Changelog
Pearlhash:
- Efficiency improvement for 30xx, 40xx, and 50xx series
- Performance improvement for 20xx and CMP 40/50hx ( lock the memory clock to 810/405 )
- Performance improvement for cmp100-210
- Added support for Zstd compression of proofs (saves bandwidth and helps with latency on the pools that support it)
Added a new feature --power-save (eg: --power-save 1 or --power-save *,1,2)
- It takes a level between 1 and 3 and works on top of your OC settings and might save some watts on all algos. Works better with a locked core clock.
- power-save is highly recommended to be tried on 5070 ti, 5080, 5090, 20xx series and CMP cards(non hbm) . Lower end 50xx series cards seems to be only working with driver version 610.43.03 (other drivers might just crash) and it doesn't help them much.
- Higher levels are not always better. for 30xx and 40xx series, the effect is more subtle and level 1 might be better than level 2 (These cards only have two levels)
- It can potentially lead to crashes and stability issues like other overclock settings.
Added --vcore-offset(e.g. -12.5) and --vcore-target( e.g. 768.75)
- These are another way of adjusting the V/F curve. You typically can get the same result with just locking core clock and adding a core offset.
Added locking memory clock to 405 on Windows
Added support for 405 memory clock on older drivers
Added native overclocking for Pascal cards