2 GPUs for one task?

\n studio-striking\n

Message boards : Number crunching : 2 GPUs for one task?
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 58
Credit: 70,255,584
RAC: 0
Message 8358 - Posted: 21 Oct 2022, 0:13:42 UTC

I see someone doing 2-Nvidia tasks, are there 2-AMD tasks? How similar do the cards have to be?
ID: 8358 · Rating: 0 · rate: Rate + / Rate - Report as offensive
.clair.

Send message
Joined: 2 Dec 15
Posts: 14
Credit: 422,065,673
RAC: 0
Message 8359 - Posted: 21 Oct 2022, 0:41:38 UTC - in response to Message 8358.  
Last modified: 21 Oct 2022, 1:15:36 UTC

Can you give a link to where you see this
As far as I know tasks never spread across GPU`s
[unless boinc haz a funky crashy moment] ;-)
edit..
Oh , the `teknikaly any topic` page on R@h
https://boinc.bakerlab.org/rosetta/forum_thread.php?id=6893&postid=107516
funky stuff
They should rename it "Problems and Technical Issues with BOINC"
ID: 8359 · Rating: 0 · rate: Rate + / Rate - Report as offensive
.clair.

Send message
Joined: 2 Dec 15
Posts: 14
Credit: 422,065,673
RAC: 0
Message 8360 - Posted: 21 Oct 2022, 1:11:18 UTC

I would like to see what actual use of CPU/GPU is per task in resource/task monitor or the like
We could aktualy have a conversation about Moo here on Moo but we would just be talking to ourselves :-)
and what fun would that be
Hellow Greg_BE I know you are out there somewhere
ID: 8360 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 58
Credit: 70,255,584
RAC: 0
Message 8361 - Posted: 21 Oct 2022, 7:48:15 UTC - in response to Message 8360.  

You could try your 730 and 1060 Nvidias together. I have no Nvidias, and currently no very close AMD models in the same machine, apart from a dual AMD card, which Boinc sees as two cards, but Moo refuses to recognise the second chip.
ID: 8361 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Greg_BE

Send message
Joined: 18 Oct 22
Posts: 5
Credit: 66,767,068
RAC: 84,677
Message 8362 - Posted: 22 Oct 2022, 8:33:53 UTC

you guys saw this all on my post in the other project.
I can import all that here if needed.
ID: 8362 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 58
Credit: 70,255,584
RAC: 0
Message 8363 - Posted: 22 Oct 2022, 9:02:50 UTC - in response to Message 8362.  

you guys saw this all on my post in the other project.
I can import all that here if needed.
Just what you're managing to do on Moo - that one task uses 2 CPU cores, 2 GPUs, they're maxed out, and what models they are. And anything else relevant.

All I wanted to know is if I could do the same with 2 AMDs. It may be Moo doesn't do AMD in dual mode. Or maybe AMD don't have that ability in the first place. Or maybe my cards are too different?
ID: 8363 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Greg_BE

Send message
Joined: 18 Oct 22
Posts: 5
Credit: 66,767,068
RAC: 84,677
Message 8364 - Posted: 22 Oct 2022, 17:39:05 UTC - in response to Message 8360.  
Last modified: 22 Oct 2022, 17:41:21 UTC

I would like to see what actual use of CPU/GPU is per task in resource/task monitor or the like
We could aktualy have a conversation about Moo here on Moo but we would just be talking to ourselves :-)
and what fun would that be
Hellow Greg_BE I know you are out there somewhere



Clair, all the data you want to look at was posted back over in the other project.
Have a look through all the Imgur links.

You have task manager text, task manager graph, screenshot of Boinctasks, MSI Afterburner graph, gees I don't remember what else, but it is all there.

BTW, aren't we talking to ourselves on the other project as well? Just the one post from the current PM of the quarter and that other guy. That's the only others.
ID: 8364 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Greg_BE

Send message
Joined: 18 Oct 22
Posts: 5
Credit: 66,767,068
RAC: 84,677
Message 8365 - Posted: 22 Oct 2022, 17:40:00 UTC - in response to Message 8363.  

you guys saw this all on my post in the other project.
I can import all that here if needed.
Just what you're managing to do on Moo - that one task uses 2 CPU cores, 2 GPUs, they're maxed out, and what models they are. And anything else relevant.

All I wanted to know is if I could do the same with 2 AMDs. It may be Moo doesn't do AMD in dual mode. Or maybe AMD don't have that ability in the first place. Or maybe my cards are too different?



