Message boards :
Number crunching :
Vega 64 Command Lines?
Message board moderation
Previous · 1 · 2
Author | Message |
---|---|
![]() ![]() ![]() Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 ![]() ![]() |
Thanks for chiming in Rick. I saw you were getting much better performance out of your Vega cards and I also thought somewhere back in the dark recesses of my memory that -cpu_lock was a viable option parameter for AMD/ATI cards. Seti@Home classic workunits:20,676 CPU time:74,226 hours ![]() ![]() A proud member of the OFA (Old Farts Association) |
![]() ![]() Send message Joined: 27 Jul 09 Posts: 83 Credit: 12,388,119 RAC: 140 ![]() ![]() |
I have 3 mb_cmdline txt files on this machine and its an AMD Ryzen 2700x yet the cmdline files have intel in the name... does that matter? Which txt file should the command lines go in on this pc? Sorry for all the questions lol ![]() ![]() All of this has happened before and will happen again -Battlestar Galactica |
![]() ![]() ![]() Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 ![]() ![]() |
Which app are you running? I assume you are running all three possibilities since the server is still trying to determine which app is fastest for you. It should land on the SoG app eventually. Why so few returned tasks? I see 200 in progress and no tasks validated, pending, inconclusive or errored. Are you keeping the machine offline until you receive configuration information. The apps do run fine without any further intervention on your part. To answer your question, put the command line into all three mb_cmdline files, the ati_sah, the ati5_sah and the ati_SoG files. That way no matter what app you run, you will have the command line getting the most out of the card. Seti@Home classic workunits:20,676 CPU time:74,226 hours ![]() ![]() A proud member of the OFA (Old Farts Association) |
![]() ![]() Send message Joined: 27 Jul 09 Posts: 83 Credit: 12,388,119 RAC: 140 ![]() ![]() |
Which app are you running? I assume you are running all three possibilities since the server is still trying to determine which app is fastest for you. It should land on the SoG app eventually. Just the standard BOINC app like my Dell server is running. Why so few returned tasks? I see 200 in progress and no tasks validated, pending, inconclusive or errored. Are you keeping the machine offline until you receive configuration information. The apps do run fine without any further intervention on your part.[/i] This is my brand new rig that I finished in January and I wanted to get the command line help before I start letting it crunch for longer periods of time but it wont be 24/7 like my Dell is... it will be working at random times when I fell like letting her have some play time haha! Also with all the RGB theres no way I could sleep with it on at night LOL! To answer your question, put the command line into all three mb_cmdline files, the ati_sah, the ati5_sah and the ati_SoG files. That way no matter what app you run, you will have the command line getting the most out of the card.[i] Thanks Keith!!! I will put that in all 3. Here is the pc with the Ryzen 2700X and the factory overclocked Sapphire Nitro+ Vega 64... that gpu is an absolute beast! ![]() ![]() ![]() ![]() ![]() All of this has happened before and will happen again -Battlestar Galactica |
![]() ![]() ![]() Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 ![]() ![]() |
BOINC is just the client which handles communications with projects. BOINC Manager is a nice gui to manage the client. Those programs having nothing to do with crunching work. Only the science apps crunch tasks. Since this is a new machine the schedulers have just seen and rarely, it doesn't know what to send you for the best science apps. So it tries all the ones at its disposal, both cpu and gpu. The cpu apps are fairly limited but the choices in gpu apps fill a page. https://setiathome.berkeley.edu/apps.php Once each gpu science app processes 11 valid, non-oveflowed, non- radar blanked tasks it can develop an APR (Average Processing Rate) for the app. Once all three of your ATI gpu app tryouts have developed an APR, the scheduler will know which app is the fastest and then will only send you work for that app. The app with the highest numerical APR is the winner. That app will be the SoG app since it is the current fastest app for Windows hosts. Seti@Home classic workunits:20,676 CPU time:74,226 hours ![]() ![]() A proud member of the OFA (Old Farts Association) |
![]() ![]() Send message Joined: 27 Jul 09 Posts: 83 Credit: 12,388,119 RAC: 140 ![]() ![]() |
For future reference and anyone needing a command line for your Vega 64 or Vega 56... the information here helped tremendously! After BOINC had time to adjust to the new command line (which took about 24hrs) new work units are coming in with an estimated 4:40'ish run time which far exceeds my expectation and is much better than the 12 mins I was getting. With that said results may vary depending on the brand/version of Vega card and if you are running stock gpu speeds (1247Mhz with 1546 Boost) or factory overclocked like my Sapphire Nitro+ (1630Mhz). Although I wouldn't recommend extreme overclocking since it can cause invalid results... just leave the card at where it was designed to run at and have fun! I have also been told you can undervolt the Vega cards to reduce your power usage but I have chosen no to since this machine wont crunch 24/7 like my Dell server does. Also you need to be feeding that beast of a gpu with 1 core of your cpu... to do that you need to edit your app_config.xml in your seti project folder with the following: <app_config> <app> <name>setiathome_v8</name> <gpu_versions> <gpu_usage>1</gpu_usage> <cpu_usage>1</cpu_usage> </gpu_versions> </app> </app_config> A huge thank you to everyone here that helped! ![]() ![]() All of this has happened before and will happen again -Battlestar Galactica |
TBar Send message Joined: 22 May 99 Posts: 5204 Credit: 840,779,836 RAC: 2,768 ![]() ![]() |
This still shows you are starving your GPU of the necessary CPU Resources. Looking above I see the CPU is still running 16 threads, forcing the GPU to compete with the CPU App for CPU time. Looking at your latest results it shows a Very small CPU time, Name: blc32_2bit_guppi_58406_04260_HIP116258_0041.20836.0.21.44.122.vlar_1 Run time: 9 min 58 sec CPU time: 51 sec Now compare that to the results from Rick: Name: blc33_2bit_guppi_58405_80305_HIP85620_0011.29743.0.21.44.147.vlar_0 Run time: 5 min 31 sec CPU time: 1 min 43 sec Notice how his GPU is using Much More CPU time even though the run-time is lower. Now compare the CPU times. Your's shows a large difference in Total Time and CPU time indicating the CPU is Over-committed causing the completion times to be slower; Name: blc32_2bit_guppi_58406_28274_HIP20533_0107.16731.0.21.44.211.vlar_0 Run time: 1 hours 30 min 46 sec CPU time: 1 hours 28 min 47 sec Ricks; Name: blc33_2bit_guppi_58405_81634_HIP85575_0015.29703.409.22.45.15.vlar_0 Run time: 1 hours 7 min 48 sec CPU time: 1 hours 7 min 34 sec Note how his times are a matter of seconds apart while Yours is Minutes apart, again, indicating Your CPU is Over-committed. It's very simple to fix this, lower your CPU tasks to 15 (or less) instead of 16, you will know you're close when the CPU times are less than 30 secs apart, as Rick's. You will see your GPU & CPU speed up, along with the CPU run-times narrow in difference. Notice the CPU Time on my AMD 570; Name: blc33_2bit_guppi_58405_80305_HIP85620_0011.23744.409.21.44.180.vlar_1 Run time: 7 min 35 sec CPU time: 2 min 52 sec Also the difference in CPU times; Name: blc33_2bit_guppi_58405_80634_HIP86036_0012.27883.818.21.44.254.vlar_0 Run time: 1 hours 38 min 33 sec CPU time: 1 hours 38 min 31 sec Obviously with only 2 secs between the above times, my CPU is Not Over-committed. |
![]() Send message Joined: 28 Nov 02 Posts: 5126 Credit: 276,046,078 RAC: 462 ![]() |
Which app are you running? I assume you are running all three possibilities since the server is still trying to determine which app is fastest for you. It should land on the SoG app eventually. I can see how you would be "delighted" with that machine! ;) Tom A proud member of the OFA (Old Farts Association). |
![]() ![]() Send message Joined: 27 Jul 09 Posts: 83 Credit: 12,388,119 RAC: 140 ![]() ![]() |
