miniZ v2.0a Equihash 144,5 125,4 210,9 150,5 192,7 BeamHash3 ProgPoW Ethash CFX - page 69.

miniZ

member

Activity: 690

Merit: 17

Quote from: thereal35 on January 09, 2019, 12:49:53 PM

Heads up,
I couldnt get this Miner to successfully run my 6x1070 rig.
Turns out that the 16gb Virtual memory i had set on Windows was not enough.
Bumped it up to 18gb (18432) and it was able to start mining.
It actually seemed like it wasnt going to run, and eventually fired up.
I think i need to bump up to at least 20gb virtual mem, but do not have that available on this rig.
Im going to hook up another SSD and set the virtual mem to it instead.

I had the same issue trying to run another miner (BEAM).
even though i have 8gb ram and all 6 GPUs have 8gb, the virtual memory was not enough on my system to run it.
(A 60gb ssd in windows does not hold up over time)

hope this helps someone.

Hi thereal35,
thank you for sharing your findings with everyone.
Indeed page file needs to be set to high values. To be on the safe side, the best is to set it approximately to the sum of all GPUs memories.

Others have also shared some thoughts/experiences on this matter. See:
https://bitcointalksearch.org/topic/m.48173601

Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: 1ncehost on January 09, 2019, 08:42:49 AM

Hello, I believe I found a regression that explains why my 1060s were running slower on newest versions.

Two of my cards have poor thermal performance and run hot. I have noticed that on one of my rigs MiniZ runs the hot cards in a different algorithm configuration (the vram usage is higher and the wattage is lower). However, in the other of my rigs, it runs the wrong card in this "hot mode".

If I am correct in understanding how MiniZ works, on this rig my hot card is getting thermally limited by the drivers (and running unoptimally) while simultaniously a cool card is getting limited by MiniZ, running unnecessarily low wattage.

Hope this helps

Hi 1ncehost,
sorry for the late reply.
Thank you for letting us know.

miniZ requires 2.7Gb of memory to run at full speed, but on Windows the cards may run sub-optimally (specially the 3Gb cards, which are close to the limit value) because the OS drivers consume more memory. If this happens miniZ runs a different mode (usually mode 2 for GTX 1060 3GB). This problem should not happen in Linux.

Cheers

thereal35

newbie

Activity: 13

Merit: 0

Heads up,
I couldnt get this Miner to successfully run my 6x1070 rig.
Turns out that the 16gb Virtual memory i had set on Windows was not enough.
Bumped it up to 18gb (18432) and it was able to start mining.
It actually seemed like it wasnt going to run, and eventually fired up.
I think i need to bump up to at least 20gb virtual mem, but do not have that available on this rig.
Im going to hook up another SSD and set the virtual mem to it instead.

I had the same issue trying to run another miner (BEAM).
even though i have 8gb ram and all 6 GPUs have 8gb, the virtual memory was not enough on my system to run it.
(A 60gb ssd in windows does not hold up over time)

hope this helps someone.

1ncehost

copper member

Activity: 20

Merit: 4

Hello, I believe I found a regression that explains why my 1060s were running slower on newest versions.

Two of my cards have poor thermal performance and run hot. I have noticed that on one of my rigs MiniZ runs the hot cards in a different algorithm configuration (the vram usage is higher and the wattage is lower). However, in the other of my rigs, it runs the wrong card in this "hot mode".

If I am correct in understanding how MiniZ works, on this rig my hot card is getting thermally limited by the drivers (and running unoptimally) while simultaniously a cool card is getting limited by MiniZ, running unnecessarily low wattage.

Hope this helps

miniZ

member

Activity: 690

Merit: 17

Quote from: KriptoFull on January 05, 2019, 12:41:20 PM

Quote from: miniZ on January 01, 2019, 10:54:19 PM

Quote from: alucard20724 on December 30, 2018, 12:15:33 AM

will you be creating a windows cuda 10 version?

