Quote from: kerney666 on December 23, 2018, 04:07:53 AM

Quote from: ku4eto on December 22, 2018, 09:42:17 PM

Quote from: MoneroCrusher on December 22, 2018, 11:04:58 AM

I'm suggesting 2 killer features that will make this incredible miner complete:
1. Include in watchdog/logfile the uptime of a crashed GPU (i.e. GPU 4 crashed, uptime: 5 hours, 12 minutes)
2. A simple fan curve in addition to the temperature watchdog (i.e. --temp_max=83,100 -> max temp of 83 while fan must be at 100% and --temp_ramp=75,70 -> when it reaches temp of 75, spin fan to 70% and increase to linearly until max defined temp is reached)

The 18.12 Adrenaline drivers have Fan Curve. But since you are on Linux, that wont help you Tongue

Yeah, on Linux a fan control mechanism using interpolation between defined points on a fan/temp curve could be a nice addition, not too much work either, it's simple scaling the fan pct using sysfs. For Windows it's getting messier now, I haven't looked into how ADL interacts with the 18.12.2 fan curve at all yet.

For the first question, the uptime for all gpus is the same, so it's logged every time the hashrates are logged (every 30 secs or what you set the log interval to). We can add it to the log output made by the watchdog when a gpu is detected dead though, it's trivial really. Personally, I don't see the big value add though

.

So can we expect the fan curve and GPU uptime log

?
First one is very helpful for obvious reasons, second one is useful for bigger farms that have to tune many single GPUs.

Another suggestion:
Split the hw error. i.e. instead of hw: x, you have hw: mem: x, core: x so we can see if an error occured on memory or if it was core related (mostly too low voltage). Would be enormously helpful for debugging/optimizing a larger farm.