HA, thanks.
I did copy the miners and it worked on one rig, but not the other
, it still comes up with the same error.
I will do a clean install on this to get it running properly.
I checked all three of them, and the net 4.7.2 was already installed on them.
Also, whats the best way to benchmark on a clean install. Do you use a default overclock for all benchmark and then a custom overclock for specific algorithms?
How do you benchmark your software and what settings do you use.
I know it all depends upon your GPu, but for my 1080Ti, its 65, 125, 400. Am having crashes sometimes, but probably due to my settings.
Interested to hear what settings you use.
Cheers
That is frustrating that the one rig is still having problems with the miner install. The miner install process that HA uses is pretty straightforward: it downloads each file and runs the appropriate extractor. It really should not vary per rig, but since 7zip files require a different extractor, they tend to have more compatibility issues with system configurations. If you are comfortable with the command prompt (or have a GUI version of 7zip on that rig), you could try manually extracting one of the miners that have an error to the Miners folder to see if the 7z extractor provides any error messages. Any miner archives that could not be extracted should be in the Miners\Downloads directory in case the user has to do a manual install.
As for the OC settings and benchmarking, most users prefer to set their default overclock settings before benchmarking. Hash Auger will revise benchmarks based on actual mining results, but that process uses averaging to prevent the results from being skewed too much by a single mining run in case there is a miner failure, communication error, atypical vardiff from the pool etc. So it can take a few mining runs before OC changes for an algoritm are fully reflected in the revised benchmarks. If you are tuning a specific algorithm, you can use the Benchmark Utility in the Devices panel to select specific algorithms and miners to re-benchmark without having to re-bench everything.
Based on my experience and discussions on the HA Discord server, 1080ti GPUs are the trickiest to keep stable when overclocking. There are some users on the Discord server with 1080tis that are stable with core clocks as high as +125. On my own 1080tis I have found that I have to limit the core clock to +75 for stability for the air-cooled models and no overclocks for the EVGA hybrids. Since the liquid-cooled 1080tis stay much cooler than the air-cooled versions, the auto-boost on those GPUs is very aggressive and any additional overclock appears to push the GPUs over the edge. I also follow the advice of the Z-Enemy and T-Rex devs and not use any overclocks for x16 due to the variable nature of the algorithms it uses can make it hard to tune and keep stable.
Depending on the miners that you are using, you may want to consider lowering your memory clock adjustment. Newer versions of T-Rex and Z-Enemy appear to prefer negative memory clocks (as low as -800) for some algorithms such as x17. Also, some 1080ti users have noticed that they can increase their core clocks more and still remain stable when they use a negative memory clock adjustment. As a starting point, you may want to leave the memory clock adjustment as 0 for algorithms that use either T-Rex or Z-Enemy and see if that improves stability.
Finally, since tuning individual GPUs for specific algorithms can be a time consuming process, you may want to take a look at the new PG Bench overclocking utility:
https://www.pgboc.com. It automates the process of testing various overclock settings per algorithm and GPU. The results are often surprising as the optimal settings may not be obvious or intuitive based on the common expectation that the highest settings possible are also the fastest.
