[JCE]Fast & stable CN/v8/Heavy/Tube/XHV miner, CPU+GPU, Vega56 1800+ RX580 1200+

[Pennywis3]

Hey JCE

I figured out a manual fix to get those rx 580 cards to hash at max speeds on Heavy and be stable, any time i see one that has lost the max speed, i pause/unpause 4 times each (8 button pushes) in quick succession and the card comes back to maximum speed.  May be something like this can be coded into the miner?

Yeah, I've been doing this since 0.33B5 and it works on all my rigs, except sometimes it takes more than 4x pause/un-pause. This also works on SRB miner, but takes a bit more time to achieve.

In the end you get all the card hashing at max speed this way.

[Iamtutut]

Mining HAVEN (heavy) for >11H

4 RX570 4G with bios mod.

Display GPU: 1250C/2040M. Intensity set to 272.
Other GPUs: 1230C/2100M; 1240C/2020M; 1230C/2030M. Intensity set to 464.

Max HR: 3196H/s
Reported HR: 3290H/s (yep, more than the theoratical perf).
Share difficulty set by the pool: 169K

HR at the pool:
- 2 hours: 3,38KH/s
- 24 hours: 1,58KH/s. That's consistent to the fact I've been mining for less than 12Hs

[ivomm]

Mining heavy/haven:
280x give stable 430 h/s.
390 are very unstable. Cards report between 500 and 800 whatever intensity I chose. Most cards gravitate towards 640 h/s, but I have some that can't go beyond 540 h/s with nearly 40% bad shares even if not moded. For comparison with xmrig all 390 cards are very stable with 660 h/s and none bad shares. On the other hand, 280x are not even working with xmrig.

[JCE-Miner]

yep, more than the theoratical perf

On a 11h run with a high diff, this is very normal.

with nearly 40% bad shares even if not moded

yeah my hybrid on the b-xyz is pretty bad on Hawaii, i want to add an option in next GPU version to enable or not the new hybrid. Disable will typically lower the peak hashrate by 1% but provide zero bad shares.

[Iamtutut]

Mining HAVEN (heavy) for >11H

4 RX570 4G with bios mod.

Display GPU: 1250C/2040M. Intensity set to 272.
Other GPUs: 1230C/2100M; 1240C/2020M; 1230C/2030M. Intensity set to 464.

Max HR: 3196H/s
Reported HR: 3290H/s (yep, more than the theoratical perf).
Share difficulty set by the pool: 169K

HR at the pool:
- 2 hours: 3,38KH/s
- 24 hours: 1,58KH/s. That's consistent to the fact I've been mining for less than 12Hs

24 hours mining with same settings:
- Max HR is 3234,55H/s
- Reported average HR by the miner is 3150H/s
- Pool says 3150 H/s

Power draw at the wall: 405-410W.

[UnclWish]

Здpaвcтвyйтe. Бyдeт paбoтaть нa hd7850 c 1гб нa v7 370-400h кoпaли.

Пoчти тaкжe дoлжнa кoпaть и нa v8.

[UnclWish]

JCE, any news about new version? Heavy algo speed stabilization and lowering power draw on v8 is very needed!

[JCE-Miner]

a new version? sure, here's the

0.33g CPU Linux re-release

Fixed JSON timer
FreeHaven coin correctly defaults to FreeHaven fork

then I go back to GPU sync fixes.

[MadSilence]

We need this awesome miner (GPU version) for Linux NOW!=)  haha I have hive rig w8ing to get this Sw on them

[livada]

a new version? sure, here's the

0.33g CPU Linux re-release

Fixed JSON timer
FreeHaven coin correctly defaults to FreeHaven fork

then I go back to GPU sync fixes.

ok but where is GPU version with lower fee?

Srb miner have better HR and smaller devfee on all algo (no in v8)

[UnclWish]

a new version? sure, here's the

0.33g CPU Linux re-release

Fixed JSON timer
FreeHaven coin correctly defaults to FreeHaven fork

then I go back to GPU sync fixes.

ok but where is GPU version with lower fee?

Srb miner have better HR and smaller devfee on all algo (no in v8)

JCE x-y version have 0,9% devfee on all GPU algo's.

Hmmm, looks llike doctor83 can do that:
"Ok, so looks like i finally resolved the ellesmere 8gb heavy algos thing.
Now they easily hash over 1.1khs.

New version in a few days"

[GKumaran]

Hey Dev,

Code:

Abort was called at 767 line in file:
D:\qb\workspace\19992\src\vpg-compute-neo\runtime\os_interface\windows\wddm.cpp

I have no clue !!

Cards : Vega 64
Windows : 10
Driver : 18.6.1
JCE : b6 version

[UnclWish]

Hey Dev,

Code:

Abort was called at 767 line in file:
D:\qb\workspace\19992\src\vpg-compute-neo\runtime\os_interface\windows\wddm.cpp

I have no clue !!

Cards : Vega 64
Windows : 10
Driver : 18.6.1
JCE : b6 version

Use directory with miner on path only with latin symbols...

[GKumaran]

Hey Dev,

Code:

Abort was called at 767 line in file:
D:\qb\workspace\19992\src\vpg-compute-neo\runtime\os_interface\windows\wddm.cpp

I have no clue !!

Cards : Vega 64
Windows : 10
Driver : 18.6.1
JCE : b6 version

Use directory with miner on path only with latin symbols...

I have no clue what latin is?
The path shown above isnt the path of miner.
The miner is plain on the desktop
My PC doesn't even have a "d:"

[JCE-Miner]

Please, take a look at the doc:

https://github.com/jceminer/cn_gpu_miner/blob/master/README.md#troubleshooting

Q. I get error D:\qb\workspace\19992\src\vpg-compute-neo\runtime\os_interface\windows\wddm.cpp
Intel GPUs are now enabled among AMD GPUs, ensure you're not applying a previous AMD configuration to an Intel GPU

Online is the 0.33b7 GPU

This is a compilation of various fixes and optims from the b-z and the 0.33g CPU. It should be slightly faster on most GPUs and supports CN-Dark and CN-FreeHaven. And it officializes the revert to low-fee 0.9% (which are no longer in the doc). The b-z is hence removed.

Those who experienced bad perfs on old cards (HD7000, 6000) take a look it may be better. I now have an average of 426 h/s on my HD7950 on Heavy while on b-z the peak was 423.

It does not contain the hashrate instability fix yet, I still have to test it more. Looks like doktor faces the same problem and found a solution, probably the same too. So same answer: more tests to do, release planned in a few days.

[JCE-Miner]

You're welcome

I have no clue what latin is?

His advice is good too: when stored in a directory with exotic characters (russian, asian...) JCE may fail to start. Latin is the technical name given to the normal A-to-Z letters, and by extension, to the 0-to-9 digits. Sometimes it's called 8.3, like the old DOS standard for filenames.
So c:\computer is a correct name, c:\pacчeтa is not.

@unclwish: i'm getting pretty good results on the new-new-new sync, an experimental release is possible in less than a day. I give up my hybrid option, i'll make it automatic (enabled on HD7000/6000, disabled otherwise)

[UnclWish]

@unclwish: i'm getting pretty good results on the new-new-new sync, an experimental release is possible in less than a day. I give up my hybrid option, i'll make it automatic (enabled on HD7000/6000, disabled otherwise)

Excellent! I'm waiting to test it!
And what about v8 high power draw? You wrote that you have thoughts to lower it...

[JCE-Miner]

sure i'll make a new experimental for both of you, the sync for Unclwish and a param --no-monitor to disable the gpu monitoring (similar to --no-adl in xmrstak)

the power draw on v8 is an order of magnitude more complex to fix than the sync, i admit TeamRed will be very hard to beat, but i think i know how.

[UnclWish]

sure i'll make a new experimental for both of you, the sync for Unclwish and a param --no-monitor to disable the gpu monitoring (similar to --no-adl in xmrstak)

the power draw on v8 is an order of magnitude more complex to fix than the sync, i admit TeamRed will be very hard to beat, but i think i know how.

Would you be able to do it on b4? b-x, b-y and b-z are more prone to the crashing issue. On b4 I can mine for about 12h. On b-z sometimes can manage calculate only one share accepted (about 5min) before the crash.

Did you tried b7 version?

[Iamtutut]

Here we go for B7. Start is better than B6, less frequency swings.

I'll update tomorrow.

With B6 I had 3300H/s average for 24Hs (heavy / Haven). 4 RX574.