EWBF's CUDA Zcash miner - page 256. | Bitcointalksearch.org

tbearhere

legendary

Activity: 3164

Merit: 1003

Quote from: krnlx on December 27, 2016, 07:42:51 AM

Quote from: EWBF_ on December 27, 2016, 07:30:01 AM

Hi everybody!

A couple of answers to your questions:
Expected speed ~360 on stock 1070 ~250 on stock 1060
Stability, miner now work with "intensity" it better utilize gpu and as result it can make strong overclock less stable.
DevFee, miner works for devfee 2% of time. But difficulty on devfee pool can be significantly lower or higher than on your pool.

It would be good, if would be intensity param, to fine tune cards.

That would be nice. I've been waiting for that for a long time.. intensity setting.

CoffeeCat

newbie

Activity: 39

Merit: 0

Quote from: m1n1ngP4d4w4n on December 28, 2016, 12:42:24 PM

Quote from: CoffeeCat on December 28, 2016, 12:35:21 PM

What you would folks say are the best clocks for 1070 and a 1080? I want to get the most hashing power while saving on electricity. (Electric costs are high for me.)

If your electricity costs are high, best keep @50% tdp, that's what i do, any other combo just isn't enough profitability.

for OC settings, you would probably have to do your own testing, because each card is different depending on design, makers, making and so on.

My own settings atm and im still working on them, is +135 core / +575 mem @50% tdp +20% vCore (6x EVGA FTW 1070 micron chip memory) yielding about 375sols/s each for 92w each in nvidia-smi, whole rig consuming 695w (cpu, mobo, fans, and psu efficiency in play). I also run with 3 ewbf process, 2x cards per process for maximum efficiency (i have to test with lower amount of process since ewbf made changes in 0.0.7b, to see if it's as efficient running only one or 2 process, would save a few watts, will see that this week-end)

Those are just general settings, you would have to make your own testings to find the sweetest spot

Hey, thanks for the above. That's in line with my 1070. Any thoughts on the 1080? You can bring the tdp down to 37% and I know the memory is different.

petrmaje

newbie

Activity: 2

Merit: 0

Hi,
I have a issue with yours miner, Linux version.
Latest working version is 0.05b.

+---------------------------------+
| EWBF's Zcash CUDA miner. 0.0.5b |
+---------------------------------+
INFO: Target: 007fffffffffffff...
INFO: Detected new work: 1482962614_64854
INFO: Detected new work: 1482962614_64854
CUDA: Device: 0 GeForce GTX 1060 6GB, 6072 MB
INFO 23:04:01: GPU0 Accepted share 297ms [A:1, R:0]
INFO 23:04:07: GPU0 Accepted share 298ms [A:2, R:0]

Newer versions crased with this:
+---------------------------------+
| EWBF's Zcash CUDA miner. 0.0.8b |
+---------------------------------+
INFO: Target: 007fffffffffffff...
INFO: Detected new work: 1482962393_64849
INFO: Detected new work: 1482962393_64849
ERROR: Cannot run workers.

Does your miner require OPENCL files? I have installed driver without OPENCL files, because I had problems with second videocard installed in my PC. Other miners (claymore, genoil) works well. Where is the problem?
Thanks
Petr

m1n1ngP4d4w4n

full member

Activity: 224

Merit: 100

CryptoLearner

Quote from: xPwnK on December 28, 2016, 02:35:29 PM

Quote from: anorganix on December 28, 2016, 02:30:55 PM

I see a lot of people using 376.09 drivers.
Is there a specific reason not to use latest drivers? I'm mining just fine with 376.33.

No, but there isn't reason to update your drivers either.
376.33 drivers don't increase your hashrate.

Well with the newest pascal cards, you can always hope something will improve, but usually it'll do nothing if your previous version isn't that far from current. I recommend doing testing, and for easyness to change between version use restore point in windows, very useful

Quote from: bensam1231 on December 28, 2016, 03:23:52 PM

Updating my bios's was a huge PITA all for about 8% more hashrate. When you have mixed manufacturers on one PC a lot of the update utilities don't respect models and will flash their bios over other manufacturers cards. That was a lot of fun digging up bios's to flash them back after that. Other manufacturers wont flash multiple cards (Asus update). It was a grab bag.

Best method is using nvflash and updating one of the cards with the manufacturers package, saving the updated bios with Nvflash, and then flashing all the cards at once by unplugging other models from the system.

In other news apparently you can't get that last couple % out of cards by running multiple instances anymore. EWBF still only uses 97-98% of my GPUs where as Nicehash uses 99-100%.

Worth noting, when running on Maxwell (970s) hashrate is 10% slower then Nicehash's miner. That's a pretty big difference. If you're running multiple Maxwell rigs it's better to stick with EQM.

Also .8 is still quite unstable. Multiple machines were completely stable with EQM, I have to reduce the memory clock a extra 100mhz to be stable in .8.

Latest version 0.0.8b with 3 process has an average GPU usage of 99% for me. (1 bat file per 2 cards on 6 cards rigs).

Stability wise, running with no rejected shares for the last 4 hours @ average 374sols/s per card. I use the same OC i was using with EQM, so far so good

Agree with you, bios updating for heterogeneous rigs is horrible...

Code:

Would love to see threading support from EWBF, as well as loggin and monitoring features for sure Wink

bensam1231

legendary

Activity: 1764

Merit: 1024

Updating my bios's was a huge PITA all for about 8% more hashrate. When you have mixed manufacturers on one PC a lot of the update utilities don't respect models and will flash their bios over other manufacturers cards. That was a lot of fun digging up bios's to flash them back after that. Other manufacturers wont flash multiple cards (Asus update). It was a grab bag.

Best method is using nvflash and updating one of the cards with the manufacturers package, saving the updated bios with Nvflash, and then flashing all the cards at once by unplugging other models from the system.

In other news apparently you can't get that last couple % out of cards by running multiple instances anymore. EWBF still only uses 97-98% of my GPUs where as Nicehash uses 99-100%.

Worth noting, when running on Maxwell (970s) hashrate is 10% slower then Nicehash's miner. That's a pretty big difference. If you're running multiple Maxwell rigs it's better to stick with EQM.

Also .8 is still quite unstable. Multiple machines were completely stable with EQM, I have to reduce the memory clock a extra 100mhz to be stable in .8.

xPwnK

newbie

Activity: 41

Merit: 0

Quote from: anorganix on December 28, 2016, 02:30:55 PM

I see a lot of people using 376.09 drivers.
Is there a specific reason not to use latest drivers? I'm mining just fine with 376.33.

No, but there isn't reason to update your drivers either.
376.33 drivers don't increase your hashrate.

anorganix

copper member

Activity: 970

Merit: 287

Per aspera ad astra

I see a lot of people using 376.09 drivers.
Is there a specific reason not to use latest drivers? I'm mining just fine with 376.33.

_javi_

hero member

Activity: 968

Merit: 624

Still a manic miner

Some workers are stopped attempting to restart....
and again and again...

No OC...

376.09 drivers..

is this ok for this miner..=
setx GPU_FORCE_64BIT_PTR 0
setx GPU_MAX_HEAP_SIZE 100
setx GPU_USE_SYNC_OBJECTS 1
setx GPU_MAX_ALLOC_PERCENT 100
setx GPU_SINGLE_ALLOC_PERCENT 100

??

Kompik

sr. member

Activity: 463

Merit: 250

Quote from: m1n1ngP4d4w4n on December 28, 2016, 01:28:53 PM

Quote from: Kompik on December 28, 2016, 01:26:40 PM

Quote

In the afterburner option you need to go in general tab, and in the compatibility properties section, check unlock voltage control "extended MSI", and unlock voltage monitoring. You should be able to OC Vcore after afterburner restarting.
For the process, i run 3 instance of EWBF each targeting 2 cards (i have basically 3 bat file with the --cuda-devices switch)

Yes I have that but it does seem to do anything! Cheesy

Did you upgraded your GPU bios ? maybe not at the latest version, i "seem" to remember it was locked @ first on my FTW ones, maybe voltage are still locked on SC version i have no idea, sorry mate Sad

Yes I did, but i will try to see if there are some new ones! You are right that it is probably locked. Thanks!

m1n1ngP4d4w4n

full member

Activity: 224

Merit: 100

CryptoLearner

Quote from: Kompik on December 28, 2016, 01:26:40 PM

Quote

In the afterburner option you need to go in general tab, and in the compatibility properties section, check unlock voltage control "extended MSI", and unlock voltage monitoring. You should be able to OC Vcore after afterburner restarting.
For the process, i run 3 instance of EWBF each targeting 2 cards (i have basically 3 bat file with the --cuda-devices switch)

Yes I have that but it does seem to do anything! Cheesy

Did you upgraded your GPU bios ? maybe not at the latest version, i "seem" to remember it was locked @ first on my FTW ones, maybe voltage are still locked on SC version i have no idea, sorry mate Sad

Kompik

sr. member

Activity: 463

Merit: 250

Quote

In the afterburner option you need to go in general tab, and in the compatibility properties section, check unlock voltage control "extended MSI", and unlock voltage monitoring. You should be able to OC Vcore after afterburner restarting.
For the process, i run 3 instance of EWBF each targeting 2 cards (i have basically 3 bat file with the --cuda-devices switch)

Yes I have that but it does seem to do anything! Cheesy

m1n1ngP4d4w4n

full member

Activity: 224

Merit: 100

CryptoLearner

Quote from: Kompik on December 28, 2016, 01:13:08 PM

Quote from: m1n1ngP4d4w4n on December 28, 2016, 12:42:24 PM

Quote from: CoffeeCat on December 28, 2016, 12:35:21 PM

What you would folks say are the best clocks for 1070 and a 1080? I want to get the most hashing power while saving on electricity. (Electric costs are high for me.)

If your electricity costs are high, best keep @50% tdp, that's what i do, any other combo just isn't enough profitability.

for OC settings, you would probably have to do your own testing, because each card is different depending on design, makers, making and so on.

My own settings atm and im still working on them, is +135 core / +575 mem @50% tdp +20% vCore (6x EVGA FTW 1070 micron chip memory) yielding about 375sols/s each for 92w each in nvidia-smi, whole rig consuming 695w (cpu, mobo, fans, and psu efficiency in play). I also run with 3 ewbf process, 2x cards per process for maximum efficiency (i have to test with lower amount of process since ewbf made changes in 0.0.7b, to see if it's as efficient running only one or 2 process, would save a few watts, will see that this week-end)

Those are just general settings, you would have to make your own testings to find the sweetest spot

How do you up the vCore on the EVGA? I have EVGA SC and when i up the core in MSI afterburner it has absolutelly no effect. To keep the 375 sols i therefore need to have at least 65% TDP with the same overclock as you do. It consumes roughly 110 - 115W for 375 sols.

In the afterburner option you need to go in general tab, and in the compatibility properties section, check unlock voltage control "extended MSI", and unlock voltage monitoring. You should be able to OC Vcore after afterburner restarting.
For the process, i run 3 instance of EWBF each targeting 2 cards (i have basically 3 bat file with the --cuda-devices switch)
as for the sols/s, this is an average, i get sometimes 380, sometimes 365/370, but yeah on average i hit about 745/750 per EWBF process (each with 2 cards) using 376.19 nvidia drivers just for info too

Kompik

sr. member

Activity: 463

Merit: 250

Quote from: m1n1ngP4d4w4n on December 28, 2016, 12:42:24 PM

Quote from: CoffeeCat on December 28, 2016, 12:35:21 PM

What you would folks say are the best clocks for 1070 and a 1080? I want to get the most hashing power while saving on electricity. (Electric costs are high for me.)

If your electricity costs are high, best keep @50% tdp, that's what i do, any other combo just isn't enough profitability.

for OC settings, you would probably have to do your own testing, because each card is different depending on design, makers, making and so on.

My own settings atm and im still working on them, is +135 core / +575 mem @50% tdp +20% vCore (6x EVGA FTW 1070 micron chip memory) yielding about 375sols/s each for 92w each in nvidia-smi, whole rig consuming 695w (cpu, mobo, fans, and psu efficiency in play). I also run with 3 ewbf process, 2x cards per process for maximum efficiency (i have to test with lower amount of process since ewbf made changes in 0.0.7b, to see if it's as efficient running only one or 2 process, would save a few watts, will see that this week-end)

Those are just general settings, you would have to make your own testings to find the sweetest spot

How do you up the vCore on the EVGA? I have EVGA SC and when i up the core in MSI afterburner it has absolutelly no effect. To keep the 375 sols i therefore need to have at least 65% TDP with the same overclock as you do. It consumes roughly 110 - 115W for 375 sols.
Also saying you run two processes for one card means 12 EWBF miners running per 6x rig?

m1n1ngP4d4w4n

full member

Activity: 224

Merit: 100

CryptoLearner

Quote from: CoffeeCat on December 28, 2016, 12:35:21 PM

What you would folks say are the best clocks for 1070 and a 1080? I want to get the most hashing power while saving on electricity. (Electric costs are high for me.)

If your electricity costs are high, best keep @50% tdp, that's what i do, any other combo just isn't enough profitability.

for OC settings, you would probably have to do your own testing, because each card is different depending on design, makers, making and so on.

My own settings atm and im still working on them, is +135 core / +575 mem @50% tdp +20% vCore (6x EVGA FTW 1070 micron chip memory) yielding about 375sols/s each for 92w each in nvidia-smi, whole rig consuming 695w (cpu, mobo, fans, and psu efficiency in play). I also run with 3 ewbf process, 2x cards per process for maximum efficiency (i have to test with lower amount of process since ewbf made changes in 0.0.7b, to see if it's as efficient running only one or 2 process, would save a few watts, will see that this week-end)

Those are just general settings, you would have to make your own testings to find the sweetest spot

CoffeeCat

newbie

Activity: 39

Merit: 0

What you would folks say are the best clocks for 1070 and a 1080? I want to get the most hashing power while saving on electricity. (Electric costs are high for me.)

t2yax

sr. member

Activity: 462

Merit: 250

Arianee:Smart-link Connecting Owners,Assets,Brands

any plans for zcoin ??

lyolyalya

sr. member

Activity: 455

Merit: 250

Quote from: m1n1ngP4d4w4n on December 28, 2016, 11:42:52 AM

Nice work for 0.0.8b, improved stability and a tad bit more hashrate.

Would love to see some more options like logging, remote json admin port and stuff

Keep up the good work !

joining=)best miner!keep on tuning=)

xPwnK

newbie

Activity: 41

Merit: 0

970 gets a bit less hashrate with 0.0.8 than 0.0.7. Does anyone else have this problem with maxwell cards?
260 sol/s with 0.0.7 and 250 sol/s with 0.0.8

m1n1ngP4d4w4n

full member

Activity: 224

Merit: 100

CryptoLearner

Nice work for 0.0.8b, improved stability and a tad bit more hashrate.

Would love to see some more options like logging, remote json admin port and stuff

Keep up the good work !

du44

full member

Activity: 251

Merit: 100

Can confirm, what 0.0.8b more stable than 0.0.7b and little more hashrate.
Thanks.

Topic: EWBF's CUDA Zcash miner - page 256. (Read 2164327 times)