That's beyond my knowledge level.
Pretty sure the web can tell you that info.
ID: 8365 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 58
Credit: 70,255,584
RAC: 0
Message 8366 - Posted: 22 Oct 2022, 17:44:47 UTC - in response to Message 8365.  

Pretty sure the web can tell you that info.
It's up to Moo's programming, which Is why I asked in here. Seems there are no Moo programmers/admins in here.
ID: 8366 · Rating: 0 · rate: Rate + / Rate - Report as offensive
.clair.

Send message
Joined: 2 Dec 15
Posts: 14
Credit: 422,065,673
RAC: 0
Message 8369 - Posted: 22 Oct 2022, 20:16:11 UTC

True , we can end up talking to arse elves :-)
ID: 8369 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 58
Credit: 70,255,584
RAC: 0
Message 8370 - Posted: 23 Oct 2022, 4:25:19 UTC - in response to Message 8369.  

True , we can end up talking to arse elves :-)
ROTFPMSL! I've never seen it written like that before.
ID: 8370 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Link
Avatar

Send message
Joined: 11 Feb 14
Posts: 117
Credit: 7,649,163
RAC: 1
Message 8371 - Posted: 23 Oct 2022, 17:43:21 UTC - in response to Message 8363.  
Last modified: 23 Oct 2022, 18:11:53 UTC

you guys saw this all on my post in the other project.
I can import all that here if needed.
Just what you're managing to do on Moo - that one task uses 2 CPU cores, 2 GPUs, they're maxed out, and what models they are. And anything else relevant.

Regardless of which of his Moo results I click on, I see this in std_err:

12:35:54 (17420): device: OpenCL: NVIDIA GPU 0: NVIDIA GeForce GTX 1080 (driver version 522.30, device version OpenCL 3.0 CUDA, 8192MB, 8192MB available, 9070 GFLOPS peak)
12:35:54 (17420): device: OpenCL: NVIDIA GPU 1 (not used): NVIDIA GeForce GTX 1050 Ti (driver version 522.30, device version OpenCL 3.0 CUDA, 4096MB, 4096MB available, 2274 GFLOPS peak)

I couldn't find any result which used the 1050 Ti. So like I posted in the Rosetta thread, it would be interesting to see how the GPU load looks like with just Moo running and with just FAH running.

If you check another computer with more than one Nvidia GPU (and Windows, Linux std_err does not have that info), for example this one, you will easily find both

18:29:10 (14364): device: OpenCL: NVIDIA GPU 0 (not used): NVIDIA GeForce GTX 1660 Ti (driver version 512.15, device version OpenCL 3.0 CUDA, 6144MB, 6144MB available, 5622 GFLOPS peak)
18:29:10 (14364): device: OpenCL: NVIDIA GPU 1: NVIDIA GeForce GTX 1080 (driver version 512.15, device version OpenCL 3.0 CUDA, 8192MB, 8192MB available, 9523 GFLOPS peak)
and
18:10:45 (4696): device: OpenCL: NVIDIA GPU 0: NVIDIA GeForce GTX 1660 Ti (driver version 512.15, device version OpenCL 3.0 CUDA, 6144MB, 6144MB available, 5622 GFLOPS peak)
18:10:45 (4696): device: OpenCL: NVIDIA GPU 1 (not used): NVIDIA GeForce GTX 1080 (driver version 512.15, device version OpenCL 3.0 CUDA, 8192MB, 8192MB available, 9523 GFLOPS peak)


My guess is Moo is running on GPU 0 and FAH on both (since the 1080 Moo runtimes are about double compared to the example computer, which has the same card as GPU 1). The 2xCPU and 2xGPU in BOINC are from app_config.xml, you can insert there any number of CPUs and GPUs, which BOINC should reserve for a task, even more than you actually have. The app will still use what it was designed for (i.e. 1 GPU).

EDIT: you see that actually in the last task manager screenshot he posted:
Moo is running on GPU 0 and uses 6.3% CPU (1 core), the two instances of FAH are using each one GPU and each one CPU core.
ID: 8371 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 58
Credit: 70,255,584
RAC: 0
Message 8372 - Posted: 24 Oct 2022, 2:59:59 UTC - in response to Message 8371.  
Last modified: 24 Oct 2022, 3:00:35 UTC

That makes sense, but why does his boinc manager show "2NV"? Did it try but fail? Can you find a computer with two identical Nvidias?
ID: 8372 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Link
Avatar

Send message
Joined: 11 Feb 14
Posts: 117
Credit: 7,649,163
RAC: 1
Message 8373 - Posted: 24 Oct 2022, 8:05:53 UTC - in response to Message 8372.  

