Author

Topic: EWBF's CUDA Zcash miner - page 194. (Read 2164329 times)

hero member
Activity: 526
Merit: 502
March 05, 2017, 03:05:38 PM
Is anyone using EWBF on linux and if so can you tell me how you are OC the cards?  I am using the nvidia-settings and I am setting core, mem and fan speed but I feel like I am missing something else.

nope, that's how you OC.

In MSI afterburner I have a PowerLimit and TempLimit, do those correlate to nvidia-settings.  I have searched and looked at many articles and cannot find anything.  Just wandering if setting the fan speed in linux is how temp limit is handled in MSI afterburner

TempLimit is the temp you allow the card to reach before shutting down.
You adjust the fan curve to maintain the temp below your TempLimit. Ideally 55c to 65c.

right now I have

nvidia-settings -a '[gpu:N]/GPUGraphicsClockOffset[3]=150'

nvidia-settings -a '[gpu:N]/GPUMemoryTransferRateOffset[3]=525'

nvidia-settings -a "[gpu:N]/GPUFanControlState=1"

nvidia-settings -a '[fan-N]/GPUTargetFanSpeed=70'

Are there any other settings to control temp?

Enable Persistence Mode
Code:
sudo nvidia-smi -pm 1 

Power limit to 90w
Code:
sudo nvidia-smi -pl 90 

Code:
watch nvidia-smi

Thanks a billion, that was what I was missing.  With the persistence, does it stay set through a reboot, or do I need to add those 2 commands to my startup script?

you need root privilege open after.local file
Code:
sudo vi /etc/init.d/after.local

and add this lines
Code:
/usr/bin/nvidia-smi -pm 1
/usr/bin/nvidia-smi -pl 90



full member
Activity: 154
Merit: 100
March 05, 2017, 02:27:06 PM
Is anyone using EWBF on linux and if so can you tell me how you are OC the cards?  I am using the nvidia-settings and I am setting core, mem and fan speed but I feel like I am missing something else.

nope, that's how you OC.

In MSI afterburner I have a PowerLimit and TempLimit, do those correlate to nvidia-settings.  I have searched and looked at many articles and cannot find anything.  Just wandering if setting the fan speed in linux is how temp limit is handled in MSI afterburner

TempLimit is the temp you allow the card to reach before shutting down.
You adjust the fan curve to maintain the temp below your TempLimit. Ideally 55c to 65c.

right now I have

nvidia-settings -a '[gpu:N]/GPUGraphicsClockOffset[3]=150'

nvidia-settings -a '[gpu:N]/GPUMemoryTransferRateOffset[3]=525'

nvidia-settings -a "[gpu:N]/GPUFanControlState=1"

nvidia-settings -a '[fan-N]/GPUTargetFanSpeed=70'

Are there any other settings to control temp?

Enable Persistence Mode
Code:
sudo nvidia-smi -pm 1 

Power limit to 90w
Code:
sudo nvidia-smi -pl 90 

Code:
watch nvidia-smi

Thanks a billion, that was what I was missing.  With the persistence, does it stay set through a reboot, or do I need to add those 2 commands to my startup script?
sr. member
Activity: 420
Merit: 250
March 05, 2017, 01:18:01 PM

These percentages tell nothing because each card has totally different TDP settings and curve. What is the consumption at the wall? When i set the settings as you have, my cards take over 160Watt. So for me, it is still much better to go 410-420 sols at 105-110W.

this.

TDP and gpuz mean NOTHING, the only correct stat is how much power computer/rig sucks at the wall.
I've had gpuz say I'm using 60% tdp, and the wall said it's sucking 150+ watts, so...



Yes that is the case. Some brands have higher usage at 55% TDP than others at 80% TDP. This is especially true for the MSI and Palit cards. But you can be sure that no 1070 can hold clock over 2050 MHZ under 150 watts.

I saw this effect between the EVGA cards and a PNY

EVGA would be seemingly reliable for the software end of power readings. Also before saying no GTX 1070 can clock over 2050mhz under 150 watts, 2050mhz isn't a power bench mark, some cards hit higher clocks and one that reaches 2120mhz has headroom to reduce the power and stay over 2050.
hero member
Activity: 526
Merit: 502
March 05, 2017, 12:44:17 PM
Is anyone using EWBF on linux and if so can you tell me how you are OC the cards?  I am using the nvidia-settings and I am setting core, mem and fan speed but I feel like I am missing something else.

nope, that's how you OC.

In MSI afterburner I have a PowerLimit and TempLimit, do those correlate to nvidia-settings.  I have searched and looked at many articles and cannot find anything.  Just wandering if setting the fan speed in linux is how temp limit is handled in MSI afterburner

TempLimit is the temp you allow the card to reach before shutting down.
You adjust the fan curve to maintain the temp below your TempLimit. Ideally 55c to 65c.

right now I have

nvidia-settings -a '[gpu:N]/GPUGraphicsClockOffset[3]=150'

nvidia-settings -a '[gpu:N]/GPUMemoryTransferRateOffset[3]=525'

nvidia-settings -a "[gpu:N]/GPUFanControlState=1"

nvidia-settings -a '[fan-N]/GPUTargetFanSpeed=70'

Are there any other settings to control temp?

