Posts by Nick Name

1) Message boards : Number crunching : Radeon VII CL_OUT_OF_HOST_MEMORY Error (Message 297) Posted 19 Jul 2020 by Nick Name Post: I tried 20.4.2 and 20.q1., both for Win7, (Win10 versions will not install, the OS says the current driver is the latest even if no driver is installed) with the same result. Both were installed after using DDU to remove the current driver. MilkyWay work runs normally so everything seems ok otherwise. This particular system is unique and I don't see any need to pursue this further since no one else has reported a problem.
2) Questions and Answers : Web site : Friends Function Header Errors (Message 267) Posted 13 Jul 2020 by Nick Name Post: I get the following at the top of the page when sending invitations / confirming friends. Warning: Cannot modify header information - headers already sent by (output started at /home/boincadm/project/html/inc/util.inc:1095) in /home/boincadm/project/html/inc/util.inc on line 257 Warning: Cannot modify header information - headers already sent by (output started at /home/boincadm/project/html/inc/util.inc:1095) in /home/boincadm/project/html/inc/util.inc on line 258 Warning: Cannot modify header information - headers already sent by (output started at /home/boincadm/project/html/inc/util.inc:1095) in /home/boincadm/project/html/inc/util.inc on line 260 Warning: Cannot modify header information - headers already sent by (output started at /home/boincadm/project/html/inc/util.inc:1095) in /home/boincadm/project/html/inc/util.inc on line 262 Warning: Cannot modify header information - headers already sent by (output started at /home/boincadm/project/html/inc/util.inc:1095) in /home/boincadm/project/html/inc/util.inc on line 264 The friend functions work but it would be nice to have this cleaned up.
3) Message boards : Number crunching : Radeon VII CL_OUT_OF_HOST_MEMORY Error (Message 266) Posted 13 Jul 2020 by Nick Name Post: Thanks for the tips. I will try this on the weekend when I have time to fix things in case there's a disaster.
4) Message boards : Number crunching : Radeon VII CL_OUT_OF_HOST_MEMORY Error (Message 261) Posted 13 Jul 2020 by Nick Name Post: Thanks for the quick reply. It's a legitimate question. I'm using 19.6.3. Upgrading to Win10 isn't happening for awhile, and I can't upgrade the driver as AMD has blocked driver installation on Win8; I get a message saying the OS isn't supported. There's probably a kludge but this system is working perfectly otherwise so I'm not going to pursue it further. I expected this would be a general problem not specific to me. The VII is mostly used on the MilkyWay project, I only tried running it again to see if it worked since the app update here. If no one else is having this issue I wouldn't worry about it.
5) Message boards : Number crunching : Radeon VII CL_OUT_OF_HOST_MEMORY Error (Message 258) Posted 12 Jul 2020 by Nick Name Post: Seeing that there's a new app that fixed the multi-GPU issues, I thought I'd try again since the previous one didn't work well on my mixed Nvidia / AMD system. Nvidia assignment now works correctly but tasks are failing on my Radeon VII with this error: <core_client_version>7.14.2</core_client_version> <![CDATA[ <message> (unknown error) - exit code -6 (0xfffffffa)</message> <stderr_txt> Received work unit: 189087579814020 Data: n1: 923, n2: 411, n3: 907, di: 0, ch: 12 clCreateCommandQueueWithProperties : CL_OUT_OF_HOST_MEMORY I'm not sure if this error refers to VRAM or system RAM. This host has 32 GB of RAM and wasn't near using the swap so either way this seems like a bug.
6) Message boards : News : We are live (Message 99) Posted 29 Jun 2020 by Nick Name Post: (BTW a problem with the AMD project exclusion is that AMD work gets downloaded even though it won't run). So why not go to the Project preferences page then and uncheck the Use AMD GPU option? Then it won't use the AMD GPU and it won't download work for it. That would have been the smart move. I didn't think about it, and also it's not a problem using the device exclusion like I normally use. Curiously, BOINC gives an error message about a missing coproc using the project app exclusion, but still downloads work for it. Work for an ignored GPU isn't downloaded even if it's selected in project preferences. At least it allowed me to determine my AMD card isn't working here.
