Topic: Simple core/memory auto-tuner script - page 2. (Read 4576 times)

I'm sending you a donation.  Not because This tool helped me speed up mining.  I got zero increase in mining.  I guess my cards are near optimally tuned.  But because I was able to optimize in another way.  Also,  I now have example code on how to talk to cgminer via APIs.  That could be useful to me some day.

Different cards.  Needing to set different Intensity, thread-concurrency, work size, etc.
And since i'm mining also on my main video card, I sometimes have to stop that miner if I have to do other gpu intensive tasks.

Just curious, why are you running two instances of cgminer on the same machine?  Different pools?

That's right.  I need to add something like: --api-port 4029  to the second cgminer, and then specify that port to the 2nd tuner.

Thanks - fixed.


Not sure I follow. If you have two cgminer programs running, one would have to have its API listening on a different port (using --api-port).
If you wanted to tune both of them at the same time, you would just launch two instances of the tuner, specifying the custom API port of the 2nd cgminer instance in the parameters of the second tuner instance.

I looked at the code and it looks safe.

Running it now.

In the readme.md file the example says to use -h 192.168.0.4, but it should be -i 192.168.0.4 (or your appropriate ip address).

I'm running two cgminers and can only tune one at a time without changing the default API address of the second cgminer.  I don't know how to do that yet.

Pretty cool.  I'll give it a shot and report back.

I didn't feel like spending ages fiddling with my GPU core and memory clock settings to find the sweet spot for Scrypt based mining, so I created a simple python script to do it for me!
The script uses the cgminer API, so you will have to enable that.

Thought I may as well share it: https://github.com/Uvedale/CGMinerTuner

The README on the git page gives all the available command line options and usage. This is a basic overview:

You specify your core clock and memory clock search ranges, and it will try combinations within those ranges, recording the achieved kh as it goes along.
Once it is done, it will give you a report on the best combinations.

There are a bunch of options you can add such as:
- max allowable temperature
- desired core:mem ratio range (to constrain search time)
- memory clock increment step
- core clock increment step
- time to wait between increment steps (longer is obviously better for more accurate results, but will impact the search time considerably)
- how many top results to show at the end

Other than possibly increasing your hash rate, it can also be used to find lower clocks that give you the same/similar hash rate.

This hasn't been thoroughly tested, so use at your own risk.
I'm sure the code can be improved considerably; I'll gladly accept pull requests!

If you find this helpful, please consider donating (especially if it increased your mining speed )

BTC: 1CM1zDcfUpg81fCJxacq2nNsZCE3G96FLH
LTC: Lej4FBr9k3HM49GJix922xyJCm1whqHmfU
FTC: 72KjWoZKM3C8S8xaujTNwcFjFbq48XNfpg