Vega56 users, please try v1.6.7 and a heavy algo (haven, bittube, heavy, italo) with :

intensity : 61
worksize : 9
threads : 2

Do you get better results than earlier ?

So what was the reason, if you figured it out ?

There is no such a thing as global kernel setting in the config, so that's why it isn't working
Kernel parameter can be set only in gpu_conf.
You should try kernel 2 on the 390x also, the 3 thread setup, because kernel 1 is now really more for 4xx 5xx and vegas.

you can find all these hefty parameters on the first page or in the readme 

To manually set up video cards, you must create a "gpu_conf" array in the config.txt file.

Example :

"gpu_conf" :
[
   { "id" : 0, "intensity" : 0, "worksize" : 8, "threads" : 2},
   { "id" : 1, "intensity" : 0, "worksize" : 8, "threads" : 2}
]

Set in config.txt

"reboot_script" : filename to a batch file in miner directory, if set it turns off built in miner reset procedure on gpu failure, and instead runs this script

If you are using the latest 1.6.7 version then logging is enabled by default, every time you start the miner it creates a .txt file.

Start miner and let it run a minute or two, then paste here the latest .txt file created.

IDK neither because its the same code as V7, only that it uses half the iterations as normalV7.
Everything else is same.

Have you tried what i suggested earlier to use 112/16/2 ?

Paste here your config.txt
Also it would be very nice to get the log file pasted here

I fixed a small bug with the new --algoswitchmintime parameter, as it was reset on algo change.
If you are using algo switching please download 1.6.7 again.

Hi, maybe you should try to set intensity by hand, because the auto setup isnt perfect
Try with the value you get from auto - 1 for start.

So your config could look like this for example:


54 is here just an example, you should use the value you get from auto - 1

kernel2 is now for 'older' cards, so thats why it selects that one automatically.
Also i got info from someone with 390x cards that for v7 3 threads and intensity 30 works best,  i couldn't test that i dont have those cards.

Do you have 290x ?

you should be able to change to kernel 1, that setting is set per card like :

"gpu_conf" :
[
   {
     "id" : 0,
     "intensity" : 124,
     "worksize" : 8,
     "threads" : 2,
     "kernel" : 1
   }
]

Write down what steps you did, and what result did you get.
Also driver version, did you use soft powerplay tables, clocks etc.

Im not really planning it because i was totally focused on windows from the start, and way too much code has to be changed/adapted and that means also a lot of new bugs.
But hey, i won't say never, who knows...

V1.6.7

- Now providing binary kernels for popular algos on mainstream cards (Vega56&64&FE, RX470&480&570&580, RX550&560)
- Auto setup for a lot of cards improved, especially for RX550, RX560
- A lot of OCL optimizations, mainly on Heavy algos (Heavy, Haven, Bittubev2, Italo)
- Removed --sendallstales parameter because it is now the default behavior (to send all stale shares)
- When using algo switching mining time is now showing total mining time, not just for the specific algo
- Now on every miner start a new log file is created, in the format of date-time in normal mining mode
- Added parameter --algoswitchmintime
- Completely setup everything for algo switching on Monero Ocean pools, you just need to edit your algo hashrates, and change wallet in the example files

+ Ok, binary kernels. Yes, the .zip grew in size a lot, because i decided to provide binary kernels for a few algos, for popular video cards.
I had to do this because of the crappy new 18.7+ drivers, where the compiler created bad .srb files.
A lot of people don't follow the big letters from the first page that says use 18.3.4 or 18.6.1 drivers.
So now i provide binary kernels for popular algos for :


- Vega 56
- Vega 64
- Vega FE
- RX 470
- RX 480
- RX 570
- RX 580
- RX 550
- RX 560


If you have some problems with these pre-compiled files (maybe blockchain drivers don't like them), just delete the BinaryKernels directory, and the miner will compile them for you, just like it was before.
I had a problem on one of the 580 rigs on BC drivers while testing, that they could not handle intensity 56 anymore, but 55 was ok with the new kernel.
Deleted the pre-compiled kernels, miner created them on BC drivers and then cards could run again on intensity 56.

+ I did a lot of optimizations in OCL kernels, and in auto setup for a lot of cards, so you should try it out. Heavy algos on Vega56 and auto setup now give you more hash by 1-2-3%! (tested on 18.6.1 drivers)

+ The --sendallstales is removed and you don't have to use it anymore, because that is now the default setup

+ For those using the algo switching, a new parameter --algoswitchmintime was added, it is used on Monero Ocean pools, and it sets the minimum time that you want an algo to be mined. The default value is 600 (10 minutes), this means that the pool shouldn't switch to another algo for minimum 10 minutes. This parameter should be entered in seconds.

+ I created all files you need for algo switching on Monero Ocean pools, so you just have to edit them now. Put the hashrate of your rig for the specific algo in algos.txt and change the XMR wallet address in pools-algoswitch.txt

18.7 , 18.8 has a screwed compiler, so no shares are found when compiled on those drivers.
But kernels compiled on older versions can be used with those drivers.

you should use 18.3.4 or 18.6.1 drivers

Guys with FAST algo on Vega's and hashrate drop after some time, can you try :

Intensity: 112
Worksize: 16
Threads: 2

This will create a logfile with the current date and time, every time you start the miner.

Try adding "adl_type" 2 in gpu_conf like :

{ "id" : 0, "intensity" : 59, "worksize" : 8, "threads" : 2,  "target_fan_speed" : 3500, "adl_type" : 2 }

Maybe that will work for you.