Hey all, been doing some noodling over mysetup and looking for some outside opinions on the pros and cons between running one instance of Claymore across all 6 480's and running 6 instances, one each per GPU? I tried doing some googling, but the relevant search terms for hunting down something like this are painfully generic and result in a ton of useless info, sadly.
So far, all the pros and cons I've come across seem to be pretty personal preference-related:
For per-GPU miners:
+ If one hangs in an unrecoverable state, the others are left happy
+ My stats and tracking allows for a bit more nuanced info and prettier stats
+ If a GPU hangs in a recoverable state and triggers a restart script, only that GPU resets it's VARDIFF when reconnecting to the pool (and has to re-scale)
+ I can fine-tune launch commands per-GPU(helpful since I have a very disparate batch of card quality)
- More script files to track and sort.. if I want to leverage reboot scripts then I also need each mining instance in it's own folder as well
- Lots of console windows ruining my clean layout
For unified miner:
+ One launch script/one install folder
+ One window to monitor
+ One data feed for reading into my monitoring tools
- All data and stats are crammed into one feed
- If a GPU hangs in an unrecoverable state, they all hang
- If a GPU hangs in a recoverable stats and triggers a restart script, my entire VARDIFF pool link resets for all GPUS
- I can still fine tune per gpu most tweaks in the command line using ","s but it's much easier to fat-finger something and command lines get crazy big.
So far everything I have accounted for seems to be pretty much down to "weighing the pros and cons of personal hassle" more than anything else. Before I made a final choice based on that however, I wanted to check and see if other folks have extra thoughts I haven't concluded yet, and any advice from personal experiences in the same area.
Thoughts?
So far, all the pros and cons I've come across seem to be pretty personal preference-related:
For per-GPU miners:
+ If one hangs in an unrecoverable state, the others are left happy
+ My stats and tracking allows for a bit more nuanced info and prettier stats
+ If a GPU hangs in a recoverable state and triggers a restart script, only that GPU resets it's VARDIFF when reconnecting to the pool (and has to re-scale)
+ I can fine-tune launch commands per-GPU(helpful since I have a very disparate batch of card quality)
- More script files to track and sort.. if I want to leverage reboot scripts then I also need each mining instance in it's own folder as well
- Lots of console windows ruining my clean layout
For unified miner:
+ One launch script/one install folder
+ One window to monitor
+ One data feed for reading into my monitoring tools
- All data and stats are crammed into one feed
- If a GPU hangs in an unrecoverable state, they all hang
- If a GPU hangs in a recoverable stats and triggers a restart script, my entire VARDIFF pool link resets for all GPUS
- I can still fine tune per gpu most tweaks in the command line using ","s but it's much easier to fat-finger something and command lines get crazy big.
So far everything I have accounted for seems to be pretty much down to "weighing the pros and cons of personal hassle" more than anything else. Before I made a final choice based on that however, I wanted to check and see if other folks have extra thoughts I haven't concluded yet, and any advice from personal experiences in the same area.
Thoughts?