GPU device assignment, 6.10.56 for x86_64-pc-linux-gnu

Message boards : Questions and problems : GPU device assignment, 6.10.56 for x86_64-pc-linux-gnu
Message board moderation

To post messages, you must log in.

AuthorMessage
Skip Da Shu
Avatar

Send message
Joined: 18 Mar 08
Posts: 38
United States
Message 33244 - Posted: 3 Jun 2010, 22:10:04 UTC

I have been noticing that on one machine the vidio cards are numbered in BOINC the opposite of the way nvidia-settings, gkrellm and nvclock number them. All of the above have the GTX-275 card as the 2nd card and it is physically in the 2nd PCIx slot. Now at one time they were reversed but I had to change them around to keep the 275 cool enough to run. Is there someplace in a config or .xml file that this possibly got set because of their earlier placement? Is there a way to correct it?

The problem seems specific to this machine because my desktop (this machine) does not seem to have this problem. This machine has two GTS-250s in it so it's a bit harder to determine but I believe the OS, device driver, and BOINCmgr all agree that the faster one is in slot1 and called GPU0 or whatever the particular software wants to call the first device. You'll see in this machines startup it lists the GFLOPS of GPU0 at 477 while GPU1 reports 470. The core clock on the GTS-250 in slot one is clocked a bit higher than the GTS-250 in slot two (771 vs 756). gkrellm, nvidia-settings and nvclock all refer to the faster card as the first device. So in this case BOINC seems to match all else.

Below is startup log of my 'c35' machine:

Thu 03 Jun 2010 03:52:30 PM CDT		Starting BOINC client version 6.10.56 for x86_64-pc-linux-gnu
Thu 03 Jun 2010 03:52:30 PM CDT		Config: use all coprocessors
Thu 03 Jun 2010 03:52:30 PM CDT		Config: GUI RPC allowed from any host
Thu 03 Jun 2010 03:52:30 PM CDT		Config: GUI RPC allowed from:
Thu 03 Jun 2010 03:52:30 PM CDT		log flags: file_xfer, sched_ops, task
Thu 03 Jun 2010 03:52:30 PM CDT		Libraries: libcurl/7.18.0 OpenSSL/0.9.8g zlib/1.2.3.3 c-ares/1.5.1
Thu 03 Jun 2010 03:52:30 PM CDT		Data directory: /var/lib/boinc-client
Thu 03 Jun 2010 03:52:31 PM CDT		Processor: 8 GenuineIntel Intel(R) Core(TM) i7 CPU         860  @ 2.80GHz [Family 6 Model 30 Stepping 5]
Thu 03 Jun 2010 03:52:31 PM CDT		Processor: 8.00 MB cache
Thu 03 Jun 2010 03:52:31 PM CDT		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 ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good xtopology tsc_reliable nonstop_tsc pni dtes64 moni
Thu 03 Jun 2010 03:52:31 PM CDT		OS: Linux: 2.6.31-21-generic
Thu 03 Jun 2010 03:52:31 PM CDT		Memory: 3.86 GB physical, 3.91 GB virtual
Thu 03 Jun 2010 03:52:31 PM CDT		Disk: 51.16 GB total, 43.65 GB free
Thu 03 Jun 2010 03:52:31 PM CDT		Local time is UTC -5 hours
Thu 03 Jun 2010 03:52:31 PM CDT		NVIDIA GPU 0: GeForce GTX 275 (driver version unknown, CUDA version 3000, compute capability 1.3, 896MB, 691 GFLOPS peak)
Thu 03 Jun 2010 03:52:31 PM CDT		NVIDIA GPU 1: GeForce GTS 250 (driver version unknown, CUDA version 3000, compute capability 1.1, 511MB, 470 GFLOPS peak)



And this is my 'c17' machine:

Thu 03 Jun 2010 09:43:56 AM CDT		Starting BOINC client version 6.10.56 for x86_64-pc-linux-gnu
Thu 03 Jun 2010 09:43:56 AM CDT		Config: use all coprocessors
Thu 03 Jun 2010 09:43:56 AM CDT		Config: GUI RPC allowed from any host
Thu 03 Jun 2010 09:43:56 AM CDT		Config: GUI RPC allowed from:
Thu 03 Jun 2010 09:43:56 AM CDT		Config:   192.168.218.17
Thu 03 Jun 2010 09:43:56 AM CDT		Config:   192.168.218.15
Thu 03 Jun 2010 09:43:56 AM CDT		log flags: file_xfer, sched_ops, task
Thu 03 Jun 2010 09:43:56 AM CDT		Libraries: libcurl/7.18.0 OpenSSL/0.9.8g zlib/1.2.3.3 c-ares/1.5.1
Thu 03 Jun 2010 09:43:56 AM CDT		Data directory: /var/lib/boinc-client
Thu 03 Jun 2010 09:43:56 AM CDT		Processor: 4 GenuineIntel Intel(R) Core(TM)2 Quad CPU    Q9550  @ 2.83GHz [Family 6 Model 23 Stepping 10]
Thu 03 Jun 2010 09:43:56 AM CDT		Processor: 6.00 MB cache
Thu 03 Jun 2010 09:43:56 AM CDT		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 ht tm pbe syscall nx lm constant_tsc arch_perfmon pebs bts rep_good pni dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr
Thu 03 Jun 2010 09:43:56 AM CDT		OS: Linux: 2.6.31-21-generic
Thu 03 Jun 2010 09:43:56 AM CDT		Memory: 3.87 GB physical, 2.00 GB virtual
Thu 03 Jun 2010 09:43:56 AM CDT		Disk: 16.97 GB total, 11.26 GB free
Thu 03 Jun 2010 09:43:56 AM CDT		Local time is UTC -5 hours
Thu 03 Jun 2010 09:43:56 AM CDT		NVIDIA GPU 0: GeForce GTS 250 (driver version unknown, CUDA version 3000, compute capability 1.1, 511MB, 477 GFLOPS peak)
Thu 03 Jun 2010 09:43:56 AM CDT		NVIDIA GPU 1: GeForce GTS 250 (driver version unknown, CUDA version 3000, compute capability 1.1, 512MB, 470 GFLOPS peak)




Here is the result of "nvclock -s" on c35 (the cruncher, GTX-275 runs the 648MHz clock):

nvclock -s
Unhandled init script entry with id '�' at c758
Card: 		Unknown Nvidia card
Card number: 	1
Mode		GPU Clock	Memory Clock
Coolbits 2D: 	300.000 MHz	100.000 MHz
Coolbits 3D: 	760.000 MHz	1150.000 MHz
Current: 	756.000 MHz	1152.000 MHz

Card: 		Unknown Nvidia card
Card number: 	2
Mode		GPU Clock	Memory Clock
Coolbits 2D: 	300.000 MHz	100.000 MHz
Coolbits 3D: 	760.000 MHz	1150.000 MHz
Current: 	648.000 MHz	1152.000 MHz


Here is the result of "nvclock -s" on c17 (the desktop):

nvclock -s
Unhandled init script entry with id '�' at c7d8
Card: 		Unknown Nvidia card
Card number: 	1
Mode		GPU Clock	Memory Clock
Coolbits 2D: 	300.000 MHz	100.000 MHz
Coolbits 3D: 	767.000 MHz	1100.000 MHz
Current: 	771.428 MHz	1101.600 MHz

Card: 		Unknown Nvidia card
Card number: 	2
Mode		GPU Clock	Memory Clock
Coolbits 2D: 	300.000 MHz	100.000 MHz
Coolbits 3D: 	767.000 MHz	1100.000 MHz
Current: 	756.000 MHz	1116.000 MHz


- da shu @ HeliOS,
"A child's exposure to technology should never be predicated on an ability to afford it."
ID: 33244 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15486
Netherlands
Message 33245 - Posted: 3 Jun 2010, 22:16:50 UTC - in response to Message 33244.  

The order is set by the way you insert them in your machine, the drivers you have installed and the library read by BOINC. I think you'll have to reinstall your drivers if you want to see them in the correct order as they are in the machine.

BOINC merely checks for the GPU library, sees if it exists on your system and then shows what - according to that library - is installed and in what order. It doesn't reorder anything.
ID: 33245 · Report as offensive
Skip Da Shu
Avatar

Send message
Joined: 18 Mar 08
Posts: 38
United States
Message 33246 - Posted: 3 Jun 2010, 23:19:15 UTC - in response to Message 33245.  

The order is set by the way you insert them in your machine, the drivers you have installed and the library read by BOINC. I think you'll have to reinstall your drivers if you want to see them in the correct order as they are in the machine.

BOINC merely checks for the GPU library, sees if it exists on your system and then shows what - according to that library - is installed and in what order. It doesn't reorder anything.


We're still missing a step someplace...

I reinstalled 195.36.24 drivers right before I posted the results. On that machine nvidia-settings reports GPU 0 as the GTS-250 and GPU 1 as the GTX-275 which corresponds to them being in slot 1 and slot 2.

Boinc still reports:

Thu 03 Jun 2010 05:24:39 PM CDT		NVIDIA GPU 0: GeForce GTX 275 (driver version unknown, CUDA version 3000, compute capability 1.3, 896MB, 691 GFLOPS peak)
Thu 03 Jun 2010 05:24:39 PM CDT		NVIDIA GPU 1: GeForce GTS 250 (driver version unknown, CUDA version 3000, compute capability 1.1, 511MB, 470 GFLOPS peak)


I am suspicious of some Xorg setting someplace but it's not in /etc/X11/xorg.conf


- da shu @ HeliOS,
"A child's exposure to technology should never be predicated on an ability to afford it."
ID: 33246 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15486
Netherlands
Message 33248 - Posted: 4 Jun 2010, 5:48:36 UTC - in response to Message 33246.  

Remind me, is there a way to uninstall drivers under Linux? For it may be that you have to get rid of them, before reinstalling that will change this. Some configuration file somewhere that sets the order, that the rest is getting their information from.
ID: 33248 · Report as offensive
Skip Da Shu
Avatar

Send message
Joined: 18 Mar 08
Posts: 38
United States
Message 33297 - Posted: 7 Jun 2010, 16:30:06 UTC - in response to Message 33248.  
Last modified: 7 Jun 2010, 16:31:19 UTC

Remind me, is there a way to uninstall drivers under Linux? For it may be that you have to get rid of them, before reinstalling that will change this. Some configuration file somewhere that sets the order, that the rest is getting their information from.


Yes, I can uninstall the drivers reboot and let boinc come up saying "GPU missing" stop it, stop the GUI and reinstall the driver. I'll give that a shot tonight. Good idea, thanx.

Skip

PS: There's always "a way" in Linux ;-) It's just a matter of do I know it... LOL
- da shu @ HeliOS,
"A child's exposure to technology should never be predicated on an ability to afford it."
ID: 33297 · Report as offensive

Message boards : Questions and problems : GPU device assignment, 6.10.56 for x86_64-pc-linux-gnu

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.