31:36 | | Host name: A-SYS
08/18/2023 21:31:36 | | Processor: 20 GenuineIntel Intel(R) Core(TM) i9-10900 CPU @ 2.80GHz [Family 6 Model 165 Stepping 5]
08/18/2023 21:31:36 | | OS: Microsoft Windows 11: Professional x64 Edition, (10.00.22621.00)
It seems that only one task is running by default. What is the sort name for these tasks, and how can I get both GPUs to work. Also how can I get the cuda tasks to run?
This old man needs some help. Right now, I'm running twin Nvidia GTX 1070s and only one is running the gavitational-wave-search data.
08/18/2023 21:31:35 | | Starting BOINC client version 7.24.1 for windows_x86_64
08/18/2023 21:31:35 | | log flags: file_xfer, sched_ops, task, cpu_sched, sched_op_debug
08/18/2023 21:31:35 | | Libraries: libcurl/8.2.1-DEV Schannel zlib/1.2.13
08/18/2023 21:31:35 | | Data directory: D:\BOINC
08/18/2023 21:31:35 | | Running under account clhar
08/18/2023 21:31:36 | | CUDA: NVIDIA GPU 0: NVIDIA GeForce GTX 1070 (driver version 531.29, CUDA version 12.1, compute capability 6.1, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | CUDA: NVIDIA GPU 1: NVIDIA GeForce GTX 1070 (driver version 531.29, CUDA version 12.1, compute capability 6.1, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | OpenCL: NVIDIA GPU 0: NVIDIA GeForce GTX 1070 (driver version 531.29, device version OpenCL 3.0 CUDA, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | OpenCL: NVIDIA GPU 1: NVIDIA GeForce GTX 1070 (driver version 531.29, device version OpenCL 3.0 CUDA, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | OpenCL: Intel GPU 0 (ignored by config): Intel(R) UHD Graphics 630 (driver version 31.0.101.2125, device version OpenCL 3.0 NEO, 6396MB, 6396MB available, 230 GFLOPS peak)
08/18/2023 21:31:36 | | Windows processor group 0: 20 processors
08/18/2023 21:31:36 | | Host name: A-SYS
08/18/2023 21:31:36 | | Processor: 20 GenuineIntel Intel(R) Core(TM) i9-10900 CPU @ 2.80GHz [Family 6 Model 165 Stepping 5]
08/18/2023 21:31:36 | | OS: Microsoft Windows 11: Professional x64 Edition, (10.00.22621.00)
It seems that only one task is running by default. What is the sort name for these tasks, and how can I get both GPUs to work. Also how can I get the cuda tasks to run?
You say you have 2 different gpu's in the pc but it lists both of them as Nvidia 1070's is that correct? Also it says further down that the Intel gpu is "ignored by config" does that mean you already have the line to 'use_all_gpus' in your cc config file?
BTW, I didn't have this problem when I was running SETI or Milkyway with these same specs. Both project ran tasks on both GTX1070s and ignored the Intel GPU.
The Intel GPU is explicitly ignored by configuration <ignore_intel_dev>0</ignore_intel_dev>.
I don't specify <use_all_gpus>0|1</use_all_gpus> as it defaults to the most capable ones are used.
Might try the "1" anyway. If Boincmgr thinks one GPU is "better" than the other. It will only run that one GPU. If you have "1" set it will run an assortment of gpu's. Like rtx 3080 ti, gtx 1080 ti and gtx 1070 ti's (one of my boxs).
What happens if you pull the active gpu?
Respectfully,
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor)
Hallo! O3MD1/F are running
)
Hallo!
O3MD1/F are running out of work. What's next? Is there an immediate replacement in the pipeline?
Kind regards and happy crunching
Martin
We're working on the next GW
)
We're working on the next GW search, "O3AS1HF" (O3 data, All-Sky, High Frequency). There's still a change to be implemented and tested in the application. We hope to get the search started next week.
However, a few workunits (~100k) of the old O3MD1 search will need to be re-issued.
BM
Hallo! The last batch of
)
Hallo!
The last batch of data for O3MD1 I got was by far too big. I still have 131 tasks in spare, but do crunch only 11,5 tasks/day, and the date of expire is tomorrow about 15:55. So about 118 tasks will expire. Is it possible to shift the date of expire for the remaining tasks. I believe I'm not the only one.
Kind regards and happy crunching
Martin
astro-marwil
)
Tachchen, ;-)
No, deadlines are set serverside the second a tasks is assigned to a host which requested more work. The best you can do now: Estimate which tasks you will finish until their expiry date (and time of day). Abort all other O3MD1 tasks manually in the BOINC manager. This way they will be scheduled to other users ASAP. Einstein implemented some clever task locality scheme into BOINC's server software. Tasks aren't assigned randomly to users, but if possible to hosts which already downloaded the corresponding raw data files. The same set of raw data files is analyzed by many (thousands?) of different tasks applying different science parameters. Therefore, I assume there will be few additional load on download servers, due to aborting such O3MD1 tasks (different story for e.g. BRP4). This way your "lost tasks" will be drained through the einstein pipeline ASAP.