This still shows you are starving your GPU of the necessary CPU Resources. Looking above I see the CPU is still running 16 threads, forcing the GPU to compete with the CPU App for CPU time. Looking at your latest results it shows a Very small CPU time, Ok... so how can I fix it? ![]() All of this has happened before and will happen again -Battlestar Galactica |
![]() ![]() ![]() Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 ![]() ![]() |
https://setiathome.berkeley.edu/prefs.php?subset=global Instead of using 100% of your cpu cores, set it to 90%. Seti@Home classic workunits:20,676 CPU time:74,226 hours ![]() ![]() A proud member of the OFA (Old Farts Association) |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13915 Credit: 208,696,464 RAC: 304 ![]() ![]() |
https://setiathome.berkeley.edu/prefs.php?subset=global Or just reserve 1 core for each GPU WU that is being processed. In your Seti project directory. app_config.xml <app_config> <app> <name>setiathome_v8</name> <gpu_versions> <gpu_usage>1.0</gpu_usage> <cpu_usage>1.0</cpu_usage> </gpu_versions> </app> <app> <name>astropulse_v7</name> <gpu_versions> <gpu_usage>0.5</gpu_usage> <cpu_usage>1.0</cpu_usage> </gpu_versions> </app> </app_config> Works for me. Grant Darwin NT |
![]() Send message Joined: 28 Nov 02 Posts: 5126 Credit: 276,046,078 RAC: 462 ![]() |
Or use both. Set your local bionc manager to use 90% of the available cores running the cpu 100% of the time. And set your app_config.xml file to use 1 cpu / gpu. Tom A proud member of the OFA (Old Farts Association). |
![]() Send message Joined: 8 Mar 19 Posts: 14 Credit: 54,481,483 RAC: 150 ![]() ![]() |
Hello RueiKe, I was wondering if you might be able to help my with my command line as well. I am running a R9 380 & I plugged in the cml you provided for the vega. (-v 1 -instances_per_device 1 -sbs 2048 -period_iterations_num 1 -tt 600 -spike_fft_thresh 4096 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64 -hp -high_perf -no_defaults_scaling -cpu_lock -nobs -tune 1 64 1 4 -no_use_sleep) I wasn't sure if it is optimal for my ATI GPU.. Any help would be great, thx in advance. |
![]() ![]() ![]() Send message Joined: 17 Feb 01 Posts: 34497 Credit: 79,922,639 RAC: 80 ![]() ![]() |
I`m running a R9 380. Use this. -sbs 384 -instances_per_device 1 -period_iterations_num 30 -spike_fft_thresh 2048 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 32 -oclfft_tune_cw 32 -cpu_lock With each crime and every kindness we birth our future. |
![]() Send message Joined: 8 Mar 19 Posts: 14 Credit: 54,481,483 RAC: 150 ![]() ![]() |
Thanks Mike, will give this a try and see how it performs.. on my ATI, I am running a mixed system(Nvidia & ATI), When I was setting up I was having trouble getting SoG to run.. it is running on my Nviodia Cards now, but on my ATI i am running Sah(HD5).. is HD5 better than SoG, and can only one type of card run SoG at a time, because it seamed like whenever I was trying to get my ATI to run SoG, my Nvidia cards would drop out.. anyways, it was a pain getting them running together, multiple driver re-installs.. so little cautious with messing with it, but still wonder if I should be running Sog instead of HD5... Anywho.. thanks again for the cml, I will plug it in for the day and see how it does to the previous one I had snagged from the forums |
![]() ![]() ![]() Send message Joined: 17 Feb 01 Posts: 34497 Credit: 79,922,639 RAC: 80 ![]() ![]() |
You are running SOG. All is fine. With each crime and every kindness we birth our future. |
![]() Send message Joined: 8 Mar 19 Posts: 14 Credit: 54,481,483 RAC: 150 ![]() ![]() |
Hi Mike, I tried your command line, but the one that I had previously was actually working better.. Your cml took 20min compard to 12:53 with this one: -v 1 -instances_per_device 1 -sbs 2048 -period_iterations_num 1 -tt 600 -spike_fft_thresh 4096 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64 -hp -high_perf -no_defaults_scaling -cpu_lock -nobs -tune 1 64 1 4 -no_use_sleep image link below https://www.dropbox.com/s/az7pucndyz5gvwh/cmdln.JPG |
![]() ![]() ![]() Send message Joined: 17 Feb 01 Posts: 34497 Credit: 79,922,639 RAC: 80 ![]() ![]() |
When you check my results they take between 10 and 12 minutes with my values. If the others work better for you just let it run. With each crime and every kindness we birth our future. |
©2025 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.