Enable Persistence Mode
Code:
sudo nvidia-smi -pm 1 

Power limit to 90w
Code:
sudo nvidia-smi -pl 90 

Code:
watch nvidia-smi
full member
Activity: 154
Merit: 100
March 05, 2017, 12:30:41 PM
Is anyone using EWBF on linux and if so can you tell me how you are OC the cards?  I am using the nvidia-settings and I am setting core, mem and fan speed but I feel like I am missing something else.

nope, that's how you OC.

In MSI afterburner I have a PowerLimit and TempLimit, do those correlate to nvidia-settings.  I have searched and looked at many articles and cannot find anything.  Just wandering if setting the fan speed in linux is how temp limit is handled in MSI afterburner

TempLimit is the temp you allow the card to reach before shutting down.
You adjust the fan curve to maintain the temp below your TempLimit. Ideally 55c to 65c.

right now I have

nvidia-settings -a '[gpu:N]/GPUGraphicsClockOffset[3]=150'

nvidia-settings -a '[gpu:N]/GPUMemoryTransferRateOffset[3]=525'

nvidia-settings -a "[gpu:N]/GPUFanControlState=1"

nvidia-settings -a '[fan-N]/GPUTargetFanSpeed=70'

Are there any other settings to control temp?
full member
Activity: 240
Merit: 100
March 05, 2017, 12:25:16 PM
Is anyone using EWBF on linux and if so can you tell me how you are OC the cards?  I am using the nvidia-settings and I am setting core, mem and fan speed but I feel like I am missing something else.

nope, that's how you OC.

In MSI afterburner I have a PowerLimit and TempLimit, do those correlate to nvidia-settings.  I have searched and looked at many articles and cannot find anything.  Just wandering if setting the fan speed in linux is how temp limit is handled in MSI afterburner

TempLimit is the temp you allow the card to reach before shutting down.
You adjust the fan curve to maintain the temp below your TempLimit. Ideally 55c to 65c.
full member
Activity: 154
Merit: 100
March 05, 2017, 12:04:06 PM
Is anyone using EWBF on linux and if so can you tell me how you are OC the cards?  I am using the nvidia-settings and I am setting core, mem and fan speed but I feel like I am missing something else.

nope, that's how you OC.

In MSI afterburner I have a PowerLimit and TempLimit, do those correlate to nvidia-settings.  I have searched and looked at many articles and cannot find anything.  Just wandering if setting the fan speed in linux is how temp limit is handled in MSI afterburner
legendary
Activity: 1068
Merit: 1020
March 05, 2017, 11:52:19 AM
Is anyone using EWBF on linux and if so can you tell me how you are OC the cards?  I am using the nvidia-settings and I am setting core, mem and fan speed but I feel like I am missing something else.

nope, that's how you OC.
full member
Activity: 154
Merit: 100
March 05, 2017, 10:09:30 AM
Is anyone using EWBF on linux and if so can you tell me how you are OC the cards?  I am using the nvidia-settings and I am setting core, mem and fan speed but I feel like I am missing something else.
sr. member
Activity: 337
Merit: 250
March 05, 2017, 09:36:11 AM

Yes that is the case. Some brands have higher usage at 55% TDP than others at 80% TDP. This is especially true for the MSI and Palit cards. But you can be sure that no 1070 can hold clock over 2050 MHZ under 150 watts.

Exactly. I found a sweet spot when underclocking it, IMHO, OC is just not worth it. 10% stronger hash rate demands around 25% more energy...
Only if one had almost free electricity, which no one really does.
sr. member
Activity: 463
Merit: 250
March 05, 2017, 09:09:13 AM

These percentages tell nothing because each card has totally different TDP settings and curve. What is the consumption at the wall? When i set the settings as you have, my cards take over 160Watt. So for me, it is still much better to go 410-420 sols at 105-110W.

this.

TDP and gpuz mean NOTHING, the only correct stat is how much power computer/rig sucks at the wall.
I've had gpuz say I'm using 60% tdp, and the wall said it's sucking 150+ watts, so...



Yes that is the case. Some brands have higher usage at 55% TDP than others at 80% TDP. This is especially true for the MSI and Palit cards. But you can be sure that no 1070 can hold clock over 2050 MHZ under 150 watts.
sr. member
Activity: 337
Merit: 250
March 05, 2017, 08:43:01 AM

These percentages tell nothing because each card has totally different TDP settings and curve. What is the consumption at the wall? When i set the settings as you have, my cards take over 160Watt. So for me, it is still much better to go 410-420 sols at 105-110W.

this.

TDP and gpuz mean NOTHING, the only correct stat is how much power computer/rig sucks at the wall.
I've had gpuz say I'm using 60% tdp, and the wall said it's sucking 150+ watts, so...

hero member
Activity: 710
Merit: 502
March 05, 2017, 08:36:53 AM
Hello
I tested version 031b on Linux (Kopiemtu 2.0) drivers 367.44 and 378, with 1070 G1 and 1060 G1 cards at minimum TDP and there is no difference in hashrate, except for 2 or 3 sols less than version 030b on the 1070 G1.

I will stay with the 030b for now.
indkt
full member
Activity: 224
Merit: 100
CryptoLearner
March 05, 2017, 05:45:04 AM
I never pushed memory clock that much, I'll give it a try,
but guys how much power are your cards sucking at this extreme OC?

I optimised it now to around 410 sols, and the whole system with 2 cards is sucking 370W, which is a great success (at extreme overclock, it used to suck up to 500W for me).
I can reach a stable 2100 core clock, but it just sucks way too much power...


It draws 85% from the 1070, the slower one. The faster one with Samsung I can go +700 memory if I wanted instead of 575

The core clock OC has no difference in power draw if you are talking about the GTX 1070. The core clock is just a clock offset while under the same power draw it would be pushed to either way. Memory OC's will push more power into the card

These percentages tell nothing because each card has totally different TDP settings and curve. What is the consumption at the wall? When i set the settings as you have, my cards take over 160Watt. So for me, it is still much better to go 410-420 sols at 105-110W.


I can only estimate it as 138w at the wall for the card based on calculating the TDP % and factoring the PSU efficiency


You can use nvidia smi to get accurate power usage (not counting the system or the psu efficiency)
sr. member
Activity: 420
Merit: 250
March 05, 2017, 05:04:57 AM
I never pushed memory clock that much, I'll give it a try,
but guys how much power are your cards sucking at this extreme OC?

I optimised it now to around 410 sols, and the whole system with 2 cards is sucking 370W, which is a great success (at extreme overclock, it used to suck up to 500W for me).
I can reach a stable 2100 core clock, but it just sucks way too much power...


It draws 85% from the 1070, the slower one. The faster one with Samsung I can go +700 memory if I wanted instead of 575

The core clock OC has no difference in power draw if you are talking about the GTX 1070. The core clock is just a clock offset while under the same power draw it would be pushed to either way. Memory OC's will push more power into the card

These percentages tell nothing because each card has totally different TDP settings and curve. What is the consumption at the wall? When i set the settings as you have, my cards take over 160Watt. So for me, it is still much better to go 410-420 sols at 105-110W.

I can only estimate it as 138w at the wall for the card based on calculating the TDP % and factoring the PSU efficiency
sr. member
Activity: 463
Merit: 250
March 05, 2017, 04:51:04 AM
I never pushed memory clock that much, I'll give it a try,
but guys how much power are your cards sucking at this extreme OC?

I optimised it now to around 410 sols, and the whole system with 2 cards is sucking 370W, which is a great success (at extreme overclock, it used to suck up to 500W for me).
I can reach a stable 2100 core clock, but it just sucks way too much power...


It draws 85% from the 1070, the slower one. The faster one with Samsung I can go +700 memory if I wanted instead of 575

The core clock OC has no difference in power draw if you are talking about the GTX 1070. The core clock is just a clock offset while under the same power draw it would be pushed to either way. Memory OC's will push more power into the card

These percentages tell nothing because each card has totally different TDP settings and curve. What is the consumption at the wall? When i set the settings as you have, my cards take over 160Watt. So for me, it is still much better to go 410-420 sols at 105-110W.
sr. member
Activity: 420
Merit: 250
March 05, 2017, 04:44:25 AM
I never pushed memory clock that much, I'll give it a try,
but guys how much power are your cards sucking at this extreme OC?

I optimised it now to around 410 sols, and the whole system with 2 cards is sucking 370W, which is a great success (at extreme overclock, it used to suck up to 500W for me).
I can reach a stable 2100 core clock, but it just sucks way too much power...


It draws 85% from the 1070, the slower one. The faster one with Samsung I can go +700 memory if I wanted instead of 575

The core clock OC has no difference in power draw if you are talking about the GTX 1070. The core clock is just a clock offset while under the same power draw it would be pushed to either way. Memory OC's will push more power into the card
full member
Activity: 224
Merit: 100
CryptoLearner
March 05, 2017, 04:06:17 AM
The card will not suck more power than you put your TDP limit, it'll just become unstable with too much OC, so to you to make your testings and find the sweet spot Smiley
sr. member
Activity: 337
Merit: 250
March 05, 2017, 03:33:12 AM
I never pushed memory clock that much, I'll give it a try,
but guys how much power are your cards sucking at this extreme OC?

I optimised it now to around 410 sols, and the whole system with 2 cards is sucking 370W, which is a great success (at extreme overclock, it used to suck up to 500W for me).
I can reach a stable 2100 core clock, but it just sucks way too much power...
sr. member
Activity: 420
Merit: 250
March 05, 2017, 01:11:44 AM
single gtx1070 palit gamerock (micron) with super jetstream bios - 480-490 sol/s ewbf 0.3.1b
downvolting&overclocking via afterburner curve


The average of the reported Sol/s is 478

You should set the fans higher to drop the temps to 55c it will up your OC to 2100mhz since Pascal cards throttle under heat
Jump to: