CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 820.

chrysophylax

legendary

Activity: 2940

Merit: 1091

--- ChainWorks Industries ---

Quote from: AzzAz on January 06, 2016, 06:41:49 AM

Quote from: chrysophylax on January 06, 2016, 06:27:15 AM

Quote from: AzzAz on January 06, 2016, 06:17:29 AM

Quote from: chrysophylax on January 06, 2016, 06:01:04 AM

hi all ...

can anyone help with this issue i have with the current build ...

im stumped as to why the 6th card on the new 'big' build of 980ti g1 and 980ti extreme refuses to work with the latest drivers ...

remove just ONE card from the lot - and it will run and mine ... and mine like a bat out of hell at that ...

the setup is exactly the same as the 750ti oc lp card setup - but these cards refuse to work with more than 5 cards ...

has anyone had an issue like this before? ... would it be an issue with onboard ram? ...

the system has 16gb high performance ram - h81probtc - g1840gpu - kingston v300 120gb sdd - 2 x corsair ax1200 psu ( 3 cards per psu - with the first psu running the power to the motherboard and ssd ) - usb3 risers ( all powered ) along with the 2 x 4pin molex power connectors ( as redundancy ) via the motherboard - 4 x gigabyte 980ti g1 cards + 2 x gigabyte 980ti extreme cards ( higher clocked ) for a total of 6 cards ...

when 5 cards of ANY combination are running - all is ok ... when the 6th card is installed - the boot refuses to load nvidia drivers ... the version of drivers plays no part in this as the drivers have been tested with different versions and they ALL do the same thing ...

fedora 23 x64 cuda 7.5 ( f23x64c75 ) and ccminer-spmod78 ...

the elusive 6th card is the only thing holding this monster of a miner back ... and is holding back the installation into its custom aluminium frame ...

any ideas? ... anyone? ..

#crysx

And if it is ProBTC mobo, did you unlock power @ BIOS? Usually it is power related, double check ( and clean) all connections

unlock power at bios? ... im not quite understanding what you mean ...

its not the connections ... as no matter which card you disconnect - it ALWAYS works when there is 5cards attached - and NEVER loads the nvidia driver when there is 6cards attached - no matter which combination of cards are plugged in to the pcie connectors ...

#crysx

I had that issue, and am also far away from my rigs so can't look at bioses, but I remember there was power limit settings (H81 ProBTC). It was MOBO powering max 4 or 5 cards for me, So I set BIOS, cleaned all connections and - it worked... ( 6x 960 cards, windows )

ok ... ill check the settings tomorrow afternoon when i get in ...

i dont remember any sort of power settings in the bios with this same motherboard on the 6 x gigabyte 750ti oc lp cards ... but they are not computer 5.2 devices ( compute 5.0 ) so i dunno if that has anything to do with it also ...

but i appreciate that info - so ill pour over the bios settings and see how that goes ...

tanx again ...

#crysx

ldw-com

full member

Activity: 229

Merit: 100

Quote from: chrysophylax on January 06, 2016, 06:27:15 AM

Quote from: AzzAz on January 06, 2016, 06:17:29 AM

Quote from: chrysophylax on January 06, 2016, 06:01:04 AM

hi all ...

can anyone help with this issue i have with the current build ...

im stumped as to why the 6th card on the new 'big' build of 980ti g1 and 980ti extreme refuses to work with the latest drivers ...

remove just ONE card from the lot - and it will run and mine ... and mine like a bat out of hell at that ...

the setup is exactly the same as the 750ti oc lp card setup - but these cards refuse to work with more than 5 cards ...

has anyone had an issue like this before? ... would it be an issue with onboard ram? ...

the system has 16gb high performance ram - h81probtc - g1840gpu - kingston v300 120gb sdd - 2 x corsair ax1200 psu ( 3 cards per psu - with the first psu running the power to the motherboard and ssd ) - usb3 risers ( all powered ) along with the 2 x 4pin molex power connectors ( as redundancy ) via the motherboard - 4 x gigabyte 980ti g1 cards + 2 x gigabyte 980ti extreme cards ( higher clocked ) for a total of 6 cards ...

when 5 cards of ANY combination are running - all is ok ... when the 6th card is installed - the boot refuses to load nvidia drivers ... the version of drivers plays no part in this as the drivers have been tested with different versions and they ALL do the same thing ...

fedora 23 x64 cuda 7.5 ( f23x64c75 ) and ccminer-spmod78 ...

the elusive 6th card is the only thing holding this monster of a miner back ... and is holding back the installation into its custom aluminium frame ...

any ideas? ... anyone? ..

#crysx

And if it is ProBTC mobo, did you unlock power @ BIOS? Usually it is power related, double check ( and clean) all connections

unlock power at bios? ... im not quite understanding what you mean ...

its not the connections ... as no matter which card you disconnect - it ALWAYS works when there is 5cards attached - and NEVER loads the nvidia driver when there is 6cards attached - no matter which combination of cards are plugged in to the pcie connectors ...

#crysx

Tried it on Windows? Just as a test?

AzzAz

legendary

Activity: 1030

Merit: 1006

Quote from: chrysophylax on January 06, 2016, 06:27:15 AM

Quote from: AzzAz on January 06, 2016, 06:17:29 AM

Quote from: chrysophylax on January 06, 2016, 06:01:04 AM

hi all ...

can anyone help with this issue i have with the current build ...

im stumped as to why the 6th card on the new 'big' build of 980ti g1 and 980ti extreme refuses to work with the latest drivers ...

remove just ONE card from the lot - and it will run and mine ... and mine like a bat out of hell at that ...

the setup is exactly the same as the 750ti oc lp card setup - but these cards refuse to work with more than 5 cards ...

has anyone had an issue like this before? ... would it be an issue with onboard ram? ...

the system has 16gb high performance ram - h81probtc - g1840gpu - kingston v300 120gb sdd - 2 x corsair ax1200 psu ( 3 cards per psu - with the first psu running the power to the motherboard and ssd ) - usb3 risers ( all powered ) along with the 2 x 4pin molex power connectors ( as redundancy ) via the motherboard - 4 x gigabyte 980ti g1 cards + 2 x gigabyte 980ti extreme cards ( higher clocked ) for a total of 6 cards ...

when 5 cards of ANY combination are running - all is ok ... when the 6th card is installed - the boot refuses to load nvidia drivers ... the version of drivers plays no part in this as the drivers have been tested with different versions and they ALL do the same thing ...

fedora 23 x64 cuda 7.5 ( f23x64c75 ) and ccminer-spmod78 ...

the elusive 6th card is the only thing holding this monster of a miner back ... and is holding back the installation into its custom aluminium frame ...

any ideas? ... anyone? ..

#crysx

And if it is ProBTC mobo, did you unlock power @ BIOS? Usually it is power related, double check ( and clean) all connections

unlock power at bios? ... im not quite understanding what you mean ...

its not the connections ... as no matter which card you disconnect - it ALWAYS works when there is 5cards attached - and NEVER loads the nvidia driver when there is 6cards attached - no matter which combination of cards are plugged in to the pcie connectors ...

#crysx

I had that issue, and am also far away from my rigs so can't look at bioses, but I remember there was power limit settings (H81 ProBTC). It was MOBO powering max 4 or 5 cards for me, So I set BIOS, cleaned all connections and - it worked... ( 6x 960 cards, windows )

chrysophylax

legendary

Activity: 2940

Merit: 1091

--- ChainWorks Industries ---

Quote from: AzzAz on January 06, 2016, 06:17:29 AM

Quote from: chrysophylax on January 06, 2016, 06:01:04 AM

hi all ...

can anyone help with this issue i have with the current build ...

im stumped as to why the 6th card on the new 'big' build of 980ti g1 and 980ti extreme refuses to work with the latest drivers ...

remove just ONE card from the lot - and it will run and mine ... and mine like a bat out of hell at that ...

the setup is exactly the same as the 750ti oc lp card setup - but these cards refuse to work with more than 5 cards ...

has anyone had an issue like this before? ... would it be an issue with onboard ram? ...

the system has 16gb high performance ram - h81probtc - g1840gpu - kingston v300 120gb sdd - 2 x corsair ax1200 psu ( 3 cards per psu - with the first psu running the power to the motherboard and ssd ) - usb3 risers ( all powered ) along with the 2 x 4pin molex power connectors ( as redundancy ) via the motherboard - 4 x gigabyte 980ti g1 cards + 2 x gigabyte 980ti extreme cards ( higher clocked ) for a total of 6 cards ...

when 5 cards of ANY combination are running - all is ok ... when the 6th card is installed - the boot refuses to load nvidia drivers ... the version of drivers plays no part in this as the drivers have been tested with different versions and they ALL do the same thing ...

fedora 23 x64 cuda 7.5 ( f23x64c75 ) and ccminer-spmod78 ...

the elusive 6th card is the only thing holding this monster of a miner back ... and is holding back the installation into its custom aluminium frame ...

any ideas? ... anyone? ..

#crysx

And if it is ProBTC mobo, did you unlock power @ BIOS? Usually it is power related, double check ( and clean) all connections

unlock power at bios? ... im not quite understanding what you mean ...

its not the connections ... as no matter which card you disconnect - it ALWAYS works when there is 5cards attached - and NEVER loads the nvidia driver when there is 6cards attached - no matter which combination of cards are plugged in to the pcie connectors ...

#crysx

chrysophylax

legendary

Activity: 2940

Merit: 1091

--- ChainWorks Industries ---

Quote from: sp_ on January 06, 2016, 06:09:57 AM

Try to replace a 980ti with a 750ti card. Does it work?
Do you have 8gig of ram in your rig?

the 6 x 750ti oc lp cards work in a factor of 6 - but they are installed ( with similar specs - apart from the psu ) as another miner ...

so in effect - no i havent tried them in THIS machine ... so that should really be the next step ...

the machine has 16gb high performance ram ( 2 x 8gb kingston hyperx 1600 ) - so it 'shouldnt' be the ram causing this ... but i have never built a machine this large as a miner ( just because i have never had the cash to - my mate does ) so anything could be the go with this ...

why do you think it could be lack of internal ram? ... i cant remember whether these boards support 4 x dimm slots on board ( as im not in the office until tomorrow afternoon ) ...

if so - ill buy another set of the 16gb kingston hyperx fury ram and see how it goes tomorrow ...

but let me tell you sp - just in the benchmark modes - these things are a hashers delight ... the extreme version is one hell of a hasher and is only $30AUD more than the g1 ... all stock - no oc ...

if its not the ram - then im STILL completely stumped ...

#crysx

AzzAz

legendary

Activity: 1030

Merit: 1006

Quote from: chrysophylax on January 06, 2016, 06:01:04 AM

hi all ...

can anyone help with this issue i have with the current build ...

im stumped as to why the 6th card on the new 'big' build of 980ti g1 and 980ti extreme refuses to work with the latest drivers ...

remove just ONE card from the lot - and it will run and mine ... and mine like a bat out of hell at that ...

the setup is exactly the same as the 750ti oc lp card setup - but these cards refuse to work with more than 5 cards ...

has anyone had an issue like this before? ... would it be an issue with onboard ram? ...

the system has 16gb high performance ram - h81probtc - g1840gpu - kingston v300 120gb sdd - 2 x corsair ax1200 psu ( 3 cards per psu - with the first psu running the power to the motherboard and ssd ) - usb3 risers ( all powered ) along with the 2 x 4pin molex power connectors ( as redundancy ) via the motherboard - 4 x gigabyte 980ti g1 cards + 2 x gigabyte 980ti extreme cards ( higher clocked ) for a total of 6 cards ...

when 5 cards of ANY combination are running - all is ok ... when the 6th card is installed - the boot refuses to load nvidia drivers ... the version of drivers plays no part in this as the drivers have been tested with different versions and they ALL do the same thing ...

fedora 23 x64 cuda 7.5 ( f23x64c75 ) and ccminer-spmod78 ...

the elusive 6th card is the only thing holding this monster of a miner back ... and is holding back the installation into its custom aluminium frame ...

any ideas? ... anyone? ..

#crysx

And if it is ProBTC mobo, did you unlock power @ BIOS? Usually it is power related, double check ( and clean) all connections

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Try to replace a 980ti with a 750ti card. Does it work?
Do you have 8gig of ram in your rig?

chrysophylax

legendary

Activity: 2940

Merit: 1091

--- ChainWorks Industries ---

hi all ...

can anyone help with this issue i have with the current build ...

im stumped as to why the 6th card on the new 'big' build of 980ti g1 and 980ti extreme refuses to work with the latest drivers ...

remove just ONE card from the lot - and it will run and mine ... and mine like a bat out of hell at that ...

the setup is exactly the same as the 750ti oc lp card setup - but these cards refuse to work with more than 5 cards ...

has anyone had an issue like this before? ... would it be an issue with onboard ram? ...