Important: Please check your account preferences (einstein website). Check "network usage" --> "store at least" [days] of work, "store up to an additional" [days] of work.
The estimated task duration logic for einstein tasks is not perfect. There are many threads discussing the reasons (quite complex) somewhere in the forums. So, if your work load changes, new science runs (like current O3MD1) starts, the estimated runtimes are often far from reality. [EDIT:] If they are way too short, your hosts requests way to much work. It isn't the einstein server assigning too much work to your computer, but your computer is requesting too much (see log of last server request at your host's page on einstein website).[/EDIT:] We (users) have to adapt. Einstein works best with very small task caches. It's not necessary to store many days as servers are very reliable, rarely offline for maintenance.
[EDIT:]
I haven't seen any of the current O3MD1 tasks. Are there CPU tasks or only GPU?Tech news here, not Einstein Café.Happy crunching.
Hallo SROOGE!Thank-you
)
Hallo SROOGE!
Thank-you for clarifying!
Kind regards and happy crunching
Martin
P.S.: Sag' bloß, Du bist auch in Bärlin zu Hause ?!
This old man needs some
)
This old man needs some help. Right now, I'm running twin Nvidia GTX 1070s and only one is running the gavitational-wave-search data.
08/18/2023 21:31:35 | | Starting BOINC client version 7.24.1 for windows_x86_64
08/18/2023 21:31:35 | | log flags: file_xfer, sched_ops, task, cpu_sched, sched_op_debug
08/18/2023 21:31:35 | | Libraries: libcurl/8.2.1-DEV Schannel zlib/1.2.13
08/18/2023 21:31:35 | | Data directory: D:\BOINC
08/18/2023 21:31:35 | | Running under account clhar
08/18/2023 21:31:36 | | CUDA: NVIDIA GPU 0: NVIDIA GeForce GTX 1070 (driver version 531.29, CUDA version 12.1, compute capability 6.1, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | CUDA: NVIDIA GPU 1: NVIDIA GeForce GTX 1070 (driver version 531.29, CUDA version 12.1, compute capability 6.1, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | OpenCL: NVIDIA GPU 0: NVIDIA GeForce GTX 1070 (driver version 531.29, device version OpenCL 3.0 CUDA, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | OpenCL: NVIDIA GPU 1: NVIDIA GeForce GTX 1070 (driver version 531.29, device version OpenCL 3.0 CUDA, 8192MB, 8192MB available, 6463 GFLOPS peak)
08/18/2023 21:31:36 | | OpenCL: Intel GPU 0 (ignored by config): Intel(R) UHD Graphics 630 (driver version 31.0.101.2125, device version OpenCL 3.0 NEO, 6396MB, 6396MB available, 230 GFLOPS peak)
08/18/2023 21:31:36 | | Windows processor group 0: 20 processors
08/18/2023 21:31:36 | | Host name: A-SYS
08/18/2023 21:31:36 | | Processor: 20 GenuineIntel Intel(R) Core(TM) i9-10900 CPU @ 2.80GHz [Family 6 Model 165 Stepping 5]
08/18/2023 21:31:36 | | OS: Microsoft Windows 11: Professional x64 Edition, (10.00.22621.00)
It seems that only one task is running by default. What is the sort name for these tasks, and how can I get both GPUs to work. Also how can I get the cuda tasks to run?
I don't buy computers, I build them!
Cliff Harding wrote: This
)
You say you have 2 different gpu's in the pc but it lists both of them as Nvidia 1070's is that correct? Also it says further down that the Intel gpu is "ignored by config" does that mean you already have the line to 'use_all_gpus' in your cc config file?
The Intel GPU is explicitly
)
The Intel GPU is explicitly ignored by configuration <ignore_intel_dev>0</ignore_intel_dev>.
I don't specify <use_all_gpus>0|1</use_all_gpus> as it defaults to the most capable ones are used.
I don't buy computers, I build them!
BTW, I didn't have this
)
BTW, I didn't have this problem when I was running SETI or Milkyway with these same specs. Both project ran tasks on both GTX1070s and ignored the Intel GPU.
I don't buy computers, I build them!
Cliff Harding wrote: The
)
Might try the "1" anyway. If Boincmgr thinks one GPU is "better" than the other. It will only run that one GPU. If you have "1" set it will run an assortment of gpu's. Like rtx 3080 ti, gtx 1080 ti and gtx 1070 ti's (one of my boxs).
What happens if you pull the active gpu?
Respectfully,
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor)