miniZ v2.0a Equihash 144,5 125,4 210,9 150,5 192,7 BeamHash3 ProgPoW Ethash CFX - page 31.

miniZ

member

Activity: 690

Merit: 17

Quote from: fistsofgod on March 17, 2020, 11:41:16 AM

is --latency option depreciated? how do i show ping time to server? I put --latency in command line and miniz errored unknown command

Hi fistsofgod,
--latency is still working.
Could you paste here your command line so that we can try to see what is going on?
Thanks for reaching us!
Cheers

fistsofgod

newbie

Activity: 164

Merit: 0

is --latency option depreciated? how do i show ping time to server? I put --latency in command line and miniz errored unknown command

miniZ

member

Activity: 690

Merit: 17

Quote from: Kringel on March 05, 2020, 08:07:58 AM

Hi,

What happened to the OC2 Beam kernel for 1070?

Quote from: sxemini on March 05, 2020, 02:58:42 AM

Hello,

--oc1/oc2 didn´t work on any algo for me. Not on Beam, not on Zhash/144,5 and so on. It says every time no oc1 or oc2 for this algo.

Hi sxemini, Kringel, and everyone,

we are sorry the message about oc1/2 usage was not clear before.

miniZ introduced a warning in version v1.5t that calls the user attention when the GPU does not support oc1/oc2, for a certain algo.

When this happens the GPU will run the default kernel, which is the best kernel we found for that GPU. It does not need oc1 or oc2 kernels for best performance.

Not all GPUs and algos have oc1/oc2 available. Some do because we realized that, for some the OC settings that you choose, those GPUs (the ones that have extra kernels available) may benefit by running a kernel other than the default.

From version to version the existence of oc1/oc2 may change. For detailed information about this, please check the dedicated information in FAQ page: https://miniz.ch/faq/

In summary, the warning is informative. There is nothing to worry. You can just ignore the message. GPU should start the default - best - kernel available.

However, do let us know if you find a regression for your GPU. We do not have access to all different GPUs available.
It can happen that you find minor improvements (or even no improvements) for a minor number of GPUs but we do not expect regressions to occur. So keep us posted if this happens.
This is the only way we can know and try to check/correct the issue.

Thanks a lot for your feedback!

Cheers

Kringel

jr. member

Activity: 151

Merit: 7

Hi,

What happened to the OC2 Beam kernel for 1070?

sxemini

member

Activity: 1558

Merit: 69

Quote from: miniZ on March 03, 2020, 05:44:42 PM

Hi everyone,

A new version miniZ v1.5t2 is out with improvements for 150,5,3 – BeamHashII.

* Improvements for 150,5,3 algo (BEAM). Major (~2-4%) for GTX 1660 Ti, 1080 Ti, and RTX GPUs.
* If mining to wallet you can write --par=beam2.

Download miniZ v1.5t2 here.

We have updated 150,5,3 performance tables (stock, eco, eco70, and max power oc) @ 150,5,3 algorithm performance page.

Few notes:
  * When adjusting your GPUs settings do not forget to test it with --oc1/oc2 since it may boost your performance a little.
--oc1/oc2 changed for 150,5,3 in the new version. For updated information read the specific topic on FAQ page.
  * On Download page you can find Cuda 10 and Cuda 8 versions (Linux and Windows).
   For Turing architectures (16xx series and RTX) you need to use Cuda 10.0 versions.

For additional information check our Usage or FAQ pages.
Let us know how it goes!

Happy mining!

Hello,

--oc1/oc2 didn´t work on any algo for me. Not on Beam, not on Zhash/144,5 and so on. It says every time no oc1 or oc2 for this algo.

miniZ

member

Activity: 690

Merit: 17

Hi everyone,

A new version miniZ v1.5t2 is out with improvements for 150,5,3 – BeamHashII.

* Improvements for 150,5,3 algo (BEAM). Major (~2-4%) for GTX 1660 Ti, 1080 Ti, and RTX GPUs.
* If mining to wallet you can write --par=beam2.

Download miniZ v1.5t2 here.

We have updated 150,5,3 performance tables (stock, eco, eco70, and max power oc) @ 150,5,3 algorithm performance page.

Few notes:
  * When adjusting your GPUs settings do not forget to test it with --oc1/oc2 since it may boost your performance a little.
--oc1/oc2 changed for 150,5,3 in the new version. For updated information read the specific topic on FAQ page.
  * On Download page you can find Cuda 10 and Cuda 8 versions (Linux and Windows).
   For Turing architectures (16xx series and RTX) you need to use Cuda 10.0 versions.

For additional information check our Usage or FAQ pages.
Let us know how it goes!

Happy mining!

miniZ

member

Activity: 690

Merit: 17

Quote from: luc1an24_nh on February 20, 2020, 03:30:32 AM

Hi miniZ
Could you please check your telemetry return values?
Sometimes I am getting invalid data even though I am mining and there is no error on my side.
I am sending `getstat` requests.
Is there any way that if you are unable to provide data atm just send previous values?
Thank you

Hi luc1an24_nh,

We had a look and could not reproduce your issue. Could you let us know which data are invalid?

Maybe if we can understand which data are invalid and under which circumstances then we can resolve this.

Thanks for the feedback!
Cheers

luc1an24_nh

newbie

Activity: 1

Merit: 0

Hi miniZ
Could you please check your telemetry return values?
Sometimes I am getting invalid data even though I am mining and there is no error on my side.
I am sending `getstat` requests.
Is there any way that if you are unable to provide data atm just send previous values?
Thank you

miniZ

member

Activity: 690

Merit: 17

Quote from: 2h4u on February 12, 2020, 06:59:27 AM

miniz,

1. how to exclude GPU 10 and 11 from the mining?
2. -cd and -cde description is a very confusing. -cd should have comma-separated list, but -cde should have a space-separated list???

Hi 2h4u,

Thanks for the feedback. Sorry that this was not clear.

1. To exclude GPU 10 and 11 you can write:

-cde=10,11
or
-cde 10 11

2. You can use both space and comma separated list with -cd and -cde. These have the same syntax.

For clarity we have updated the table in Usage page.

Hope this helps!

Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: hogwash.89m on February 11, 2020, 06:00:18 PM

Hi miniz,
Last 2 days I was mining with 1.5s version without issue, only there was one crash, and then for like 15 hours it was OK again.

Quote

[FATAL ] GPU[3]: CUDA error 77 'an illegal memory access was encountered' in line 1013

This popped out in log, this was with kind a extreme optimization and OC. It's strange that error is for all GPUs, I hope it's error for pool.
It was happening before that once in 2-3 days there are miner crash and batch loop starts it again.

Hi hogwash.89m,
thanks for the information.

The error you get really looks like OC related. It can be that it’s just one of the GPUs that is causing this.
You can try and tweak just one and see if is becomes stable…if so we suggest starting with the GPU[3] since it is usually the first reporting the error that causes the crash.

Let us know how it goes.
Cheers

2h4u

newbie

Activity: 55

Merit: 0

miniz,

1. how to exclude GPU 10 and 11 from the mining?
2. -cd and -cde description is a very confusing. -cd should have comma-separated list, but -cde should have a space-separated list???

hogwash.89m

newbie

Activity: 32

Merit: 0

Quote from: miniZ on February 09, 2020, 04:37:44 PM

Hi hogwash.89m,
Sorry to hear that there have been some issues going on.

Just to completely discard the OCs interference, have you tried mining in stock settings?
Could you let us know what algorithm are you running? And to which pool?

We will have a look into it, maybe we can reproduce this behavior.
Thanks for the feedback!
Cheers

Hi miniz,

Sorry, I didn't write which algo is in question.

150,5,3 – BeamHashII

sunpool

Also I didn't log the output so I can't show if there had been any error in console.

Last 2 days I was mining with 1.5s version without issue, only there was one crash, and then for like 15 hours it was OK again.