That makes sense, but why does his boinc manager show "2NV"?
Like I said, probably that's what he has put in his app_config.xml, probably when trying to get Moo to run on both cards. But that should be done via cc_config.xml, with <use_all_gpus>1</use_all_gpus>. Just guessing, but I don't see anything else, that could cause this.


Can you find a computer with two identical Nvidias?
Not hard, just browse the top computers list and you get for example this one. Same here, some tasks run on GPU 0, some on GPU 1.
ID: 8373 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 58
Credit: 70,255,584
RAC: 0
Message 8374 - Posted: 24 Oct 2022, 8:22:20 UTC - in response to Message 8373.  
Last modified: 24 Oct 2022, 8:23:10 UTC

Can you find a computer with two identical Nvidias?
Not hard, just browse the top computers list and you get for example this one. Same here, some tasks run on GPU 0, some on GPU 1.
I didn't know where to look for such things. I can't see your 2nd link at the moment, the server isn't responding, although your first link is ok. Are you saying you can't get both GPUs to use Moo, or you get one task on each? I can't run Moo on two Tahiti AMD chips, which are on the same card. Boinc treats them as two seperate GPUs, and so does MSI Afterburner, but Moo doesn't notice the second one.
ID: 8374 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Link
Avatar

Send message
Joined: 11 Feb 14
Posts: 117
Credit: 7,649,163
RAC: 1
Message 8375 - Posted: 24 Oct 2022, 8:47:39 UTC - in response to Message 8374.  
Last modified: 24 Oct 2022, 8:49:54 UTC

I can't see your 2nd link at the moment, the server isn't responding, although your first link is ok.
Yeah, it was difficult to open it for me too, you need to wait like half minute before it actually starts to load, this computer has long task list.

Or check dnetc_r72_1666176892_12_768_0 and dnetc_r72_1666176803_12_768_0, that are currently the two first tasks from the list.


Are you saying you can't get both GPUs to use Moo, or you get one task on each?
You can get one task on each GPU.


I can't run Moo on two Tahiti AMD chips, which are on the same card. Boinc treats them as two seperate GPUs, and so does MSI Afterburner, but Moo doesn't notice the second one.
Post your start up messages from BOINC, in particular the part where BOINC detects the GPUs, your entire cc_config.xml and your Moo app_config.xml if you have one.
ID: 8375 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 58
Credit: 70,255,584
RAC: 0
Message 8376 - Posted: 24 Oct 2022, 9:50:55 UTC - in response to Message 8375.  

I don't have a Moo app config, here are the other two:


Glass

1			24-10-2022 06:13 AM	Starting BOINC client version 7.20.2 for windows_x86_64	
2			24-10-2022 06:13 AM	log flags: file_xfer, sched_ops, task	
3			24-10-2022 06:13 AM	Libraries: libcurl/7.84.0-DEV Schannel zlib/1.2.12	
4			24-10-2022 06:13 AM	Data directory: C:\ProgramData\BOINC	
5			24-10-2022 06:13 AM	Running under account peter	
6			24-10-2022 06:13 AM	OpenCL: AMD/ATI GPU 0: AMD Radeon HD 7900 Series (driver version 2841.19, device version OpenCL 1.2 AMD-APP (2841.19), 3072MB, 3072MB available, 3277 GFLOPS peak)	
7			24-10-2022 06:13 AM	OpenCL: AMD/ATI GPU 1: AMD Radeon HD 7900 Series (driver version 2841.19, device version OpenCL 1.2 AMD-APP (2841.19), 3072MB, 3072MB available, 3277 GFLOPS peak)	
8			24-10-2022 06:13 AM	OpenCL: Intel GPU 0: Intel(R) UHD Graphics 630 (driver version 26.20.100.7642, device version OpenCL 2.1 NEO, 19596MB, 19596MB available, 221 GFLOPS peak)	
9			24-10-2022 06:13 AM	OpenCL CPU: Intel(R) Core(TM) i5-8600K CPU @ 3.60GHz (OpenCL driver vendor: Intel(R) Corporation, driver version 7.6.0.0814, device version OpenCL 2.1 (Build 0))	
10			24-10-2022 06:13 AM	Windows processor group 0: 6 processors	
11			24-10-2022 06:13 AM	Host name: Glass	
12			24-10-2022 06:13 AM	Processor: 6 GenuineIntel Intel(R) Core(TM) i5-8600K CPU @ 3.60GHz [Family 6 Model 158 Stepping 10]	
13			24-10-2022 06:13 AM	Processor features: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss htt tm pni ssse3 fma cx16 sse4_1 sse4_2 movebe popcnt aes f16c rdrandsyscall nx lm avx avx2 vmx smx tm2 pbe fsgsbase bmi1 hle	
14			24-10-2022 06:13 AM	OS: Microsoft Windows 11: Professional x64 Edition, (10.00.22000.00)	
15			24-10-2022 06:13 AM	Memory: 47.84 GB physical, 54.84 GB virtual	
16			24-10-2022 06:13 AM	Disk: 237.77 GB total, 159.30 GB free	
17			24-10-2022 06:13 AM	Local time is UTC +1 hours	
18			24-10-2022 06:13 AM	No WSL found.	
19			24-10-2022 06:13 AM	VirtualBox version: 5.2.44	





<cc_config>
    <log_flags>
        <file_xfer>1</file_xfer>
        <sched_ops>1</sched_ops>
        <task>1</task>
        <app_msg_receive>0</app_msg_receive>
        <app_msg_send>0</app_msg_send>
        <async_file_debug>0</async_file_debug>
        <benchmark_debug>0</benchmark_debug>
        <checkpoint_debug>0</checkpoint_debug>
        <coproc_debug>0</coproc_debug>
        <cpu_sched>0</cpu_sched>
        <cpu_sched_debug>0</cpu_sched_debug>
        <cpu_sched_status>0</cpu_sched_status>
        <dcf_debug>0</dcf_debug>
        <disk_usage_debug>0</disk_usage_debug>
        <file_xfer_debug>0</file_xfer_debug>
        <gui_rpc_debug>0</gui_rpc_debug>
        <heartbeat_debug>0</heartbeat_debug>
        <http_debug>0</http_debug>
        <http_xfer_debug>0</http_xfer_debug>
        <idle_detection_debug>0</idle_detection_debug>
        <mem_usage_debug>0</mem_usage_debug>
        <network_status_debug>0</network_status_debug>
        <notice_debug>0</notice_debug>
        <poll_debug>0</poll_debug>
        <priority_debug>0</priority_debug>
        <proxy_debug>0</proxy_debug>
        <rr_simulation>0</rr_simulation>
        <rrsim_detail>0</rrsim_detail>
        <sched_op_debug>0</sched_op_debug>
        <scrsave_debug>0</scrsave_debug>
        <slot_debug>0</slot_debug>
        <state_debug>0</state_debug>
        <statefile_debug>0</statefile_debug>
        <suspend_debug>0</suspend_debug>
        <task_debug>0</task_debug>
        <time_debug>0</time_debug>
        <trickle_debug>0</trickle_debug>
        <unparsed_xml>0</unparsed_xml>
        <work_fetch_debug>0</work_fetch_debug>
    </log_flags>
    <options>
        <abort_jobs_on_exit>0</abort_jobs_on_exit>
        <allow_multiple_clients>1</allow_multiple_clients>
        <allow_remote_gui_rpc>1</allow_remote_gui_rpc>
        <client_version_check_url>https://boinc.berkeley.edu/download.php?xml=1</client_version_check_url>
        <client_new_version_text></client_new_version_text>
        <client_download_url>https://boinc.berkeley.edu/download.php</client_download_url>
        <disallow_attach>0</disallow_attach>
        <dont_check_file_sizes>0</dont_check_file_sizes>
        <dont_contact_ref_site>0</dont_contact_ref_site>
        <lower_client_priority>0</lower_client_priority>
        <dont_suspend_nci>0</dont_suspend_nci>
        <dont_use_vbox>0</dont_use_vbox>
    <exclude_gpu>
        <url>http://einstein.phys.uwm.edu/</url>
        <device_num>11</device_num>
    </exclude_gpu>
    <exclude_gpu>
        <url>http://einstein.phys.uwm.edu/</url>
        <device_num>12</device_num>
    </exclude_gpu>
    <exclude_gpu>
        <url>https://www.worldcommunitygrid.org/</url>
        <device_num>10</device_num>
    </exclude_gpu>
    <exclude_gpu>
        <url>https://www.worldcommunitygrid.org/</url>
        <device_num>11</device_num>
    </exclude_gpu>
        <exit_after_finish>0</exit_after_finish>
        <exit_before_start>0</exit_before_start>
        <exit_when_idle>0</exit_when_idle>
        <fetch_minimal_work>0</fetch_minimal_work>
        <fetch_on_update>0</fetch_on_update>
        <force_auth>default</force_auth>
        <http_1_0>0</http_1_0>
        <http_transfer_timeout>300</http_transfer_timeout>
        <http_transfer_timeout_bps>10</http_transfer_timeout_bps>
        <max_event_log_lines>2000</max_event_log_lines>
        <max_file_xfers>8</max_file_xfers>
        <max_file_xfers_per_project>8</max_file_xfers_per_project>
        <max_stderr_file_size>0</max_stderr_file_size>
        <max_stdout_file_size>0</max_stdout_file_size>
        <max_tasks_reported>0</max_tasks_reported>
        <ncpus>-1</ncpus>
        <network_test_url>https://www.google.com/</network_test_url>
        <no_alt_platform>0</no_alt_platform>
        <no_gpus>0</no_gpus>
        <no_info_fetch>0</no_info_fetch>
        <no_opencl>0</no_opencl>
        <no_priority_change>0</no_priority_change>
        <os_random_only>0</os_random_only>
        <process_priority>-1</process_priority>
        <process_priority_special>-1</process_priority_special>
<proxy_info>
    <socks_server_name></socks_server_name>
    <socks_server_port>80</socks_server_port>
    <http_server_name></http_server_name>
    <http_server_port>80</http_server_port>
    <socks5_user_name></socks5_user_name>
    <socks5_user_passwd></socks5_user_passwd>
    <http_user_name></http_user_name>
    <http_user_passwd></http_user_passwd>
    <no_proxy></no_proxy>
</proxy_info>
        <rec_half_life_days>10.000000</rec_half_life_days>
        <report_results_immediately>0</report_results_immediately>
        <run_apps_manually>0</run_apps_manually>
        <save_stats_days>30</save_stats_days>
        <skip_cpu_benchmarks>0</skip_cpu_benchmarks>
        <simple_gui_only>0</simple_gui_only>
        <start_delay>0.000000</start_delay>
        <stderr_head>0</stderr_head>
        <suppress_net_info>0</suppress_net_info>
        <unsigned_apps_ok>0</unsigned_apps_ok>
        <use_all_gpus>0</use_all_gpus>
        <use_certs>0</use_certs>
        <use_certs_only>0</use_certs_only>
        <vbox_window>0</vbox_window>
    </options>
</cc_config>
ID: 8376 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Link
Avatar

