Closed UselessGuru closed 3 years ago
Will do. Hopefully next week. Any specific card you're interested in? Because for some algorithms amount of memory needed is different for each card.
I am developing a miner file for MultiPoolMiner. It will run on basically any NVIDIA cards, so so we should cover the most current card generations. It will run all algos with either 1 one 2 threads. Will the requred memory amount just double?
Because for some algorithms amount of memory needed is different for each card.
Is there a way to calculate the amount from the amount of installed memory?
I wouldn't recommend running two workers per device, it won't gain that much more performance, but it will cause more instability. Some algorithms will even crash, like lyra2rev2, and as you said yes, the amount of needed memory will double. All algorithms are tuned and tested for most of the NVIDIA cards, so there shouldn't be any problem related to not enough GPU memory.
Also.. You can always check how much GPU memory is each algorithm using with device.get method. Check the gpu_memory_used field.
Only Daggerhashimoto is supported for cards with 6+ GB of memory. Closing.
Could you please document the minimum memory requirements (RAM & graphics card mem) for each supported algorithm? This for single and multiple worker configurations.
Thank you very much!