the system has 16gb high performance ram - h81probtc - g1840gpu - kingston v300 120gb sdd - 2 x corsair ax1200 psu ( 3 cards per psu - with the first psu running the power to the motherboard and ssd ) - usb3 risers ( all powered ) along with the 2 x 4pin molex power connectors ( as redundancy ) via the motherboard - 4 x gigabyte 980ti g1 cards + 2 x gigabyte 980ti extreme cards ( higher clocked ) for a total of 6 cards ...

when 5 cards of ANY combination are running - all is ok ... when the 6th card is installed - the boot refuses to load nvidia drivers ... the version of drivers plays no part in this as the drivers have been tested with different versions and they ALL do the same thing ...

fedora 23 x64 cuda 7.5 ( f23x64c75 ) and ccminer-spmod78 ...

the elusive 6th card is the only thing holding this monster of a miner back ... and is holding back the installation into its custom aluminium frame ...

any ideas? ... anyone? ..

#crysx

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

Quote from: induktor on January 05, 2016, 03:02:29 PM

Quote from: sp_ on January 05, 2016, 01:58:58 AM

On AMD you probobly undervolted your cards to get them to use less power. On Nvidia you change the TDP of the card. You can do it in software with the nvidia-smi tool, or modify the bios. (voids warranty)

yes I modded the bios myself, 0.8V, stable at 860Mhz, 1250 mem (sapphire 7950 dual X oc)
are you saying that i can run under linux, before loading ccminer, a nvidia-smi sentence that will change in real time the TDP of every card??? WOW!!! I thought that the only way was modifying the BIOS!, do you have any link / tuto or hint how to?, I will get right on it!.

To modify your bios to change the TDP you can follow these steps:

http://cryptomining-blog.com/1014-how-to-increase-the-geforce-gtx-750-ti-power-target-limit/

But a software solution is bether. Upgrade to the latest drivers.

Quote from: joblo on January 05, 2016, 07:30:11 PM

It works on Linux. You might also want to set persistence mode: nvidia-smi -pm 1 -i 0

sp_

legendary

Activity: 2954

Merit: 1087

Team Black developer

The cryptomining blog has compared release 74 with release 78 on a gtx 970 card x86 build(compute 5.2):

blakecoin +8%
lyra2 +6%
quark +3.8%
lyra2v2 +1.3%
x11 +0.2%

I have done alot of work to reach these speeds with cuda 7.5 ( 41 commits):

https://github.com/sp-hash/ccminer/commits/windows

But I am not done...

x13, x15, qubit and neoscrypt are still behind.

dominuspro

full member

Activity: 201

Merit: 100

Quote from: induktor on January 05, 2016, 09:05:19 PM

Doesn't work in any of my rigs, probably a driver issue:

Code:

root@room2:~# nvidia-smi -q -d POWER -i 0

==============NVSMI LOG==============

Timestamp : Wed Jan 6 03:03:10 2016
Driver Version : 346.59
Attached GPUs : 3
GPU 0000:01:00.0
Power Readings
Power Management : N/A
Power Draw : N/A
Power Limit : N/A
Default Power Limit : N/A
Enforced Power Limit : N/A
Min Power Limit : N/A
Max Power Limit : N/A
Power Samples
Duration : N/A
Number of Samples : N/A
Max : N/A
Min : N/A
Avg : N/A

and when I issue a change in any of the cards:

Code:

root@room2:~# nvidia-smi -pl 60 -i 1
Changing power management limit is not supported for GPU: 0000:02:00.0.
Treating as warning and moving on.
All done.

GPU are all 750Ti (gigabyte and EVGA FTW)

Probably a driver issue or maybe it has something in common with coolbits 28...

joblo

legendary

Activity: 1470

Merit: 1114

Quote from: induktor on January 05, 2016, 09:05:19 PM

Doesn't work in any of my rigs, probably a driver issue:

Code:

root@room2:~# nvidia-smi -q -d POWER -i 0

==============NVSMI LOG==============

Timestamp : Wed Jan 6 03:03:10 2016
Driver Version : 346.59
Attached GPUs : 3
GPU 0000:01:00.0
Power Readings
Power Management : N/A
Power Draw : N/A
Power Limit : N/A
Default Power Limit : N/A
Enforced Power Limit : N/A
Min Power Limit : N/A
Max Power Limit : N/A
Power Samples
Duration : N/A
Number of Samples : N/A
Max : N/A
Min : N/A
Avg : N/A

and when I issue a change in any of the cards:

Code:

root@room2:~# nvidia-smi -pl 60 -i 1
Changing power management limit is not supported for GPU: 0000:02:00.0.
Treating as warning and moving on.
All done.

GPU are all 750Ti (gigabyte and EVGA FTW)

Works on my 750ti with driver 352.63.

induktor

hero member

Activity: 710

Merit: 502

Doesn't work in any of my rigs, probably a driver issue:

Code:

root@room2:~# nvidia-smi -q -d POWER -i 0

==============NVSMI LOG==============

Timestamp : Wed Jan 6 03:03:10 2016
Driver Version : 346.59
Attached GPUs : 3
GPU 0000:01:00.0
Power Readings
Power Management : N/A
Power Draw : N/A
Power Limit : N/A
Default Power Limit : N/A
Enforced Power Limit : N/A
Min Power Limit : N/A
Max Power Limit : N/A
Power Samples
Duration : N/A
Number of Samples : N/A
Max : N/A
Min : N/A
Avg : N/A

and when I issue a change in any of the cards:

Code:

root@room2:~# nvidia-smi -pl 60 -i 1
Changing power management limit is not supported for GPU: 0000:02:00.0.
Treating as warning and moving on.
All done.

GPU are all 750Ti (gigabyte and EVGA FTW)

joblo

legendary

Activity: 1470

Merit: 1114

Quote from: dominuspro on January 05, 2016, 04:20:43 PM

Quote from: induktor on January 05, 2016, 03:02:29 PM

are you saying that i can run under linux, before loading ccminer, a nvidia-smi sentence that will change in real time the TDP of every card??? WOW!!! I thought that the only way was modifying the BIOS!, do you have any link / tuto or hint how to?, I will get right on it!.

I just tried in windows and it works. I'm pretty sure it should work also in linux but cannot test it because I don't have a ready linux install anymore. Setting the core and mem clocks(+changing to p0) works exactly the same way on linux and windows. Tested on 960, 970 and 980.

Those are the commands:
nvidia-smi -q -d POWER -i 0 -shows the actual setting and possible settings limits

nvidia-smi -pl 180 -i 0 -sets the power limit of gpu 0 to 180W

The successfull result:

Code:

Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. All rights reserved.

C:\Windows\system32>cd /

C:\>cd "Program Files"

C:\Program Files>cd "NVIDIA Corporation"

C:\Program Files\NVIDIA Corporation>cd NVSMI

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -q -d POWER -i 0

==============NVSMI LOG==============

Timestamp : Tue Jan 05 22:05:19 2016
Driver Version : 359.06

Attached GPUs : 5
GPU 0000:01:00.0
   Power Readings
   Power Management : Supported
   Power Draw : 13.98 W
   Power Limit : 196.15 W
   Default Power Limit : 163.46 W
   Enforced Power Limit : 196.15 W
   Min Power Limit : 100.00 W
   Max Power Limit : 196.15 W
   Power Samples
   Duration : N/A
   Number of Samples : N/A
   Max : N/A
   Min : N/A
   Avg : N/A

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -pl 180 -i 0
Power limit for GPU 0000:01:00.0 was set to 180.00 W from 196.15 W.
All done.

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -q -d POWER -i 0

==============NVSMI LOG==============

Timestamp : Tue Jan 05 22:05:40 2016
Driver Version : 359.06

Attached GPUs : 5
GPU 0000:01:00.0
   Power Readings
   Power Management : Supported
   Power Draw : 13.74 W
   Power Limit : 180.00 W
   Default Power Limit : 163.46 W
   Enforced Power Limit : 180.00 W
   Min Power Limit : 100.00 W
   Max Power Limit : 196.15 W
   Power Samples
   Duration : N/A
   Number of Samples : N/A
   Max : N/A
   Min : N/A
   Avg : N/A

C:\Program Files\NVIDIA Corporation\NVSMI>

It works on Linux. You might also want to set persistence mode: nvidia-smi -pm 1 -i 0

chrysophylax

legendary

Activity: 2940

Merit: 1091

--- ChainWorks Industries ---

Quote from: tbearhere on January 05, 2016, 06:29:32 PM

crysx
Are you talking about the 6 980ti's on one rig?

yup ...

so far only 5 x gigabyte 980ti g1 / extreme will work at any one time - and the 6th kills the drivers for some reason ...

the system is 4 x gigabyte 980ti g1 + 2 x gigabyte 980ti extreme ...

it doesnt matter which pcie you remove from the collection - it will only work with 5 cards - and the as soon as the 6th card goes in the driver refuses to load ...

#crysx

tbearhere

legendary

Activity: 3164

Merit: 1003

crysx
Are you talking about the 6 980ti's on one rig?

chrysophylax

legendary

Activity: 2940

Merit: 1091

--- ChainWorks Industries ---

Quote from: tbearhere on January 05, 2016, 05:59:54 PM

Quote from: dominuspro on January 05, 2016, 04:20:43 PM

Quote from: induktor on January 05, 2016, 03:02:29 PM

are you saying that i can run under linux, before loading ccminer, a nvidia-smi sentence that will change in real time the TDP of every card??? WOW!!! I thought that the only way was modifying the BIOS!, do you have any link / tuto or hint how to?, I will get right on it!.

I just tried in windows and it works. I'm pretty sure it should work also in linux but cannot test it because I don't have a ready linux install anymore. Setting the core and mem clocks(+changing to p0) works exactly the same way on linux and windows. Tested on 960, 970 and 980.

Those are the commands:
nvidia-smi -q -d POWER -i 0 -shows the actual setting and possible settings limits

nvidia-smi -pl 180 -i 0 -sets the power limit of gpu 0 to 180W

The successfull result:

Code:

Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. All rights reserved.

C:\Windows\system32>cd /

C:\>cd "Program Files"

C:\Program Files>cd "NVIDIA Corporation"

C:\Program Files\NVIDIA Corporation>cd NVSMI

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -q -d POWER -i 0

==============NVSMI LOG==============

Timestamp : Tue Jan 05 22:05:19 2016
Driver Version : 359.06

Attached GPUs : 5
GPU 0000:01:00.0
   Power Readings
   Power Management : Supported
   Power Draw : 13.98 W
   Power Limit : 196.15 W
   Default Power Limit : 163.46 W
   Enforced Power Limit : 196.15 W
   Min Power Limit : 100.00 W
   Max Power Limit : 196.15 W
   Power Samples
   Duration : N/A
   Number of Samples : N/A
   Max : N/A
   Min : N/A
   Avg : N/A

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -pl 180 -i 0
Power limit for GPU 0000:01:00.0 was set to 180.00 W from 196.15 W.
All done.

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -q -d POWER -i 0

==============NVSMI LOG==============

Timestamp : Tue Jan 05 22:05:40 2016
Driver Version : 359.06

Attached GPUs : 5
GPU 0000:01:00.0
   Power Readings
   Power Management : Supported
   Power Draw : 13.74 W
   Power Limit : 180.00 W
   Default Power Limit : 163.46 W
   Enforced Power Limit : 180.00 W
   Min Power Limit : 100.00 W
   Max Power Limit : 196.15 W
   Power Samples
   Duration : N/A
   Number of Samples : N/A
   Max : N/A
   Min : N/A
   Avg : N/A

C:\Program Files\NVIDIA Corporation\NVSMI>

Wow thx
Did anyone try this on a 980ti ?

Ps Cryptsy is off the market now.

almost finished my mates machine - having issues with the last card and the nvidia drivers accepting it - but can test this on his system the way it is while i still have it ...

will do that later this afternoon / tonight if i can get the chance - unless someone beats me to it ... which is more likely the case ...

#crysx

tbearhere

legendary

Activity: 3164

Merit: 1003

Quote from: dominuspro on January 05, 2016, 04:20:43 PM

Quote from: induktor on January 05, 2016, 03:02:29 PM

are you saying that i can run under linux, before loading ccminer, a nvidia-smi sentence that will change in real time the TDP of every card??? WOW!!! I thought that the only way was modifying the BIOS!, do you have any link / tuto or hint how to?, I will get right on it!.

I just tried in windows and it works. I'm pretty sure it should work also in linux but cannot test it because I don't have a ready linux install anymore. Setting the core and mem clocks(+changing to p0) works exactly the same way on linux and windows. Tested on 960, 970 and 980.

Those are the commands:
nvidia-smi -q -d POWER -i 0 -shows the actual setting and possible settings limits

nvidia-smi -pl 180 -i 0 -sets the power limit of gpu 0 to 180W

The successfull result:

Code:

Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. All rights reserved.

C:\Windows\system32>cd /

C:\>cd "Program Files"

C:\Program Files>cd "NVIDIA Corporation"

C:\Program Files\NVIDIA Corporation>cd NVSMI

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -q -d POWER -i 0

==============NVSMI LOG==============

Timestamp : Tue Jan 05 22:05:19 2016
Driver Version : 359.06

Attached GPUs : 5
GPU 0000:01:00.0
   Power Readings
   Power Management : Supported
   Power Draw : 13.98 W
   Power Limit : 196.15 W
   Default Power Limit : 163.46 W
   Enforced Power Limit : 196.15 W
   Min Power Limit : 100.00 W
   Max Power Limit : 196.15 W
   Power Samples
   Duration : N/A
   Number of Samples : N/A
   Max : N/A
   Min : N/A
   Avg : N/A

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -pl 180 -i 0
Power limit for GPU 0000:01:00.0 was set to 180.00 W from 196.15 W.
All done.

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -q -d POWER -i 0

==============NVSMI LOG==============

Timestamp : Tue Jan 05 22:05:40 2016
Driver Version : 359.06

Attached GPUs : 5
GPU 0000:01:00.0
   Power Readings
   Power Management : Supported
   Power Draw : 13.74 W
   Power Limit : 180.00 W
   Default Power Limit : 163.46 W
   Enforced Power Limit : 180.00 W
   Min Power Limit : 100.00 W
   Max Power Limit : 196.15 W
   Power Samples
   Duration : N/A
   Number of Samples : N/A
   Max : N/A
   Min : N/A
   Avg : N/A

C:\Program Files\NVIDIA Corporation\NVSMI>

Wow thx
Did anyone try this on a 980ti ?

Ps Cryptsy is off the market now.

dominuspro

full member

Activity: 201

Merit: 100

Quote from: induktor on January 05, 2016, 03:02:29 PM

are you saying that i can run under linux, before loading ccminer, a nvidia-smi sentence that will change in real time the TDP of every card??? WOW!!! I thought that the only way was modifying the BIOS!, do you have any link / tuto or hint how to?, I will get right on it!.

I just tried in windows and it works. I'm pretty sure it should work also in linux but cannot test it because I don't have a ready linux install anymore. Setting the core and mem clocks(+changing to p0) works exactly the same way on linux and windows. Tested on 960, 970 and 980.

Those are the commands:
nvidia-smi -q -d POWER -i 0 -shows the actual setting and possible settings limits

nvidia-smi -pl 180 -i 0 -sets the power limit of gpu 0 to 180W

The successfull result:

Code:

Microsoft Windows [Version 6.1.7601]
Copyright (c) 2009 Microsoft Corporation. All rights reserved.

C:\Windows\system32>cd /

C:\>cd "Program Files"

C:\Program Files>cd "NVIDIA Corporation"

C:\Program Files\NVIDIA Corporation>cd NVSMI

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -q -d POWER -i 0

==============NVSMI LOG==============

Timestamp : Tue Jan 05 22:05:19 2016
Driver Version : 359.06

Attached GPUs : 5
GPU 0000:01:00.0
   Power Readings
   Power Management : Supported
   Power Draw : 13.98 W
   Power Limit : 196.15 W
   Default Power Limit : 163.46 W
   Enforced Power Limit : 196.15 W
   Min Power Limit : 100.00 W
   Max Power Limit : 196.15 W
   Power Samples
   Duration : N/A
   Number of Samples : N/A
   Max : N/A
   Min : N/A
   Avg : N/A

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -pl 180 -i 0
Power limit for GPU 0000:01:00.0 was set to 180.00 W from 196.15 W.
All done.

C:\Program Files\NVIDIA Corporation\NVSMI>nvidia-smi -q -d POWER -i 0

==============NVSMI LOG==============

Timestamp : Tue Jan 05 22:05:40 2016
Driver Version : 359.06

Attached GPUs : 5
GPU 0000:01:00.0
   Power Readings
   Power Management : Supported
   Power Draw : 13.74 W
   Power Limit : 180.00 W
   Default Power Limit : 163.46 W
   Enforced Power Limit : 180.00 W
   Min Power Limit : 100.00 W
   Max Power Limit : 196.15 W
   Power Samples
   Duration : N/A
   Number of Samples : N/A
   Max : N/A
   Min : N/A
   Avg : N/A

C:\Program Files\NVIDIA Corporation\NVSMI>

bathrobehero

legendary

Activity: 2002

Merit: 1051

ICO? Not even once.

Quote from: induktor on January 05, 2016, 03:02:29 PM

Quote from: sp_ on January 05, 2016, 01:58:58 AM

On AMD you probobly undervolted your cards to get them to use less power. On Nvidia you change the TDP of the card. You can do it in software with the nvidia-smi tool, or modify the bios. (voids warranty)

yes I modded the bios myself, 0.8V, stable at 860Mhz, 1250 mem (sapphire 7950 dual X oc)
are you saying that i can run under linux, before loading ccminer, a nvidia-smi sentence that will change in real time the TDP of every card??? WOW!!! I thought that the only way was modifying the BIOS!, do you have any link / tuto or hint how to?, I will get right on it!.
I decide not to modify the BIOS because if I fuck something up, there is no backup, in AMD 7950 there is a dual bios switch, if you fuck up your card (which I did more than 100 times when researching this) just move the switch to position 2, power cycle, move switch to 1, reflash, power cycle and good to go!, but nvidia has no switch!.
GPU's in my country are EXTREMELY COSTLY, and there is no such a thing like warranty here, if you fuck it up, it's gone, one card to the bin basket. Sad

You can recover basically any screwed up BIOS in DOS with nvflash, you just need to plug your monitor into another GPU (onboard).

You can even shut down your computer midway flashing a card and still be able to reflash it next boot.

Quote from: induktor on January 05, 2016, 03:21:38 PM

bathrobehero what is your regular ambient temperature?, I am in a very hot area, that could explain why my cards where working at 75C and yours at 60

I had 12CM spacing between them, and a 3000 rpm fan between every card to prevent that one card heat up the other one, still, in quark specially, it was impossible to lower the temperature below 70C, in lyra2v2 yes, usually where at 60 probably, don't remember exactly.

true WD Green dies fast! for offline storage they are ok, but not for much else, even if you modify the parking time (i usually do, to 30 or 60 seconds), I always use black or red drives which has better warranty and lasts longer too (one of the few companies that has true warranty in my country, WD, and is excellent!)

still mining kills a lot of hard drives, until i switched to usb flash, I killed about 10 or 11 wd black, reds, and seagate drives beyond repair, keep in mind that it is the mix of 24/7 operation, a LOT of power outages (here power is extremely unstable, we have short brownouts at least 10 per day) and at least 1 power outage a week, 1 of 4 may last days.
(reason why I spend a ridiculous amount of money in inverter generators, smart UPS, solar panels, power regulators, inverters, I even have a low voltage line across the apartment for the essentials (battery powered)) add to all that, high temperatures and high humidity and you get....HELL !! hehe Grin

About the SSD, mining OS does not write a lot, just moving the logs and temps to ramdisk is enough, a crappy kingston V300 will last at least a couple of years easy.

of course someone mention that the 840 Pro still works, of course, it is one of the best SSD consumer drives in the market! I have one in my workstation and is fantastic and very fast!, but at almost twice the cost of the kingston, it worth it for your workstation but not for a miner IMHO.

I had an excellent experience booting of flash USB drive, almost all of my stability and disk problems went away since i boot off USB flash drive, so I will keep doing it

hehe. (I know i am a stubborn sob

)

Sorry, I meant to write 70°C, not 60 for the max temp for the 970 OC cards (I have 60°C temp limit on the 750 Ti's). The ambient temp now in the room where the rigs are is 21°C and the hottest 970 OC card is 62°C (80-85% fan). If I remove the temp target on the minis, they climb to 74-77°C.
I have no idea about the humidity.

Power here also isn't the best but it's nowhere near as bad as yours so I guess that's killing your harddrives. The two WD greens I have and used to mine burst for over a year (stopped about 6 months ago) have no isses so far and are at 800 load cycle (5 min parking time). I don't store anything important on them though.

Yes, the 840 Pro is great but mostly because it's fast but it still uses TLC NANDs and the crappy V300's which I have in all mining rigs have MLC NANDs which should be way more durable.

I always had terrible experience with USB drives though, maybe I just used really crappy ones.

Just to be slightly on topic: ccMiner SPMOD Release CUDA 6.5 vs CUDA 7.5 Performance Comparison

Topic: CCminer(SP-MOD) Modded NVIDIA Maxwell / Pascal kernels. - page 820. (Read 2347664 times)