Hi alucard20724,
we can add that if it is really necessary. Why would you need a windows cuda 10 version for?
Cheers

Hi!Tell me what the parameter shows in the minner I/s

Hi KriptoFull,

I/s refers to iterations per second. This is roughly half the Sol/s (1 Sol/s = 0.5 I/s), since each equihash iteration/run generates aproximately, on average, 2 solutions. This value can vary but the most probable values range ~0-4 (but can be more). This is why I/s generally stabilizes faster than the Sol/s value, and the latter should always be compared after a few hours.

Cheers

KriptoFull

newbie

Activity: 394

Merit: 0

Quote from: miniZ on January 01, 2019, 10:54:19 PM

Quote from: alucard20724 on December 30, 2018, 12:15:33 AM

will you be creating a windows cuda 10 version?

Hi alucard20724,
we can add that if it is really necessary. Why would you need a windows cuda 10 version for?
Cheers

Hi!Tell me what the parameter shows in the minner I/s

miniZ

member

Activity: 690

Merit: 17

Quote from: alucard20724 on December 30, 2018, 12:15:33 AM

will you be creating a windows cuda 10 version?

Hi alucard20724,
we can add that if it is really necessary. Why would you need a windows cuda 10 version for?
Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: supervitek on December 23, 2018, 03:58:20 PM

Hi MiniZ,

Your miner is closing with an error, any idea where can I check what kind of error is happening?

I tried to use it on 2 1050 Ti cards and it is always closing after 20-30 mins. I updated the system to a last driver - no success. I used 9.2 and 8 versions - same problem.

Also if I set --intensity 80 (any intensity) the miners stuck like this:

Thank you!

P.S. This is how it stuck:

Hi supervitek,
thanks for reporting this.
Does this problem happen only when using the intensity?
To rule out a possible bug with intensity option we suggest to run miniZ without setting this option. Let us know how did it go.

We also suggest you to use the nocolor option (v1.0i++), for a cleaner console output. v1.0h does not have the nocolor option implemented.
However, you can use another console. We tried ConEmu terminal emulator and it worked fine. There are others available, which should also work OK.

Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: supervitek on December 24, 2018, 07:08:02 AM

Quote from: 1ncehost on December 24, 2018, 03:20:02 AM

I have GTX 1060s running at half power and stock clocks. H was the fastest build on them. "I" ran about 5% slower and i++ ran slower still. The time sample size for i and i++ could be small enough to not be meaningful (I didn't run I for long), but the sample size between H and I++ was meaningful. H was considerably faster than I++.

That is interesting, one of my card in a RIG is 1060 3 Gb and "I++" is faster than "H" (I tried H cause I have non stop closing of the program with no error message in about 20 mins - 1 hour).

Quote from: 1ncehost on December 24, 2018, 03:20:02 AM

I have GTX 1060s running at half power and stock clocks. H was the fastest build on them. "I" ran about 5% slower and i++ ran slower still. The time sample size for i and i++ could be small enough to not be meaningful (I didn't run I for long), but the sample size between H and I++ was meaningful. H was considerably faster than I++.

Hi 1ncehost, supervitek,
Thanks for reporting this. That is interesting. We tested on a 1060 3Gb, v1.0i++ (https://miniz.ch/features/) is significantly faster than v1.0h (https://miniz.ch/2018/11/30/my-sol-s/).
There is some variability since these values depend on brand, temperature, stock/OC settings...
We are working on a new version that should be faster than both v1.0h and v1.0h

Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: rekoe on December 23, 2018, 12:05:27 AM

Very fast. Thanks to the developers.

Hi rekoe,
thank you!

Quote from: zorkilion on December 27, 2018, 07:41:53 AM

I like your miner. Testing now

Hi zorkilion,
thanks. Then share with us how it went.

