Posts by Keith Myers
|
21)
(Message 7984)
Posted 16 Sep 2023 by 
Keith Myers
Post: To use a aarch64 Pi on Universe you need to add the 32 bit libraries, i.e. , armhf, because all the Universe apps are 32bit for the cpu apps. Do this: sudo dpkg --add-architecture armhf sudo apt update sudo apt install libstdc++6:armhf libgomp1:armhf libboinc7:armhf sudo reboot |
|
22)
(Message 7867)
Posted 20 Aug 2023 by
Keith Myers
Post: Ah, my mistake. I should have recognized you posted the card detection output of an older Boinc client which has the inability to report no more than 4GB of memory for Nvidia cards because it uses older 32bit only detection. That flaw was remedied in the latest client versions. The 1.02 app is not at fault here. It is Boinc itself. And the flaw is only in reporting. Any app will use all of the actual card memory. I didn't independently verify your card VRAM capacity. Your 6GB of VRAM is more than enough to run 1X tasks with no issues. So the problem lies elsewhere. |
|
23)
(Message 7863)
Posted 18 Aug 2023 by
Keith Myers
Post: Yes, the work availability is very cyclical and predictable BG should input the task availability cycle frequency into their selection criteria since they appear to only start new Sprints when it coincides with some international sporting contest. Best to schedule sprints when projects have just refilled the RTS cache. In this current situation, with the Sprint starting at the bottom of the Asteroids work availability schedule, it should have been excluded from the pool of choices. |
|
24)
(Message 7858)
Posted 16 Aug 2023 by
Keith Myers
Post: Hi, You should get the 102 version app which handles CC between 3.0 and 8.0 Even though the CUDA55 app is still in the applications page it has been deprecated. Some tasks need more than 4GB but I have most tasks using around 2900 to 3800MB per task. The scheduler might look at your 4GB and just decide to not send you any on the chance it might send you one of the larger tasks needing more than 4GB. Just guessing here. |
|
25)
(Message 7855)
Posted 31 Jul 2023 by
Keith Myers
Post: Assume there is some pre-processing of the work units that the admins and devs have to do with the original data. Might not be anyone working the weekend to fill the pipeline in advance. Never more than a day for no work availability. If you can't stand having no work for a day, suggest you have 0 resource share backup projects or just set a bigger cache. |
|
26)
(Message 7853)
Posted 28 Jul 2023 by
Keith Myers
Post: You should blame BOINC for most of your issue. Boinc and the project software can't develop an accurate estimation of task progress or time remaining until you have returned 11 validated tasks for each application you run. As a new user, you may still have not reached that threshold. |
|
27)
(Message 7849)
Posted 17 Jul 2023 by
Keith Myers
Post: out of work again |
|
28)
(Message 7823)
Posted 19 Jun 2023 by
Keith Myers
Post: out of work again |
|
29)
(Message 7809)
Posted 24 Apr 2023 by
Keith Myers
Post: Pretty obvious. Don't run your C program debug at the same time as crunching Asteroids. |
|
30)
(Message 7797)
Posted 9 Apr 2023 by
Keith Myers
Post: You've relinquished host and project control to your account manager BAM. You will have to make changes there or remove the account manager and revert to local and project preferences control. |
|
31)
(Message 7773)
Posted 10 Mar 2023 by
Keith Myers
Post: The application name is just period_search. |
|
32)
(Message 7761)
Posted 8 Mar 2023 by
Keith Myers
Post: Oh, No! I see we are out of work again. Hope the admins follow this thread. |
|
33)
(Message 7759)
Posted 3 Mar 2023 by
Keith Myers
Post: I would guess not much of any difference. TN-Grid compiled 3 different apps from the same source code with just the change in SIMD instruction flags. They offer SSE2, AVX2 and FMA applications. The AVX2 and FMA applications are about equal but the SSE2 application loses out by about 10% in performance compared to the higher level SIMD instructions sets. Without specific changes in the code optimized for each SIMD class, I doubt there would be much difference between the AVX app and the proposed AVX2 app. |
|
34)
(Message 7731)
Posted 9 Feb 2023 by
Keith Myers
Post: Oh, No! I see we are out of work again. Hope the admins follow this thread. |
|
35)
(Message 7729)
Posted 8 Feb 2023 by
Keith Myers
Post: Is this just for errored tasks? Don't you get the board identified in the stderr.txt output for validated tasks? Like this: <core_client_version>7.19.0</core_client_version> <![CDATA[ <stderr_txt> BOINC client version 7.19.0 BOINC GPU type 'NVIDIA', deviceId=0, slot=14 CUDA version: 11080 CUDA Device number: 0 CUDA Device: NVIDIA GeForce RTX 3080 12037MB CUDA Device driver: 525.78.01 Compute capability: 8.6 Shared memory per Block | per SM: 49152 | 102400 Multiprocessors: 70 Resident blocks per multiprocessor: 16 Grid dim: 1120 = 70*16 Block dim: 128 11:17:52 (229427): called boinc_finish(0) </stderr_txt> ]] And this is what I get for an errored task: <core_client_version>7.19.0</core_client_version> <![CDATA[ <message> exceeded elapsed time limit 48606.68 (138002300.00G/2839.16G)</message> <stderr_txt> BOINC client version 7.19.0 BOINC GPU type 'NVIDIA', deviceId=1, slot=2 malloc(): invalid size (unsorted) SIGABRT: abort called BOINC client version 7.19.0 BOINC GPU type 'NVIDIA', deviceId=0, slot=2 malloc(): invalid size (unsorted) SIGABRT: abort called </stderr_txt> ]]> |
|
36)
(Message 7725)
Posted 29 Jan 2023 by
Keith Myers
Post: Oh, No! I see we are out of work again. Hope the admins follow this thread. |
|
37)
(Message 7722)
Posted 24 Jan 2023 by
Keith Myers
Post: Hi, asking for help from the project admins here. I am not able to add a co-admin to my team. All I get is a database error "unable to handle request" The user's email is validated in the database because it is the same address he uses to login to the project. Would you please investigate this please. I am posting here as I haven't received any reply from the PM sent to the admin yet. |
|
38)
(Message 7706)
Posted 15 Jan 2023 by
Keith Myers
Post: You are using the incorrect task limitation in your app_info or app_config. Project_max_concurrent applies to all tasks, both cpu and gpu in a project. If you want to limit the number of tasks being crunched for each device type, use the max_concurrent statement in each app_version section. Read the configuration options documentation. https://boinc.berkeley.edu/wiki/Client_configuration#Application_configuration |
|
39)
(Message 7705)
Posted 15 Jan 2023 by
Keith Myers
Post: BOINC CreditNew algorithm rewards credit based on the number of FLOPs involved in the calculation. NOT based on the time taken to perform those calculations. A task run on a cpu and the same task run on a gpu involve crunching through the exact same number of calculations so the credit rewarded is the same. Doesn't matter that the gpu doing parallel calculations can run through them much faster that the single thread run on the cpu. It is up to each project to define which credit algorithm to follow and install in their servers. https://boinc.berkeley.edu/trac/wiki/CreditOptions https://boinc.berkeley.edu/trac/wiki/CreditNew |
|
40)
(Message 7684)
Posted 10 Jan 2023 by
Keith Myers
Post: I see we are out of work again. Hope the admins can get some more stuffed into the work generators. |
Previous 20 · Next 20