Fan control - Page 4

sinatosk · 12-28-08, 10:45 AM

part 2...

Quote:

==8086==
==8086== ERROR SUMMARY: 6 errors from 6 contexts (suppressed: 19 from 1)
==8086==
==8086== 1 errors in context 1 of 6:
==8086== Invalid write of size 1
==8086== at 0x80598A1: parse_bit_pll_table (bios.c:579)
==8086== by 0x8059F69: parse_bit_structure (bios.c:834)
==8086== by 0x805A718: parse_bios (bios.c:1093)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
==8086== Address 0x4039a9f is 11 bytes after a block of size 556 alloc'd
==8086== at 0x4004BA2: calloc (vg_replace_malloc.c:397)
==8086== by 0x805A6F0: parse_bios (bios.c:1091)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
==8086==
==8086== 1 errors in context 2 of 6:
==8086== Invalid write of size 1
==8086== at 0x8059878: parse_bit_pll_table (bios.c:578)
==8086== by 0x8059F69: parse_bit_structure (bios.c:834)
==8086== by 0x805A718: parse_bios (bios.c:1093)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
==8086== Address 0x4039a9e is 10 bytes after a block of size 556 alloc'd
==8086== at 0x4004BA2: calloc (vg_replace_malloc.c:397)
==8086== by 0x805A6F0: parse_bios (bios.c:1091)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
==8086==
==8086== 1 errors in context 3 of 6:
==8086== Invalid write of size 1
==8086== at 0x805984F: parse_bit_pll_table (bios.c:577)
==8086== by 0x8059F69: parse_bit_structure (bios.c:834)
==8086== by 0x805A718: parse_bios (bios.c:1093)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
==8086== Address 0x4039a9d is 9 bytes after a block of size 556 alloc'd
==8086== at 0x4004BA2: calloc (vg_replace_malloc.c:397)
==8086== by 0x805A6F0: parse_bios (bios.c:1091)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
==8086==
==8086== 1 errors in context 4 of 6:
==8086== Invalid write of size 1
==8086== at 0x8059826: parse_bit_pll_table (bios.c:576)
==8086== by 0x8059F69: parse_bit_structure (bios.c:834)
==8086== by 0x805A718: parse_bios (bios.c:1093)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
==8086== Address 0x4039a9c is 8 bytes after a block of size 556 alloc'd
==8086== at 0x4004BA2: calloc (vg_replace_malloc.c:397)
==8086== by 0x805A6F0: parse_bios (bios.c:1091)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
==8086==
==8086== 1 errors in context 5 of 6:
==8086== Invalid write of size 4
==8086== at 0x8059622: parse_bit_pll_table (bios.c:563)
==8086== by 0x8059F69: parse_bit_structure (bios.c:834)
==8086== by 0x805A718: parse_bios (bios.c:1093)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
==8086== Address 0x4039a98 is 4 bytes after a block of size 556 alloc'd
==8086== at 0x4004BA2: calloc (vg_replace_malloc.c:397)
==8086== by 0x805A6F0: parse_bios (bios.c:1091)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
==8086==
==8086== 1 errors in context 6 of 6:
==8086== Invalid write of size 4
==8086== at 0x80595D4: parse_bit_pll_table (bios.c:562)