Quote

[ 2d 7h15m43s|Feb 11 08:16:11] S:6645/1 257(257.1)Sol/s 779(773.6)W [beam.sunpool.top]
0>GTX 1080 Ti ` 100% [59øC/47%] 27.94 I/s 55.8(55.9)Sol/s 176(173.7)W clk=1632MHz mclk=5702MHz Sol/W=0.32
1>GTX 1080 ` 100% [59øC/47%] 20.54 I/s 40.8(40.8)Sol/s 124(122.4)W clk=1556MHz mclk=5200MHz Sol/W=0.33
2>GTX 1080 `99.9% [57øC/45%] 20.14 I/s 40.4(39.9)Sol/s 118(118.5)W clk=1531MHz mclk=5200MHz Sol/W=0.34
3>GTX 1080 ` 100% [57øC/44%] 20.36 I/s 40.1(40.4)Sol/s 121(120.5)W clk=1556MHz mclk=5200MHz Sol/W=0.34
4>GTX 1080 ` 100% [51øC/40%] 19.98 I/s 38.7(39.5)Sol/s 117(115.7)W clk=1518MHz mclk=5200MHz Sol/W=0.34
5>GTX 1080 ` 100% [58øC/46%] 20.46 I/s 40.4(40.7)Sol/s 123(122.3)W clk=1556MHz mclk=5200MHz Sol/W=0.33
[FATAL ] GPU[3]: CUDA error 77 'an illegal memory access was encountered' in line 1013
[INFO ] Disconnecting from beam.sunpool.top
[FATAL ] GPU[4]: CUDA error 77 'an illegal memory access was encountered' in line 1013
[FATAL ] GPU[2]: CUDA error 77 'an illegal memory access was encountered' in line 1013
[FATAL ] GPU[5]: CUDA error 77 'an illegal memory access was encountered' in line 1013
[FATAL ] GPU[1]: CUDA error 77 'an illegal memory access was encountered' in line 1013
[FATAL ] GPU[0]: CUDA error 77 'an illegal memory access was encountered' in line 1013

This popped out in log, this was with kind a extreme optimization and OC. It's strange that error is for all GPUs, I hope it's error for pool.
It was happening before that once in 2-3 days there are miner crash and batch loop starts it again.

Now I started 1.5t without any OC, only power limit down to ~180W for 1080ti and ~120W for 1080.
I don't like that extra heat and noise Grin

.

I will report in day or two how is going.

miniZ

member

Activity: 690

Merit: 17

Quote from: rgsnedds on February 01, 2020, 04:19:24 PM

many thanks for all your hard work miniZ!

Hi rgsnedds,
thanks for your support!
Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: hogwash.89m on February 08, 2020, 07:45:59 PM

I think there is a problem with 1.5t version. Huh

Around 2. Feb I updated to 1.5t version, and from then 3 times I got one card "minining" with 0 Sol/s (or 0 point something).
First time I thought that it could be faulty pci-e riser, they are almost 2 years old, then I restated miner with same setting (also same OC) and then again after like 12-24 hours different card drop to 0 Sol/s.

Then I lower OC settings to confirm if it's something wrong with cards, and then after almost 2 days 3rd card drop to 0 Sol/s, in all cases power consumption also dropped to idle.

All 3 cars are GTX 1080, in a rig with 1x 1080ti + 5x 1080, with fee-time set to 40 sec and OC1 but later I saw that that is not supported on 1080.

So I reverted to 1.5s to confirm if it's hardware problem or problem with new miner version.
There wasn't any changes to driver or other software.

On 1.5s I was mining for more than 2 months without a problem.
Only minor defect on 1.5s is that miner was switching to dev fee but cards actually wasn't mining (gpu usage was 0, power consumption idle) but luckly they always returned to mining properly to pool.

Hi hogwash.89m,
Sorry to hear that there have been some issues going on.

Just to completely discard the OCs interference, have you tried mining in stock settings?
Could you let us know what algorithm are you running? And to which pool?

We will have a look into it, maybe we can reproduce this behavior.
Thanks for the feedback!
Cheers

hogwash.89m

newbie

Activity: 32

Merit: 0

I think there is a problem with 1.5t version. Huh

Around 2. Feb I updated to 1.5t version, and from then 3 times I got one card "minining" with 0 Sol/s (or 0 point something).
First time I thought that it could be faulty pci-e riser, they are almost 2 years old, then I restated miner with same setting (also same OC) and then again after like 12-24 hours different card drop to 0 Sol/s.

Then I lower OC settings to confirm if it's something wrong with cards, and then after almost 2 days 3rd card drop to 0 Sol/s, in all cases power consumption also dropped to idle.

All 3 cars are GTX 1080, in a rig with 1x 1080ti + 5x 1080, with fee-time set to 40 sec and OC1 but later I saw that that is not supported on 1080.

So I reverted to 1.5s to confirm if it's hardware problem or problem with new miner version.
There wasn't any changes to driver or other software.

On 1.5s I was mining for more than 2 months without a problem.
Only minor defect on 1.5s is that miner was switching to dev fee but cards actually wasn't mining (gpu usage was 0, power consumption idle) but luckly they always returned to mining properly to pool.

rgsnedds

member

Activity: 952

Merit: 17

raskul

many thanks for all your hard work miniZ!

miniZ

member

Activity: 690

Merit: 17

Quote from: monster7321 on January 31, 2020, 08:39:37 AM

Don't improvements: 150,5 Verified on 1080 and 1080ti graphics cards!

Hi monster7321,

thank you for the feedback, but could you clarify if you had improvements?

Or regression? Sad

Improvements depend on GPU. 150,5 (GRIMM and CASH algo) improved for most GPUs (particularly for Turing architectures).

Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: ?? on ??

Why isnt there a Cuda 9.1/9.2 version and you drop down to Cuda 8? This is the only miner out there that has CUDA 8 then jumps to Cuda 10?

I dont have cuda 10 yet and cuda 8 severely degrades my performance.

Hi teammaf,

Which NVIDIA driver version do you have installed?

To run miniZ CUDA10 you just need to have NVIDIA driver >=411.31 (Windows) or >=410.48 (Linux).
For driver <=391 the miniZ CUDA8 will run at full speed.

miniZ CUDA version runs independently of the CUDA version you may have installed.
It is the NVIDIA driver version that matters.

Cheers

monster7321

newbie

Activity: 1

Merit: 0

Quote from: miniZ on January 29, 2020, 01:25:31 PM

Hi everyone,

Thank you for your patience! A new version miniZ v1.5t is out!

We took a bit longer this time (sorry for the silence), however we kept working and did our best to make miniZ a fast and friendly equihash miner!

* Major improvements: 150,5 and 210,9. Up to 4% depending on algo and GPU.
  Also for RTX 2070 (up to 8% on 192,7), 1660 Ti.
* Minor improvements in all other algos. Up to 2% depending on algo and GPU.
* Better support for GTX 1660 Ti and GTX 1070.
* Improved stability.
* Fixed telemetry on Chrome browser.
* Fixed CPU load when mining Beam on Windows.
* Cleaner program exit/termination.
* Added --color and --fee-time to command line options. (see Usage page)

Download miniZ v1.5t here.

Few notes:
  * When adjusting your GPUs settings do not forget to test it with --oc1/oc2 since it may boost your performance a little. --oc1/oc2 changed for some GPUs and algos in the new version.
For updated information read the specific topic on FAQ page.
  * On Download page you can find Cuda 10 and Cuda 8 versions (Linux and Windows).
For Turing architectures (16xx series and RTX) you need to use Cuda 10.0 versions.

For additional information check our Usage or FAQ pages.
Let us know how it goes!

Happy mining!

Don't improvements: 150,5 Verified on 1080 and 1080ti graphics cards!

miniZ

member

Activity: 690

Merit: 17

Hi everyone,

Thank you for your patience! A new version miniZ v1.5t is out!

We took a bit longer this time (sorry for the silence), however we kept working and did our best to make miniZ a fast and friendly equihash miner!

* Major improvements: 150,5 and 210,9. Up to 4% depending on algo and GPU.
  Also for RTX 2070 (up to 8% on 192,7), 1660 Ti.
* Minor improvements in all other algos. Up to 2% depending on algo and GPU.
* Better support for GTX 1660 Ti and GTX 1070.
* Improved stability.
* Fixed telemetry on Chrome browser.
* Fixed CPU load when mining Beam on Windows.
* Cleaner program exit/termination.
* Added --color and --fee-time to command line options. (see Usage page)

Download miniZ v1.5t here.

Few notes:
  * When adjusting your GPUs OC settings check if your GPU supports oc1/oc2 (FAQ). If so it may boost your performance a little.
--oc1/oc2 changed for some GPUs, for 150,5,3 (BEAM) in the new version.
   For updated and more detailed information read the specific topic on FAQ page.

  * On Download page you can find Cuda 10 and Cuda 8 versions (Linux and Windows).
   For Turing architectures (16xx series and RTX) you need to use Cuda 10.0 versions.

For additional information check our Usage or FAQ pages.
Let us know how it goes!

Happy mining!

Topic: miniZ v2.0a Equihash 144,5 125,4 210,9 150,5 192,7 BeamHash3 ProgPoW Ethash CFX - page 31. (Read 59781 times)