Quote from: lost2217 on December 27, 2018, 04:39:09 AM

msi armor 2070 63-65 Sol/s

Hi lost2217,
thank you for letting us know! Feedback on RTX is very much appreciated!

miniZ

member

Activity: 690

Merit: 17

Quote from: fistsofgod on December 23, 2018, 04:34:51 PM

0>GTX 1070 [52°C/61%]
1>GTX 1070 [45°C/55%]
2>GTX 1070 [46°C/55%]

with the space between card name and temp/fan could you put in per gpu rejected share number?

Hi fistsofgod,
we'll do that for the next version.

Quote from: fistsofgod on January 01, 2019, 06:13:45 PM

192,7

We are working on improving the 144,5, but we plan to work on that in a near future

Cheers

fistsofgod

newbie

Activity: 164

Merit: 0

192,7

alucard20724

sr. member

Activity: 703

Merit: 272

will you be creating a windows cuda 10 version?

zorkilion

newbie

Activity: 3

Merit: 0

I like your miner. Testing now

lost2217

newbie

Activity: 8

Merit: 0

msi armor 2070 63-65 Sol/s

supervitek

newbie

Activity: 147

Merit: 0

Quote from: 1ncehost on December 24, 2018, 03:20:02 AM

Quote from: miniZ on December 21, 2018, 09:07:17 PM

Hi 1ncehost,

Sorry for the inconvenience... we try to make sure that there are no regressions on new releases, but there are so many cards/overclock setting combinations that this might happend.
Which cards do you have on your rigs and which OS are you using? The changes from i+ to i++ are minimal, it would be interesting to understand what caused the problem. Thanks for your feedback!

Cheers and happy mining!

I have GTX 1060s running at half power and stock clocks. H was the fastest build on them. "I" ran about 5% slower and i++ ran slower still. The time sample size for i and i++ could be small enough to not be meaningful (I didn't run I for long), but the sample size between H and I++ was meaningful. H was considerably faster than I++.

That is interesting, one of my card in a RIG is 1060 3 Gb and "I++" is faster than "H" (I tried H cause I have non stop closing of the program with no error message in about 20 mins - 1 hour).

1ncehost

copper member

Activity: 20

Merit: 4

Quote from: miniZ on December 21, 2018, 09:07:17 PM

Hi 1ncehost,

Sorry for the inconvenience... we try to make sure that there are no regressions on new releases, but there are so many cards/overclock setting combinations that this might happend.
Which cards do you have on your rigs and which OS are you using? The changes from i+ to i++ are minimal, it would be interesting to understand what caused the problem. Thanks for your feedback!

Cheers and happy mining!

I have GTX 1060s running at half power and stock clocks. H was the fastest build on them. "I" ran about 5% slower and i++ ran slower still. The time sample size for i and i++ could be small enough to not be meaningful (I didn't run I for long), but the sample size between H and I++ was meaningful. H was considerably faster than I++.

fistsofgod

newbie

Activity: 164

Merit: 0

0>GTX 1070 [52°C/61%]
1>GTX 1070 [45°C/55%]
2>GTX 1070 [46°C/55%]

with the space between card name and temp/fan could you put in per gpu rejected share number?

supervitek

newbie

Activity: 147

Merit: 0

Hi MiniZ,

Your miner is closing with an error, any idea where can I check what kind of error is happening?

I tried to use it on 2 1050 Ti cards and it is always closing after 20-30 mins. I updated the system to a last driver - no success. I used 9.2 and 8 versions - same problem.

Also if I set --intensity 80 (any intensity) the miners stuck like this:

https://i.ibb.co/qMhQSrF/miniz-stuck-on-intensity.png

Thank you!

P.S. This is how it stuck:

https://i.ibb.co/hczrFqc/miniz-error-stop.png

rekoe

newbie

Activity: 5

Merit: 0

Very fast. Thanks to the developers.

Topic: miniZ v2.0a Equihash 144,5 125,4 210,9 150,5 192,7 BeamHash3 ProgPoW Ethash CFX - page 69. (Read 59781 times)