Hi Patrike, today I come to make a suggestion that I know is not going to be easy to do and you may not agree, but I present it to you.
My idea is to create an OC AUTOTUNE. Each brand of card, each motherboard, bios configuration, memory quantity and speed, slower or faster CPU etc ...
Getting the best OC for an AL-go is very difficult, and more so if I have 6 totally different and multi-brand rigs.
My idea starts as an addition to Benchmark. You choose all the cards of the Rig, you choose the currency and therefore the Al-GO, you choose the time, 15 -30 minutes for coins that vary little the hash, and 1 hour for more changing algorithms like x16r.
Here comes the change, a separate addition to AUTO OC. Where I define 6 values
Maximum minimum PL for example 70-90 Always in jumps of 5.
Minimum core maximum for example 30-140 in jumps of 5
Minimum maximum memory for example -500 +600 jumps of 50
This is just an example, they can be other jumps and other numbers to choose by user or leave it pre-defined. Here are 2 variants
Brute Force: test all possible configurations
Scanning. Several different tests are done, about 10 for example, and then just optimize between the 3 central sets, to find the best configuration
If a test fails, it is considered as a failure and the process must continue.
The OC must take into account 3 variants, HASH, SHARED, CONSUMPTION and temperature
The best oc is not the one that has the most hash, but the one that has a good hash with a shared that is not little or much. And avoid exceeding the maximum temperature.
Finally, 3 possible winners are selected, and a minimum test of 30 minutes or 1 hour is done to each one, to check stability and temperature, and then the best of the three that you would give per winner for that rig and AL is chosen -GO.
What I say can take more than 1 day of testing, I keep it in mind, but if I undermine what is more or less profitable, I don't mind mining it fixedly, than doing the AUTO OC.
Because many times you get better results with less consumption (this should only be informative), more hash with less temperature. Less Crash of miners, more stability.
For those who have 10 totally equal rigs, they only have to do it in a rig and extrapolate them to others. For those of us who have varied rigs, it is torture, and I prefer it alone. Obviously it only does it when an al-go or currency is in profit and I take advantage of the Auto OC with the mining of that currency.
I know it sounds complicated, it is. But one of the links that the program has left to be 360 degrees in the GPU mining world.
Getting better hash, or adapting the miners for summer heat with an AUTO OC al-go to al-go at least makes you have one of the best configurations you can have for that rig in question. With multi-brand cards, ones with 8 or 16 gigabytes, the latency configuration in Bios, the chip set, etc. It is impossible that an OC is worth to all
Sometimes I notice that there are cards that work below what they can but others in the same rig are going full. Finding the intermediate point in that OC gives me stability and accuracy. Above all, extract the maximum performance from my equipment. And something that the competition is light years away. And in the end if you plan well it should not be so complicated. We already start from the base that an AUTO OC will take hours or more than a day, I already have it in mind, but it undermines what is in profit that I tell it.
Obviously it is just an outline and surely there are better ways to do it, but without testing it in each configuration it is difficult, maybe I am wrong