7) Message boards : News : We are live (Message 95) Posted 28 Jun 2020 by Nick Name Post: It is working for me ...snip... If it's still not working, can you post your cc_config.xml for us to look at? It didn't work, most likely to do with the AMD card. I run separate clients for each GPU plus another for the CPU because BOINC's scheduling annoys me. Normally I use the device exclusions. <ignore_ati_dev>0</ignore_ati_dev> <ignore_nvidia_dev>0</ignore_nvidia_dev> This didn't work so it makes sense the project exclusion doesn't work either. (BTW a problem with the AMD project exclusion is that AMD work gets downloaded even though it won't run). I think it's an AMD issue because removing all exclusions gets AMD work, but it's not actually running on the GPU. I have a similar problem with this system running Folding@Home work. I tried this with the project exclusion, device exclusion and both together. The result was always the same. Most aren't running Nvidia and AMD in the same system so I'll just wait and hope the app gets fixed. Thanks for your help.
8) Message boards : News : We are live (Message 62) Posted 27 Jun 2020 by Nick Name Post: Until the app gets fixed, for those with multiple GPU systems, add this to your cc_config.xml. That way you can free up the rest of your GPUs to run other projects. You will need to quit/restart BOINC for the changes to take effect.: <exclude_gpu> <url>project_URL</url> [<device_num>N</device_num>] [<type>NVIDIA\|ATI\|intel_gpu</type>] [<app>appname</app>] </exclude_gpu> It runs on GPU "0" of each type by default. So exclude GPUs 1+ for each type of additional GPUs you have in your machine. For example, Here are the entries I have on a Machine with three Nvidia GPUs: Does that actually work? It doesn't work here, both general device and project exclusions are ignored by the app. All GPUs are detected. 6/27/2020 1:24:06 PM CUDA: NVIDIA GPU 0: GeForce RTX 2080 Ti (driver version 445.75, CUDA version 11.0, compute capability 7.5, 4096MB, 3018MB available, 15276 GFLOPS peak) 6/27/2020 1:24:06 PM CUDA: NVIDIA GPU 1: GeForce RTX 2080 Ti (driver version 445.75, CUDA version 11.0, compute capability 7.5, 4096MB, 3018MB available, 15276 GFLOPS peak) 6/27/2020 1:24:06 PM OpenCL: NVIDIA GPU 0: GeForce RTX 2080 Ti (driver version 445.75, device version OpenCL 1.2 CUDA, 11264MB, 3018MB available, 15276 GFLOPS peak) 6/27/2020 1:24:06 PM OpenCL: NVIDIA GPU 1: GeForce RTX 2080 Ti (driver version 445.75, device version OpenCL 1.2 CUDA, 11264MB, 3018MB available, 15276 GFLOPS peak) 6/27/2020 1:24:06 PM OpenCL: AMD/ATI GPU 0: AMD Radeon VII (driver version 2841.19 (PAL,HSAIL), device version OpenCL 2.0 AMD-APP (2841.19), 16192MB, 16192MB available, 13839 GFLOPS peak) The project exclusion is correctly read by BOINC: minecrafthome 6/27/2020 1:24:06 PM Config: excluded GPU. Type: NVIDIA. App: kaktwoos. Device: 0 But the app is still running on device 0. I'm running BOINC 7.14.2. If it's working for other Nvidia users then the AMD card must be confusing things. I haven't had this problem on other projects though.
9) Message boards : News : We are live (Message 37) Posted 26 Jun 2020 by Nick Name Post: Until the app gets fixed, for those with multiple GPU systems, add this to your cc_config.xml. That way you can free up the rest of your GPUs to run other projects. You will need to quit/restart BOINC for the changes to take effect.: <exclude_gpu> <url>project_URL</url> [<device_num>N</device_num>] [<type>NVIDIA\|ATI\|intel_gpu</type>] [<app>appname</app>] </exclude_gpu> It runs on GPU "0" of each type by default. So exclude GPUs 1+ for each type of additional GPUs you have in your machine. For example, Here are the entries I have on a Machine with three Nvidia GPUs: Does that actually work? This should but doesn't. <ignore_nvidia_dev>0</ignore_nvidia_dev> This from init_data in the slot folder: <gpu_type>NVIDIA</gpu_type> <gpu_device_num>1</gpu_device_num> <gpu_opencl_dev_index>1</gpu_opencl_dev_index> <gpu_usage>1.000000</gpu_usage> <ncpus>0.997799</ncpus> seems to indicate it should be running on the 2nd GPU, but it isn't.