miniZ v2.0a Equihash 144,5 125,4 210,9 150,5 192,7 BeamHash3 ProgPoW Ethash CFX - page 40.

MineOrDie

jr. member

Activity: 75

Merit: 1

Quote from: darkneorus on August 25, 2019, 11:09:16 AM

Quote from: Dr_Victor on August 25, 2019, 10:31:48 AM

Quote from: darkneorus on August 24, 2019, 05:23:23 PM

something is broken with 150,5,3 on P104-100 GPUs. the hashrate is low, 20-30 Sol/s max, expected 40+.

Feel free to add +500 for memory, use the pill and enjoy 40+

pill does nothing for P104-100 GPUs. it's useful only for 1080 series.
GMiner shows 40-42 H/s with mem at stock clock.
something is clearly wrong with 150,5,3 implementation.
on 150,5 algo GMiner and miniZ hashrates were about equal, on 150,5,3 miniZ is far behind.

Use --oc2.

darkneorus

jr. member

Activity: 238

Merit: 3

Quote from: miniZ on August 25, 2019, 01:39:54 PM

Could you just confirm that mode 3 was selected? Should appear something like this in the beginning of output:
miniZ<150,5,3>[3:1:00:3933]: Selecting GPU#0[0] P104-100

yup

Code:

Optimisation:   oc1[0 1 2 3 4 5 6 7 8]
Temp. limit:    [90°C]
miniZ<150,5,3>[3:1:00:5998]: Selecting GPU#6[6] P106-100
miniZ<150,5,3>[3:1:00:3933]: Selecting GPU#7[7] P104-100
miniZ<150,5,3>[3:1:00:3933]: Selecting GPU#2[2] P104-100
miniZ<150,5,3>[3:1:00:3933]: Selecting GPU#8[8] P104-100
miniZ<150,5,3>[3:1:00:3933]: Selecting GPU#3[3] P104-100
miniZ<150,5,3>[3:1:00:5998]: Selecting GPU#4[4] P106-100
miniZ<150,5,3>[3:1:00:3933]: Selecting GPU#0[0] P104-100
miniZ<150,5,3>[3:1:00:3933]: Selecting GPU#1[1] P104-100
miniZ<150,5,3>[3:1:00:3933]: Selecting GPU#5[5] P104-100

GPU Bus load, reported by GPU-Z, also varies from 7 to 16%, with worst performing GPUs using less bus bandwidth.

miniZ

member

Activity: 690

Merit: 17

Quote from: darkneorus on August 25, 2019, 01:13:09 PM

also, forgot to mention that for some reason GPUs 0, 1 and 2 performance is even lower that the others, and it was like this in previous builds too.
there are no similar issues when miniZ is mining other algos.

the driver version is 416.34.

Hi darkneorus,
Ok. This is a bit unexpected, but good to know. P106 does not look good as well.

Could you just confirm that mode 3 was selected? Should appear something like this in the beginning of output:
miniZ<150,5,3>[3:1:00:3933]: Selecting GPU#0[0] P104-100

We also noticed that you have quite range of performances (and power) on your rig.

We need to think about these GPUs for a while.
Thanks a lot for your help!
Cheers

darkneorus

jr. member

Activity: 238

Merit: 3

also, forgot to mention that for some reason GPUs 0, 1 and 2 performance is even lower that the others, and it was like this in previous builds too.
there are no similar issues when miniZ is mining other algos.

the driver version is 416.34.

without any parameters

Code:

[ 0d 0h 3m00s] 203(202.6)Sol/s 1199(1234.3)W
 0>P104-100     ` 100% [45°C/80%]* 9.22 I/s 18.3(18.3)Sol/s 128(141.0)W clk=1885MHz mclk=5005MHz Sol/W=0.13
 1>P104-100     ` 100% [47°C/84%]  7.59 I/s 14.9(14.9)Sol/s 101(106.0)W clk=1847MHz mclk=5005MHz Sol/W=0.14
 2>P104-100     ` 100% [46°C/82%]  9.20 I/s 17.9(17.9)Sol/s 120(134.3)W clk=1898MHz mclk=5005MHz Sol/W=0.13
 3>P104-100     ` 100% [57°C/100%]*14.01 I/s 29.3(29.3)Sol/s 150(157.6)W clk=1847MHz mclk=5005MHz Sol/W=0.19
 4>P106-100     ` 100% [56°C/100%]*10.21 I/s 19.7(19.7)Sol/s 117(107.8)W clk=1847MHz mclk=4006MHz Sol/W=0.18
 5>P104-100     ` 100% [60°C/100%] 13.96 I/s 27.8(27.8)Sol/s 155(164.5)W clk=1847MHz mclk=5005MHz Sol/W=0.17
 6>P106-100     ` 100% [56°C/100%]*10.43 I/s 21.0(21.0)Sol/s  99(105.7)W clk=1860MHz mclk=4006MHz Sol/W=0.20
 7>P104-100     ` 100% [51°C/94%] 14.06 I/s 27.9(27.9)Sol/s 149(154.3)W clk=1860MHz mclk=5005MHz Sol/W=0.18
 8>P104-100     ` 100% [59°C/100%]*14.13 I/s 27.4(27.4)Sol/s 180(163.1)W clk=1860MHz mclk=5005MHz Sol/W=0.17

--mode=3 --oc1

Code:

[ 0d 0h 3m09s] 197(197.3)Sol/s 1213(1193.5)W
 0>P104-100     ` 100% [45°C/80%]  9.22 I/s 18.5(18.5)Sol/s 115(130.8)W clk=1898MHz mclk=5005MHz Sol/W=0.14
 1>P104-100     ` 100% [47°C/84%]  7.75 I/s 14.8(14.8)Sol/s 140(113.9)W clk=1873MHz mclk=5005MHz Sol/W=0.13
 2>P104-100     ` 100% [46°C/82%]  9.22 I/s 18.3(18.3)Sol/s 129(130.9)W clk=1911MHz mclk=5005MHz Sol/W=0.14
 3>P104-100     ` 100% [57°C/100%] 13.76 I/s 27.2(27.2)Sol/s 155(155.9)W clk=1860MHz mclk=5005MHz Sol/W=0.17
 4>P106-100     ` 100% [54°C/98%]  9.03 I/s 18.3(18.3)Sol/s  95( 96.6)W clk=1873MHz mclk=4006MHz Sol/W=0.19
 5>P104-100     ` 100% [61°C/100%] 13.68 I/s 26.7(26.7)Sol/s 144(149.1)W clk=1784MHz mclk=5005MHz Sol/W=0.18
 6>P106-100     ` 100% [54°C/98%]  9.06 I/s 18.3(18.3)Sol/s 109(100.4)W clk=1898MHz mclk=4006MHz Sol/W=0.18
 7>P104-100     ` 100% [51°C/90%] 13.78 I/s 27.5(27.5)Sol/s 170(155.5)W clk=1822MHz mclk=5005MHz Sol/W=0.18
 8>P104-100     ` 100% [60°C/100%]*13.80 I/s 27.5(27.5)Sol/s 156(160.5)W clk=1847MHz mclk=5005MHz Sol/W=0.17

--mode=3 --oc2

Code:

[ 0d 0h 3m09s] 196(195.9)Sol/s 1228(1202.5)W
 0>P104-100     ` 100% [44°C/77%]  9.22 I/s 18.6(18.6)Sol/s 111(121.1)W clk=1873MHz mclk=5005MHz Sol/W=0.15
 1>P104-100     ` 100% [46°C/82%]  7.73 I/s 15.3(15.3)Sol/s 107(114.0)W clk=1885MHz mclk=5005MHz Sol/W=0.13
 2>P104-100     ` 100% [45°C/77%]  9.20 I/s 17.5(17.5)Sol/s 164(134.3)W clk=1885MHz mclk=5005MHz Sol/W=0.13
 3>P104-100     ` 100% [56°C/100%] 13.79 I/s 27.7(27.7)Sol/s 151(152.4)W clk=1822MHz mclk=5005MHz Sol/W=0.18
 4>P106-100     ` 100% [54°C/96%]  9.04 I/s 17.9(17.9)Sol/s 106(101.7)W clk=1873MHz mclk=4006MHz Sol/W=0.18
 5>P104-100     ` 100% [60°C/100%]*13.72 I/s 26.6(26.6)Sol/s 169(168.3)W clk=1860MHz mclk=5005MHz Sol/W=0.16
 6>P106-100     ` 100% [53°C/98%]  9.07 I/s 18.1(18.1)Sol/s  96( 98.7)W clk=1898MHz mclk=4006MHz Sol/W=0.18
 7>P104-100     ` 100% [49°C/88%]*13.77 I/s 26.9(26.9)Sol/s 159(150.5)W clk=1835MHz mclk=5005MHz Sol/W=0.18
 8>P104-100     ` 100% [58°C/100%] 13.81 I/s 27.7(27.7)Sol/s 163(161.5)W clk=1885MHz mclk=5005MHz Sol/W=0.17

--oc2

Code:

[ 0d 0h 3m00s] 204(203.7)Sol/s 1233(1231.7)W
 0>P104-100     ` 100% [46°C/80%]  9.22 I/s 17.3(17.3)Sol/s 150(137.2)W clk=1911MHz mclk=5005MHz Sol/W=0.13
 1>P104-100     ` 100% [48°C/86%]  7.62 I/s 14.9(14.9)Sol/s 112(121.2)W clk=1873MHz mclk=5005MHz Sol/W=0.12
 2>P104-100     ` 100% [47°C/84%]  9.19 I/s 18.0(18.0)Sol/s 143(130.7)W clk=1898MHz mclk=5005MHz Sol/W=0.14
 3>P104-100     ` 100% [58°C/100%] 14.01 I/s 28.2(28.2)Sol/s 150(162.3)W clk=1797MHz mclk=5005MHz Sol/W=0.17
 4>P106-100     ` 100% [56°C/100%] 10.16 I/s 20.1(20.1)Sol/s  93(101.8)W clk=1860MHz mclk=4006MHz Sol/W=0.20
 5>P104-100     ` 100% [61°C/100%] 13.92 I/s 28.1(28.1)Sol/s 168(152.8)W clk=1784MHz mclk=5005MHz Sol/W=0.18
 6>P106-100     ` 100% [56°C/100%] 10.30 I/s 20.1(20.1)Sol/s 107(109.8)W clk=1873MHz mclk=4006MHz Sol/W=0.18
 7>P104-100     ` 100% [53°C/96%]*14.05 I/s 28.3(28.3)Sol/s 154(157.3)W clk=1809MHz mclk=5005MHz Sol/W=0.18
 8>P104-100     ` 100% [60°C/100%] 14.09 I/s 28.4(28.4)Sol/s 156(158.5)W clk=1822MHz mclk=5005MHz Sol/W=0.18

with exactly the same OC settings GMiner performs 25 H/s on P106-100s and 41 H/s on P104-100s.

miniZ

member

Activity: 690

Merit: 17

Quote from: darkneorus on August 25, 2019, 11:53:45 AM

Windows 10, CUDA 10. would be happy to test it

Hi darkneorus,
You can download the beta file @ https://miniz.ch/download/

Test it with --oc1, it will load mode 3.

Do you see any improvement?

Thanks for your collaboration.

Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: 2h4u on August 25, 2019, 11:33:40 AM

Quote from: miniZ on August 25, 2019, 11:05:24 AM

Hi 2h4u,
thanks for the feedback.
We will keep working on this.
Do you have any improvemement with the segmentation faults?

Yep looks like they have gone.

Hi 2h4u,
that is nice.

Quote from: 2h4u on August 25, 2019, 11:38:21 AM

I say it would be great if miniZ does it itself. 5 mins without OC, 5 mins on OC1, next 5 mins on OC2, then check the difference, set best and report to user.
You said that I/s'es stabilize kinda quick, didn't you?

We also thought about it and will see if there is a way.