==8086== by 0x8059F69: parse_bit_structure (bios.c:834)
==8086== by 0x805A718: parse_bios (bios.c:1093)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
==8086== Address 0x4039a94 is 0 bytes after a block of size 556 alloc'd
==8086== at 0x4004BA2: calloc (vg_replace_malloc.c:397)
==8086== by 0x805A6F0: parse_bios (bios.c:1091)
==8086== by 0x805A594: read_bios (bios.c:1054)
==8086== by 0x804DDD1: parse_config (config.c:471)
==8086== by 0x804D8C6: open_config (config.c:307)
==8086== by 0x8057A00: init_nvclock (back_linux.c:109)
==8086== by 0x804A0FD: main (nvclock.c:605)
--8086--
--8086-- supp: 19 dl-hack3-1
==8086==
==8086== IN SUMMARY: 6 errors from 6 contexts (suppressed: 19 from 1)
==8086==
==8086== malloc/free: in use at exit: 1,340 bytes in 26 blocks.
==8086== malloc/free: 154 allocs, 128 frees, 107,706 bytes allocated.
==8086==
==8086== searching for pointers to 26 not-freed blocks.
==8086== checked 98,640 bytes.
==8086==
==8086== LEAK SUMMARY:
==8086== definitely lost: 0 bytes in 0 blocks.
==8086== possibly lost: 0 bytes in 0 blocks.
==8086== still reachable: 1,340 bytes in 26 blocks.
==8086== suppressed: 0 bytes in 0 blocks.
==8086== Rerun with --leak-check=full to see details of leaked memory.
--8086-- memcheck: sanity checks: 8 cheap, 2 expensive
--8086-- memcheck: auxmaps: 0 auxmap entries (0k, 0M) in use
--8086-- memcheck: auxmaps_L1: 0 searches, 0 cmps, ratio 0:10
--8086-- memcheck: auxmaps_L2: 0 searches, 0 nodes
--8086-- memcheck: SMs: n_issued = 29 (464k, 0M)
--8086-- memcheck: SMs: n_deissued = 0 (0k, 0M)
--8086-- memcheck: SMs: max_noaccess = 65535 (1048560k, 1023M)
--8086-- memcheck: SMs: max_undefined = 0 (0k, 0M)
--8086-- memcheck: SMs: max_defined = 59 (944k, 0M)
--8086-- memcheck: SMs: max_non_DSM = 29 (464k, 0M)
--8086-- memcheck: max sec V bit nodes: 1 (0k, 0M)
--8086-- memcheck: set_sec_vbits8 calls: 2 (new: 1, updates: 1)
--8086-- memcheck: max shadow mem size: 768k, 0M
--8086-- translate: fast SP updates identified: 5,452 ( 87.3%)
--8086-- translate: generic_known SP updates identified: 475 ( 7.6%)
--8086-- translate: generic_unknown SP updates identified: 318 ( 5.0%)
--8086-- tt/tc: 11,611 tt lookups requiring 12,079 probes
--8086-- tt/tc: 11,611 fast-cache updates, 3 flushes
--8086-- transtab: new 5,345 (118,903 -> 1,647,738; ratio 138:10) [0 scs]
--8086-- transtab: dumped 0 (0 -> ??)
--8086-- transtab: discarded 6 (153 -> ??)
--8086-- scheduler: 875,055 jumps (bb entries).
--8086-- scheduler: 8/57,270 major/minor sched events.
--8086-- sanity: 9 cheap, 2 expensive checks.
--8086-- exectx: 769 lists, 179 contexts (avg 0 per list)
--8086-- exectx: 307 searches, 145 full compares (472 per 1000)
--8086-- exectx: 0 cmp2, 61 cmp4, 0 cmpAll
--8086-- errormgr: 14 supplist searches, 176 comparisons during search
--8086-- errormgr: 25 errlist searches, 109 comparisons during search

and thats all of it

**Thunderbird** · 12-28-08, 11:45 AM

Valgrind contains an x86 / x86-64 emulator and runs the app through that. This way it can intercept and solve memory issues. I'll study the report and add some fixes to nvclock. I'll report back later. Could you email me the file 'bios0.rom'? I think my bios parsing code is just failing on yours and that's what causes the issues.

**Thunderbird** · 12-28-08, 11:57 AM

I don't think I need the bios anymore. I found a bios on the web and did some tests. It appeared that there are more PLL tables than on other cards (or other systems were less picky about memory). It should work fine in cvs now. Could you try it?

sinatosk · 12-28-08, 11:47 PM

I'll give it a quick try now... but I have to go to work soon. I may be a while before I respond though...

sinatosk · 12-29-08, 06:28 PM

well... it seems to all work now. Some of the info that it showing seems to concern me and also incorrect

The nvidia powermizer is currently set at 0 so my gpu/shader/memory should be showing 300/600/100 but I'm seeing 300/1296/1188 yet when powermizer is 3 I should see 655/1404/1125 but I see 648/1404/1134. I'm not sure if this is a problem with your application.

I have a GTX 260 that has 216 stream units not 192. I can see why this is though... in your loop you loop it 8 times at max and you multiply it by 24 which ofcourse leads to the 192 value

my pci-e current rate is 2x instead of 16x... no idea why :s maybe thats why my rendering performance is lower in fedora 10. in Ubuntu 8.04.1 and 8.10 at a certain part ( radiosity I think it was ) I was getting 850 to 900 fps in lightsmark 2008... only 600 in Fedora 10. I don't think this is a problem with your application

the rpm rate are reporting much lower than whats windows ( gpu-z and everest ) is reporting

I will continue to keep Fedora 10 for a little while then I will try Ubuntu 8.10 out again with nvclock to see if it works

but thanks for all your help Thunderbird.

sinatosk · 12-29-08, 06:43 PM

oh forgot to add the output

This is powermizer 0

Quote:

-- General info --
Card: nvidia GeForce GTX 260
Architecture: GT200 A2
PCI id: 0x5e2
GPU clock: 300.856 MHz
Bustype: PCI-Express

-- Shader info --
Clock: 1296.000 MHz
Stream units: 192 (11111111b)
ROP units: 28 (01111111b)
-- Memory info --
Amount: 896 MB
Type: 448 bit DDR3
Clock: 1188.000 MHz

-- PCI-Express info --
Current Rate: 2X
Maximum rate: 16X

-- Sensor info --
Sensor: Analog Devices ADT7473
Board temperature: 34C
GPU temperature: 42C
Fanspeed: 327 RPM
Fanspeed mode: manual
PWM duty cycle: 40.0%

-- VideoBios information --
Version: 62.00.1a.00.16
Signon message: GT200 P651 SKU 0009 VGA BIOS
Performance level 0: gpu 300MHz/shader 600MHz/memory 100MHz/1.11V/100%
Performance level 1: gpu 400MHz/shader 800MHz/memory 300MHz/1.12V/100%
Performance level 2: gpu 655MHz/shader 1404MHz/memory 1125MHz/1.12V/100%
VID mask: 3
Voltage level 0: 1.06V, VID: 3
Voltage level 1: 1.11V, VID: 2
Voltage level 2: 1.12V, VID: 1
Voltage level 3: 1.15V, VID: 0

and this is powermizer 3

Quote:

-- General info --
Card: nvidia GeForce GTX 260
Architecture: GT200 A2
PCI id: 0x5e2
GPU clock: 648.000 MHz
Bustype: PCI-Express

-- Shader info --
Clock: 1404.000 MHz
Stream units: 192 (11111111b)
ROP units: 28 (01111111b)
-- Memory info --
Amount: 896 MB
Type: 448 bit DDR3
Clock: 1134.000 MHz

-- PCI-Express info --
Current Rate: 2X
Maximum rate: 16X

-- Sensor info --
Sensor: Analog Devices ADT7473
Board temperature: 35C
GPU temperature: 45C
Fanspeed: 327 RPM
Fanspeed mode: manual
PWM duty cycle: 40.0%

-- VideoBios information --
Version: 62.00.1a.00.16
Signon message: GT200 P651 SKU 0009 VGA BIOS
Performance level 0: gpu 300MHz/shader 600MHz/memory 100MHz/1.11V/100%
Performance level 1: gpu 400MHz/shader 800MHz/memory 300MHz/1.12V/100%
Performance level 2: gpu 655MHz/shader 1404MHz/memory 1125MHz/1.12V/100%
VID mask: 3
Voltage level 0: 1.06V, VID: 3
Voltage level 1: 1.11V, VID: 2
Voltage level 2: 1.12V, VID: 1
Voltage level 3: 1.15V, VID: 0

**Thunderbird** · 12-30-08, 03:40 AM

I have no idea to what degree the drivers are using the performance table for powermizer in the drivers. The last performance entry seems to be correct as you can't set all possible clock values the 1125 is the same as the 1134. I don't know why the other clock values don't match. To be honest I find 100MHz in one of them a bit too low, so either the bios is buggy or nvclock but I don't think it is nvclock as else I would have seen more issues.

**Thunderbird** · 12-30-08, 04:33 AM

The cvs version reports the pipelines correctly.

sinatosk · 12-30-08, 12:12 PM

hmmm... I still see it reporting 192 stream units instead of 216

**Thunderbird** · 12-30-08, 12:16 PM

Can you show me the line NV_1540 of the nvclock -D output?

sinatosk · 12-30-08, 12:30 PM

will do... just installing a game in windows right now

sinatosk · 12-30-08, 12:41 PM

NV_1540 (0x1540): f77f01ff

12-28-08, 11:45 AM	#38
Thunderbird Join Date: Jul 2002 Location: Netherlands, Europe Posts: 2,105	Re: Fan control Valgrind contains an x86 / x86-64 emulator and runs the app through that. This way it can intercept and solve memory issues. I'll study the report and add some fixes to nvclock. I'll report back later. Could you email me the file 'bios0.rom'? I think my bios parsing code is just failing on yours and that's what causes the issues.

12-28-08, 11:57 AM	#39
Thunderbird Join Date: Jul 2002 Location: Netherlands, Europe Posts: 2,105	Re: Fan control I don't think I need the bios anymore. I found a bios on the web and did some tests. It appeared that there are more PLL tables than on other cards (or other systems were less picky about memory). It should work fine in cvs now. Could you try it?

12-28-08, 11:47 PM	#40
sinatosk Registered User Join Date: May 2008 Posts: 32	Re: Fan control I'll give it a quick try now... but I have to go to work soon. I may be a while before I respond though...

12-29-08, 06:28 PM	#41
sinatosk Registered User Join Date: May 2008 Posts: 32	Re: Fan control well... it seems to all work now. Some of the info that it showing seems to concern me and also incorrect The nvidia powermizer is currently set at 0 so my gpu/shader/memory should be showing 300/600/100 but I'm seeing 300/1296/1188 yet when powermizer is 3 I should see 655/1404/1125 but I see 648/1404/1134. I'm not sure if this is a problem with your application. I have a GTX 260 that has 216 stream units not 192. I can see why this is though... in your loop you loop it 8 times at max and you multiply it by 24 which ofcourse leads to the 192 value my pci-e current rate is 2x instead of 16x... no idea why :s maybe thats why my rendering performance is lower in fedora 10. in Ubuntu 8.04.1 and 8.10 at a certain part ( radiosity I think it was ) I was getting 850 to 900 fps in lightsmark 2008... only 600 in Fedora 10. I don't think this is a problem with your application the rpm rate are reporting much lower than whats windows ( gpu-z and everest ) is reporting I will continue to keep Fedora 10 for a little while then I will try Ubuntu 8.10 out again with nvclock to see if it works but thanks for all your help Thunderbird.

12-30-08, 03:40 AM	#43
Thunderbird Join Date: Jul 2002 Location: Netherlands, Europe Posts: 2,105	Re: Fan control I have no idea to what degree the drivers are using the performance table for powermizer in the drivers. The last performance entry seems to be correct as you can't set all possible clock values the 1125 is the same as the 1134. I don't know why the other clock values don't match. To be honest I find 100MHz in one of them a bit too low, so either the bios is buggy or nvclock but I don't think it is nvclock as else I would have seen more issues.

12-30-08, 04:33 AM	#44
Thunderbird Join Date: Jul 2002 Location: Netherlands, Europe Posts: 2,105	Re: Fan control The cvs version reports the pipelines correctly.

12-30-08, 12:12 PM	#45
sinatosk Registered User Join Date: May 2008 Posts: 32	Re: Fan control hmmm... I still see it reporting 192 stream units instead of 216

12-30-08, 12:16 PM	#46
Thunderbird Join Date: Jul 2002 Location: Netherlands, Europe Posts: 2,105	Re: Fan control Can you show me the line NV_1540 of the nvclock -D output?

12-30-08, 12:30 PM	#47
sinatosk Registered User Join Date: May 2008 Posts: 32	Re: Fan control will do... just installing a game in windows right now

12-30-08, 12:41 PM	#48
sinatosk Registered User Join Date: May 2008 Posts: 32	Re: Fan control NV_1540 (0x1540): f77f01ff

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
GPU fan always at 100%, can not change the thermal settings	legluondunet	NVIDIA Linux	26	05-04-13 05:56 AM
9800 GTX and Fan Speed Control	HarryMan	NVIDIA Linux	0	05-25-12 11:50 PM
in 307.11 don't read Fan speed with nvidia-settings	sl1pkn07	NVIDIA Linux	2	05-21-12 10:08 AM
Jester2K Detonator 30.xx To 40.xx Control Panel Fix	sh*tshot	NVIDIA Windows Graphics Drivers	1	10-06-02 02:36 AM
Motherboard temperature control under Linux?	Klaus-P	NVIDIA Linux	2	09-02-02 02:53 AM