Send message
Joined: 11 Feb 14
Posts: 117
Credit: 7,649,163
RAC: 1
Message 8377 - Posted: 24 Oct 2022, 10:50:45 UTC - in response to Message 8376.  
Last modified: 24 Oct 2022, 11:04:10 UTC

<use_all_gpus>0</use_all_gpus> must be <use_all_gpus>1</use_all_gpus>

Restart BOINC after the change and it should use both GPUs.

Your <exclude_gpu> tags are wrong as well, you have only ATI device 0 and 1 and Intel device 0, I'm not even sure what you try to achive there. Ever had more than 10 GPUs in that system?
ID: 8377 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Mr P Hucker
Avatar

Send message
Joined: 5 Jul 11
Posts: 58
Credit: 70,255,584
RAC: 0
Message 8378 - Posted: 24 Oct 2022, 11:22:40 UTC - in response to Message 8377.  

I tried changing to <use_all_gpus>1</use_all_gpus>, although other projects have been using both GPUs ok. Moo still tries to use both GPUs, but I can see in MSI Afterburner only 1 GPU is used. I vaguely remember someone might have said both tasks end up running on the same GPU and it's a bug in Moo, and I also vaguely remember it might also apply to two seperate cards.

I've never had more than 5 GPUs on that machine. My exclude tags are meant to be wrong. It's to save me time. I used to forbid for example GPU 0 from a project, and when I want to disable that command, I just add the preceding 1 to make it 10, which it ignores since I don't have a GPU 10.
ID: 8378 · Rating: 0 · rate: Rate + / Rate - Report as offensive
Link
Avatar

Send message
Joined: 11 Feb 14
Posts: 117
Credit: 7,649,163
RAC: 1
Message 8379 - Posted: 24 Oct 2022, 11:29:08 UTC - in response to Message 8378.  
Last modified: 24 Oct 2022, 11:31:11 UTC

I vaguely remember
that eventually you might need to disable crossfire, so the card becomes really two separate cards. I remember from Milkyway some discussions about cards with 2 GPUs on them, maybe you should try to search there in the forums.
ID: 8379 · Rating: 0 · rate: Rate + / Rate - Report as offensive
1 · 2 · Next

Message boards : Number crunching : 2 GPUs for one task?


 
Copyright © 2011-2024 Moo! Wrapper Project