Topic: [XMR] JCE Miner Cryptonight/forks, now with GPU! - page 76. (Read 90841 times)

Unfortunately, it still won't work for my win 7-64, blockchain drivers, RX 470-8GB with integrated graphics on MB.  Now the output says:

0.32 vs 0.31c

7x Vega 56 rig on celeron 3855U (very slow cpu) with adrenalin 18.6.1. drivers

14210h/s compared to 14380 on 0.31c
in the beginning (first few minutes) got 5-6 invalid shares but after that no bad shares
super fast compiling, miner started in like 30 seconds

i will keep it running for hashrate to stabilize and report after few hours
 

I give 30min 0.32 on vega i get ~10h/s(2145h/s cryptonight v7) more compare to 0.31f but 8 Rejected Shares (24%) : Low Dificulty 
on rx550/460 it runs ~5 h/s slower (cryptonight v7)
driver 18.5.1

Cheers for the awesome cpu miner   

Thanks

It's true i log very few overall info, and nothing about bad shares (well, they are displayed at the moment they are found, but not accumulated, so you don't know how many bad share you had). The yellow log may have one extra line for bad shares.

Online is the very experimental 0.32

It uses a completely new OpenCL generator that is blazing fast (the generator, not the mining code) : compile now takes about one second but it's very possible it adds regressions, since everything is new. I tried to keed all perfs at max (even on RX550) but i cannot be sure. It was stable on my rig, but i haven't a lot of time to test.

It also introduces the -g parameter to filter the GPUs to mine on. Example: -g 0,2 means: don't mine on a GPU other than 0 or 2.
The auto-config should be a little better.

Hi everybody,

now already some more time has passed with testing JCE miner and all I can say this is hands down the best and most reliable miner at least for CN7 and Vega cards.

All my machines had uptime of 200+ hours before  shut down yesterday to change the pool. There was 0 cards with hash drop, and I got 10 total invalid shares (GPU issue, nothing with miner)

I am still on 031C version because I am far away from home so I didn´t bother with remote changing

JCE, i guess now there is already many users and miners testing your miner with all kinds of suggestions, questions etc, don´t get overwhelmed and give up, just keep developing, i hope you receive now decent money from the miner fees to support yourself/family

In the beginning I was suggesting features like watchdog but considering I had 0 issues I don´t even know if this is necessary

There is few things I would like to see if possible to implement:
1. GPU displaying its BUS Id
2. Invalid shares statistic in Results, for example instead of Valid shares found 865, just write valid shares 865/888
3. Statistic of invalid shares in Results, for example if there is 15 invalid shares, to list how many was GPU fault and network fault.

Maybe you already did this in the later versions but I did not upgrade yet

I hope you will make soon enough money to get yourself a Vega card for testing

Yea those numbers are for CN-Saber which is a bit heavier algo than CN-Heavy too.  Those are 944,1 thread on 4g cards and 832,2 threads on 8gb cards.

Use config file in wich describe only your 3 gpu's. Like this:

"gpu_threads_conf" :
[
     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 8, "delta" : 8, "epsilon" : 8, "zeta" : 8, "index" : 0, "multi_hash":832 },
     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 8, "delta" : 8, "epsilon" : 8, "zeta" : 8, "index" : 0, "multi_hash":832 },
     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 8, "delta" : 8, "epsilon" : 8, "zeta" : 8, "index" : 1, "multi_hash":832 },
     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 8, "delta" : 8, "epsilon" : 8, "zeta" : 8, "index" : 1, "multi_hash":832 },
     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 8, "delta" : 8, "epsilon" : 8, "zeta" : 8, "index" : 2, "multi_hash":832 },
     { "mode" : "GPU", "worksize" : 8, "alpha" : 64, "beta" : 8, "gamma" : 8, "delta" : 8, "epsilon" : 8, "zeta" : 8, "index" : 2, "multi_hash":832 },
]

Of course not. I thought you wrote about v7 algo. On heavy algo max is 944 for 8Gb. On higher values speed is drops due to lack of vmemory.
For heavy algo on 8Gb best multihash is 832 (and maybe 864) and dual threads.

any have a config? for mining only 1 / 2 /3 gpu
i am try -d command not run

I had the settings swapped in my post for 4gb and 8gb cards.  1836 for CN-Heavy/Saber dual threads.  Will that even work?

Yes

use the "index" command in gpu_threads_conf section
gpu index numbers start at 0

Baz

i want mining only on 3 from 5 cards
is this possible?

Waiting for this possibility!

I don't think that autoconfig option is so important... Often every card needs to fine tune...

Still waiting heavy algo optimization to stop use SRB and go to JCE miner. Until use JCE only with normalv7 algo and based on it coins...

@JCE-Miner

Any ETA for Linux GPU/CPU version ?

On 8Gb cards 944 isn't best multi_hash. Try to use 1836.

I Advise to use double thread on all gpu, even the 4gb one. i always get better perf with double thread, even on low-mem cards.

good idea about the gpu select, i'll add param -g similar to Claymore -di
-g 0,2 : autoconfig only gpu 0 and 2.

@JCE-Miner keep up the good work.  I get the below on 6 x RX 580 4gb, 2 x RX 570 4gb and 1 x RX 570 8gb on CN-Saber.

4gb card

8gb card

can you add the ability to select gpus in --auto mode?