Awesome, thank you . Now i can put the settings in the first post so everyone can see it.

Thanks, could you tell me the hashrate too?

Can someone with 550 and 560 cards tell me which config is the best for V7 and Heavy ?

Added info in tips section here on how to turn off crossfire and switch to compute mode.

18.6.1 is good for CN V7 on Vega cards, gives the most hash. While is worse on heavy than BC for example.
Amd stuff.. It is nearly impossible to make a miner that works good on every driver.

So that's why people shouldn't switch drivers every time a new one comes out, just stick to the one that's working.

you got an error, there is a * before "gpu_conf".

if driver supports 13 gpu then SRBMiner supports 13 gpu. There are guys here using it with 13 gpu's.

I missed your previous post, sorry. I will look into it.

edit : it's cryptonight v7, so you can use 'normalv7' to mine it.

If you take 1.6.4 and install 18.6.1 and let the miner build the binaries it would work good. The problem here is the 18.7.1 drivers.

i did not touch cl kernels from 1.6.3 to 1.6.4, so probably you built on one driver , now with another, and maybe the compiler unrolls differently, that would explain the increased size.

yes,
Accepted(Stale)/Total shares: 42(1)/42 (100.0%)

You can use nicehash+heavy way before 1.6.4, i don't know where do you get this info from. Stale shares are not sent when using nicehash cause nicehash rejects ALL stale shares, so no point in sending them. 10-11% is much i agree.

Don't have to rebuild kernels after changing drivers, miner will work, but i would suggest you to rebuild them to be sure you are totally compatible with the drivers.

If the pool pays for stale shares then you will benefit from using --sendallstales, if it's not accepting stales you will get that stale share rejected. So there is nothing for you to loose.
What can happen is if you are mining on low difficulty and you find a lot of shares / minute, and your pool does not accept stale shares, and a job switch occurs and miner sends all those shares from the previous job, and the pool rejects them, and because you had a lot of rejects in short time, it can ban you for a time. This is the worse that can happen, but on pools that have vardiff set OK, and you find approx 6-10 shares / minute you can only win with this option turned on.

You probably got a lot stale shares, download latest 1.6.4 v and use --sendallstales parameter in bat, it will probably increase your effective hashrate

Nope, no downsides, except it won't cleanup some leftovers, but who cares, the OS will 

You probably downloaded in those few minutes between when i uploaded it and updated the post here

That is normal because it shows you the number of GPU's, not threads.
When miner starts you can see on top of window something like :

GPU0: Radeon RX Vega [gfx900] [8176 MB][Intensity 120.0][W: 8][T: 2][K: 1][BUS: 1]

The bolded part means : Threads: 2, so its all ok

Do you see now what's wrong? I bolded it for you

config.txt


start.bat


Works. Just tried it.

V1.6.4
- Added possibility/support to set GPU'S with cmd parameters directly in start.bat
- Added parameter --sendallstales , which enables sending of every stale share
- Minor bug fixes

+ So command line setup of GPU'S is finally here. Here are the parameters that can be used:

--ccryptonighttype value (algo to use)
--cgpuid value (gpu id, comma separated values, use --listdevices to see available)
--cgpuintensity value (gpu intensity, comma separated values)
--cgputhreads value (number of gpu threads, comma separated values)
--cgpuworksize value (gpu worksize, comma separated values)
--cgputargettemperature value (gpu temperature, comma separated values)
--cgputargetfanspeed value (gpu fan speed in RPM, comma separated values)
--cgpuadltype value (gpu adl to use (1 or 2), comma separated values)
--cgpukernel value (gpu kernel to use (1 or 2), comma separated values)

Parameters : --cgpuid, --cgpuintensity and --cgputhreads MUST BE USED, the others are optional
If you want to set everything in cmd line, you still need to have an empty config.txt file (which contains just : {}, or any other parameter like api etc etc )


Ex. for GPU and POOL setup from CMD:

Use 1 GPU with id 0 , intensity 120, 2 threads on algo cryptonight v7 on nanopool:
SRBMiner-CN.exe --ccryptonighttype normalv7 --cgpuid 0 --cgpuintensity 120 --cgputhreads 2 --cpool xmr-eu1.nanopool.org:14444 --cwallet 4A5hJyu2FvuM2azexYssHW2odrNCNWVqLLmzCowrA57xGJLNufXfzVgcMpAy3YWpzZSAPALhVH4Ed7x o6RZYyw2bUtbm12g.donation


Use 5 GPUS with id 0,1,2,3,4 , intensities 56,56,55,58,55, 2 threads for each GPU, on algo cryptonight v7 on nanopool:
SRBMiner-CN.exe --ccryptonighttype normalv7 --cgpuid 0,1,2,3,4 --cgpuintensity 56,56,55,58,55 --cgputhreads 2,2,2,2,2 --cpool xmr-eu1.nanopool.org:14444 --cwallet 4A5hJyu2FvuM2azexYssHW2odrNCNWVqLLmzCowrA57xGJLNufXfzVgcMpAy3YWpzZSAPALhVH4Ed7x o6RZYyw2bUtbm12g.donation


+ --sendallstales is now a default parameter in start.bat
When a new job arrives to the miner, and a result from a previous job is sent to the pool, this is called a stale share. A low percent of pools accepts stale shares, they generally reject the share with an 'Invalid job id' error.
Now in SRB i made this logic : if a new job arrives, a share for a previous job is not sent IF it's not sent in a period less than a second since the new job arrived. It looks like some pools accept stale shares even for a few seconds after a new job was dispatched.
This parameter turns off the 1 second rule, and sends every stale share there is. This could potentially lead to an increase in effective share rate, but also you can get a lot more rejected shares, it all depends on the pool you are using.

Oh yeah, you can also get shares rejected when a pool switch/reconnect/job timeout happens, in situations where the pool difficulty is low and you produce a lot of shares, but in the process of doing the calculations (which can take 1-2 sec depending on your hardware) the connection to the pool drops, these share are tried to be sent while you are not logged on the pool, so you will get now a nice warning that a network error has occured and your shares wont be sent.