Cheers

darkneorus

jr. member

Activity: 238

Merit: 3

Quote from: miniZ on August 25, 2019, 11:53:03 AM

Quote from: darkneorus on August 25, 2019, 11:20:46 AM

should --mode set the first digit here?

if so then it's not working for P104-100s. the log above is miniZ launched with --mode=4. I don't really know what it means, you did not describe this option, but P104-100s always use the digit "7" there, no matter what mode value I set.

hope it helps

Hi darkneorus,
Thanks. This helps!
Mode can be used to tune the best solver for the card. It looks like there was only one (and not the best) solver for your card (P104).
Which OS are you using? Maybe we can build a beta version for you to try?
Cheers

Windows 10, CUDA 10. would be happy to test it

miniZ

member

Activity: 690

Merit: 17

Quote from: darkneorus on August 25, 2019, 11:20:46 AM

should --mode set the first digit here?

if so then it's not working for P104-100s. the log above is miniZ launched with --mode=4. I don't really know what it means, you did not describe this option, but P104-100s always use the digit "7" there, no matter what mode value I set.

hope it helps

Hi darkneorus,
Thanks. This helps!
Mode can be used to tune the best solver for the card. It looks like there was only one (and not the best) solver for your card (P104).
Which OS are you using? Maybe we can build a beta version for you to try?
Cheers

2h4u

newbie

Activity: 55

Merit: 0

Quote from: miniZ on August 25, 2019, 11:04:25 AM

However, the best is to try both because the OC settings used (core and mem clock) have big impact on this.

I say it would be great if miniZ does it itself. 5 mins without OC, 5 mins on OC1, next 5 mins on OC2, then check the difference, set best and report to user.
You said that I/s'es stabilize kinda quick, didn't you?

2h4u

newbie

Activity: 55

Merit: 0

Quote from: miniZ on August 25, 2019, 11:05:24 AM

Hi 2h4u,
thanks for the feedback.
We will keep working on this.
Do you have any improvemement with the segmentation faults?

Yep looks like they have gone.

darkneorus

jr. member

Activity: 238

Merit: 3

Quote from: miniZ on August 24, 2019, 07:14:04 PM

Quote from: darkneorus on August 24, 2019, 06:41:47 PM

Quote from: miniZ on August 24, 2019, 06:05:49 PM

Quote from: darkneorus on August 24, 2019, 05:23:23 PM

something is broken with 150,5,3 on P104-100 GPUs. the hashrate is low, 20-30 Sol/s max, expected 40+.

Hi darkneorus,
if you try adding --oc1 to your command line do you get the same hashrate?
Thanks.
Cheers

yup. way lower than it should be

Hi darkneorus,
thanks for the information.

We don't have your card so, for the moment we cannot test it.
Maybe you can help us understand this GPU

You could try test other mode values by adding --mode=x, with x=[1,2,4,6,8,9] to your command line, and see if you can get a better hashrate.

Let us know which performs better.
Cheers

no success.

should --mode set the first digit here?

Quote

miniZ<150,5,3>[4:1:00:5998]: Selecting GPU#4[4] P106-100
miniZ<150,5,3>[7:1:00:3933]: Selecting GPU#0[0] P104-100
miniZ<150,5,3>[7:1:00:3933]: Selecting GPU#2[2] P104-100
miniZ<150,5,3>[7:1:00:3933]: Selecting GPU#3[3] P104-100
miniZ<150,5,3>[7:1:00:3933]: Selecting GPU#7[7] P104-100
miniZ<150,5,3>[7:1:00:3933]: Selecting GPU#8[8] P104-100
miniZ<150,5,3>[7:1:00:3933]: Selecting GPU#1[1] P104-100
miniZ<150,5,3>[4:1:00:5998]: Selecting GPU#6[6] P106-100
miniZ<150,5,3>[7:1:00:3933]: Selecting GPU#5[5] P104-100

if so then it's not working for P104-100s. the log above is miniZ launched with --mode=4. I don't really know what it means, you did not describe this option, but P104-100s always use the digit "7" there, no matter what mode value I set.

hope it helps

darkneorus

jr. member

Activity: 238

Merit: 3

Quote from: Dr_Victor on August 25, 2019, 10:31:48 AM

Quote from: darkneorus on August 24, 2019, 05:23:23 PM

something is broken with 150,5,3 on P104-100 GPUs. the hashrate is low, 20-30 Sol/s max, expected 40+.

