Quote from: pandas_talks on April 21, 2014, 01:56:44 PM

btw, are we going to make a page with the settings for the most popular GPUs?

There already is this post:

Quote from: digitalixz on April 14, 2014, 09:04:16 AM

Hello Guys,

I can share some mining configs with you.

First you have to download Yacminer - https://github.com/Thirtybird/YACMiner/releases/download/3.5.0-yac2/yacminer-3.5.0-yac2-x86.zip
These configs works on systems with more than 4GB RAM.

Config for 290x - yacminer --scrypt-chacha --nfmin 4 --nfmax 30 --starttime 1389028879 --worksize 128 -g 1 --lookup-gap 3 --buffer-size 3532 -R 10496
Config for 290 - yacminer --scrypt-chacha --nfmin 4 --nfmax 30 --starttime 1389028879 --worksize 128 -g 1 --lookup-gap 3 --buffer-size 3420 -X 32
Config for 280x - yacminer --scrypt-chacha --nfmin 4 --nfmax 30 --starttime 1389028879 --worksize 128 -g 1 --lookup-gap 3 --buffer-size 2590 -R 7680
Config for 280 - yacminer --scrypt-chacha --nfmin 4 --nfmax 30 --starttime 1389028879 --worksize 128 -g 1 --lookup-gap 3 --buffer-size 2394 -X 32
Config for 270x - yacminer --scrypt-chacha --nfmin 4 --nfmax 30 --starttime 1389028879 --worksize 128 -g 1 --lookup-gap 3 --buffer-size 1648 -R 4864
Config for 7950 - yacminer --scrypt-chacha --nfmin 4 --nfmax 30 --starttime 1389028879 --worksize 128 -g 1 --lookup-gap 3 --buffer-size 2394 -X 32
Config for 7970 - yacminer --scrypt-chacha --nfmin 4 --nfmax 30 --starttime 1389028879 --worksize 128 -g 1 --lookup-gap 3 --buffer-size 2590 -R 7680

Tips on tuning your GPU to mine Microcoin
--buffer-size XXXX (-B XXXX):
This setting sets the size of the OpenCL buffer to allocate to each thread. This option overrides --thread-concurrency, and should be used instead of it if your miner supports it. This value should be based on the amount of available memory on your GPU. This setting is NOT affected by lookup-gap meaning that other tuning parameters that you want to test will not force you to recalculate your buffer-size. The only setting that will make you want to change this is the --gpu-threads (-g) setting. As buffer-size is the amount of memory to allocate per thread, if you increase the number of threads you are launching, you will want to decrease the buffer-size of each thread to ensure all of the memory resides in Dedicated GPU memory instead of rolling over to Dynamic.
As the NFactor for Microcoin increases, the amount of memory that is needed to perform a single calculation rises. The more memory that can be allocated per thread, the better the performance that typically can be achieved.
SUMMARY: adjust this value first - Run GPUz of HWInfo to monitor the amount of dedicated memory used on the card until it moves into the dynamic column, then reduce it back down. YACMiner will tell you at the start how much memory has been allocated per thread

--lookup-gap X:
This tunes a compromise between ram usage and performance. Performance peaks at a gap of 2, but increasing the gap may allow you to run at a higher rate as more shaders in the card get utilized. Setting lookup gap overrides the default of 4.
SUMMARY: Start at 2, and try 4 through 8. Adjust your intensity setting (-R/-X/-I) to higher values as the lookup-gap increases.

--worksize XXX (-w 256):
Has a minor effect, should be a multiple of 32 up to 256 maximum. This sets the smallest size of work being sent to the GPU, and on older cards and/or drivers affects the values allowed in Raw Intensity.
SUMMARY: Higher values are better for cards with higher shader counts and 7XXX series cards and newer. Lower values for older cards give finer ranges of Raw Intensity to maximize performance.

--rawintensity XXXX (-R XXXX) This setting allows for a much more preceise intensity setting and can recognize significant gains when mining Microcoin
Note: make sure to use much lower raw-intensity values when you increase cpu threads (-g).
The original Intensity setting is currently used to spawn GPU threads as a simple 2^value setting.

I:8 = 128 threads
I:8 = 256 threads
I:9 = 512 threads
I:10 = 1024 threads
I:11 = 2048 threads
I:12 = 4096 threads
I:13 = 8192 threads

Notice how the higher settings increase thread count tremendously.
It's now much easier to control thread intensity specifically and it potentially allows for a uniform way of setting the intensity on your system.

Valid values for raw-intensity are safest to be set as a multiple of --worksize. Newer drivers and/or cards may allow for adjustment in smaller increments - in some cases, this can be adjusted in in increments of 1.

--xintensity XX (-X XX)
When you have enough GPU memory allocated to fully utilize all shaders on your GPU, use this for intensity control as it will spawn threads as a multiple of how many shaders are on your GPU. This setting will be best starting at 4, and increases can be had by going up to 32 or as high as 128. Very large values for this may cause instability or driver crashes for very little gain - consider 128 to be the max.
Optional parameters for tuning:
--gpu-threads X (-g X):
Some people who have systems where the GPUs have large amounts of memory and low amounts of system memory will need to run multiple CPU threads so that all of the GPU memory can be allocated. This is because the OpenCL buffer is built in system memory and passed to the GPU. This defaults to 1, however values of 1, 2, 3 & 4 can all produce similar results.
SUMMARY: Run 1 unless your card has more than 2GB of memory and you cannot allocate all of that memory with just 1 thread.
Other available parameters that have better alternatives:
--thread-concurrency XXXXX:
This option may be depreciated in your mining software depreciated. It is still available for those people who like to do math on really high numbers that have to change when any other parameter needs to change. I strongly recommend utilizing the --buffer-size option instead. This setting adjusts the size of the OpenCL buffer to allocate to each thread. This value should be based on the amount of available memory on your GPU. This setting is affected by lookup-gap as well, utilizing the following formula for lookup-gaps of 1,2 & 4

MB per thread = thread-concurrency * 128 / (1024 * lookup-gap)
A good example of this: lookup-gap of 2, thread-concurrency of 8192 = 512 MB

As the NFactor for a Microcoin increases, the amount of memory that is needed to perform a single calculation rises. The more memory that can be allocated per thread, the better the performance that typically can be achieved.
SUMMARY: Use --buffer-size instead

--intensity XX (-I XX)
Just like in bitcoin mining, Microcoin mining takes an intensity, however the scale goes from 0 to 20 to mimic the "Aggression" used in mtrlt's reaper. The reason this is crucial is that too high an intensity can actually be disastrous with Microcoin because it can run out of memory and you will receive HW errors. High intensities start writing over the same ram and it is highly dependent on the GPU, but they can start actually DECREASING your hashrate, or even worse, start producing garbage with HW errors skyrocketing.
SUMMARY: Setting this for reasonable hashrates is mandatory. This setting is not recommended for Microcoin - use --raw-intensity

All the info is from Ultracoin forum.

But yes, creating a new thread would be good for coin. Call it "Big microCoin Settings thread" for example.