I only have a few XFX cards spread across three rigs. I did like them as they have a slightly higher hash rate, but I can say with certainty now that any instability with those three rigs is 100% from the XFX cards. The fans on XFX cards refuse to respond properly to the AMD drivers (I tried all driver versions, BTW newest does not work). I have one card that will sit with one fan off past 90C. The other two need external airflow just to stay below 70C. I missed this for the last 10 months as I have extra airflow on all my rigs and push them hard so they auto restart often. All my Nitro cars run at <60C, all the time, on same rigs. Perhaps they have fixed this in the 580 or Vega but I have moved to more efficient GPU’s for Zcash.
I get irritated when I can smell cards burning but one fan refuses to come on (every XFX card I have does this). Does not matter what you change in the BIOS or the AMD drivers. I will never buy these cards again for mining, but it does not sound like this is an issue for Windows miners.
I have moved all XFX cards to a test rig. My rigs all auto restart, so I plan to run them into the ground until they burn up. The test rig is auto restarting about every 10 min, hitting 90C and crashing. Way better than these cards restarting three of my main rigs for one hot GPU every 10 min.
My Nvidia rigs do not have this problem, software automatically reduces power if a GPU hits 70C, then increases it again if the temps drop again. Deals with day and night temp fluctuations nicely but airflow over multiple rigs is critical for farm cooling, can be a pain, each building HVAC system is different.
These XFX card have been running 10 months this way so perhaps they will just keep going, but I plan to push them hard just for spite…
did you ever stop to think its a defect in the card / sensor / fan and you should RMA it and get it or them fixed? Sounds like it’s a bad batch or a defect.
If they worked correctly when you got them but don’t know ( see above) Get them fixed / replaced
They were not all bought from the same vendor or at the same time so unlikely a bad batch. Its not worth it to me to return them as its only four cards and they have paid for themselves many times over. Besides this is a perfect situation to test some new code improvements to deal with older rigs. These AMD RX480 rigs only get ~2200 Sol/S. With all four of the XFX cards moved to a test rig, replaced with Nitro cards, those rigs are now back to 100% stability and not restarting at all. So I can forget about them again.
With a strategically placed extra fan on the XFX rig it helps quite a bit, the temps are lower even on the card with only one fan running consistently. I am now writing / testing new code on that rig that will kill a miner instance when a GPU gets too hot instead of just restarting the whole rig, it will then restart that one instance after the GPU cools down, and restart the rig if that fails. If that happens 3 times in 15 min it will permanently shut down that GPU and send me a text msg (but will continue mining on the remaining GPU’s). This is a good test bed for this code before I consider moving it to my other AMD rigs.
My main AMD rigs will shut down a whole rig if any GPU get too hot, I assumed a fan failure when I wrote that. So now if this new code reliably deals with these XFX cards, then that will be a big plus as my AMD RX480’s are getting older, and will start to need fan replacements soon. I already needed to blow out the dust twice on these GPU’s, the temps started to creep up in the logs.