Feel free to add +500 for memory, use the pill and enjoy 40+

pill does nothing for P104-100 GPUs. it's useful only for 1080 series.
GMiner shows 40-42 H/s with mem at stock clock.
something is clearly wrong with 150,5,3 implementation.
on 150,5 algo GMiner and miniZ hashrates were about equal, on 150,5,3 miniZ is far behind.

miniZ

member

Activity: 690

Merit: 17

Quote from: Dr_Victor on August 25, 2019, 10:31:48 AM

Quote from: darkneorus on August 24, 2019, 05:23:23 PM

something is broken with 150,5,3 on P104-100 GPUs. the hashrate is low, 20-30 Sol/s max, expected 40+.

Feel free to add +500 for memory, use the pill and enjoy 40+

Hi Dr_Victor,
thanks for the suggestion!
Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: Divinity666 on August 25, 2019, 09:05:36 AM

So, v1.5q2 seems to have fix issue with low h\r in 144.5 on my 1080s. After <5min run I even see higher I\s, though by 0.15% Grin

But at least it is on pair with v1.3

Quote from: invizion on August 24, 2019, 09:36:36 PM

Question, does OC1/OC2 work for all algos? What determines "power saving" to use OC1 vs OC2?

oc1 gives you max power saving and efficiency.

Hi Divinity666,
thanks for letting us know that the regression is reverted!

Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: lordcryptocurrency on August 25, 2019, 01:00:16 AM

Hi, I recently did an update on windows 10 and mining is no longer stable, windows automatically restarts.
I did not have this problem, started to appear after windows 10 upgrade

I have 2 1080TI, my settings are:
Power: 65
Core: +75
Memory: -475

All help is welcome, in advance I thank

Hi lordcryptocurrency,
that is unfortunate. Maybe the driver was updated and you need to adjust your OC values.

We did try something different from your values. Maybe worth giving it a try.
pl=165W (more or less your 65%) and core=+100, memory=+300.
With --oc2 we got ~49 Sol/s.

Hope this helps.
Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: 2h4u on August 25, 2019, 12:44:13 AM

Quote from: miniZ on August 24, 2019, 06:32:52 PM

* (Hopefully!) Fixed slow start for some GPUs/rigs.

Hmm... With rhminer as 2nd miner:

2:33 from start and still not all GPUs started:

Hi 2h4u,
thanks for the feedback.
We will keep working on this.
Do you have any improvemement with the segmentation faults?
Cheers

miniZ

member

Activity: 690

Merit: 17

Quote from: invizion on August 24, 2019, 09:36:36 PM

v1.5q2 fixed the issue thanks!

Question, does OC1/OC2 work for all algos? What determines "power saving" to use OC1 vs OC2?

Hi invizion,
thanks! Good to know!
oc1/oc2 is implemented for almost all GPUs and algorithms (not yet 96,5).
For each algorithm, there are still a few GPUs that may not have the oc implemented.

Generally, oc1 performs better when we lower power from the stock settings level, oc2 when it is increased.
However, the best is to try both because the OC settings used (core and mem clock) have big impact on this.

We plan to work a bit more on this and extend oc to all gpus and algos.
We realize that also it will be nice to have a option to show which gpus have them implemented. We will work on this

Cheers

Dr_Victor

sr. member

Activity: 954

Merit: 250

Quote from: darkneorus on August 24, 2019, 05:23:23 PM

something is broken with 150,5,3 on P104-100 GPUs. the hashrate is low, 20-30 Sol/s max, expected 40+.

Feel free to add +500 for memory, use the pill and enjoy 40+

Divinity666

jr. member

Activity: 312

Merit: 2

So, v1.5q2 seems to have fix issue with low h\r in 144.5 on my 1080s. After <5min run I even see higher I\s, though by 0.15% Grin

But at least it is on pair with v1.3

Quote from: invizion on August 24, 2019, 09:36:36 PM

Question, does OC1/OC2 work for all algos? What determines "power saving" to use OC1 vs OC2?

oc1 gives you max power saving and efficiency.

lordcryptocurrency

newbie

Activity: 1

Merit: 0

Hi, I recently did an update on windows 10 and mining is no longer stable, windows automatically restarts.
I did not have this problem, started to appear after windows 10 upgrade

I have 2 1080TI, my settings are:
Power: 65
Core: +75
Memory: -475

All help is welcome, in advance I thank

Topic: miniZ v2.0a Equihash 144,5 125,4 210,9 150,5 192,7 BeamHash3 ProgPoW Ethash CFX - page 40. (Read 60004 times)