Bitcoin Forum

I am sure many here has probably had this problem. I have an XFX RX 580 that keeps overheating.
I have changed the thermal paste and memory pads several times. Cleaned the heat sink thoroughly and replaced the fans, but still the same problem.
Is the card just dead? Or is there something one can do?

FWIW, I am running these values:
-core clock     1100
-core voltage  950
-mem clock    2200
-mem voltage 950

2200 MHz memory is a pretty high overclock for a Bios modded card. As explained in the guide below, run HWinfo in sensors only mode to check for memory errors and adjust the memory overclock until there are no memory errors or just a few. You should also be able to lower the voltage to 850 mV, which will use less power and the cards will run cooler.

http://mining.help

Decrease core and mem voltage 50 mv. If works normally(not temperature, only mining) - decrease next 50 mv. Repeat. Find your lowest voltage and look at temperature. If it is still high - decrease core and memory clock until it will be fine.
But i think that problem with thermal pads and thermal paste.

Try to change the cooler too if you believe the card is still good. There are a dozen custom cooler, including a cpu cooler that probably worth a try to cool down your card.

@2200 for Mem clock will crash most of my rx580 graphic cards, why cant you just reduce that back to default and see how the GPU performs? My RX580s are only stable at 2100MH and one at 2000MH max, anything over this the PC will freeze or crash

Tell us what temperatures do you read? I had one rx 470 msi armor was overheating to 85-90C the problem was with radiator. Heatpipe wasnt working. Changed radiator working normal now for many years. I had radiator from xfx rx480 didnt match but i cut it and made it match :) now it is working 47-50C :)
If it is around 65-70C then your card is normal. maybe your room is too hot.

Correct setting.
Check the pearl barley tubes in the cooling system. Heat the heat pipe on one side and check the temperature on the other side. If on the other hand it does not heat up, then the heat pipes are faulty.
Or buy coolers Delta for 0.5-1A :) You will not have problems with overheating

Thanks for all the replies, but I have finally solved it. After using many different clock speeds, voltage settings, etc, I finally discovered the memory timing levels.
Wow, what a difference! Now the same card runs at 30mh/s and at a constant 75 degrees (its hot summer in Hong Kong).
Now this rig, 7 x rx580, runs at 207mh/s using 580W power.All temparetures around 75. Nice and stable.

core clock: 1150
core voltage: 850
mem clock: 2125
mem voltage: 900
mem timing level 2

All these settings are now done in Radeon software, rather than Phoenix. I seem to get better results this way.

75 is ok but not recommended for long use.it is better get working them at 70C add some fan speed. or add extra fans

I saw several times where the wrong setting of heat sink after changing thermal paste and pads caused high temperature on GPU's. You should need to be sure that the heat sinks 100% leaning on the GPU chip. you must not put a thicker thermal pad than is a gap.

7x rx580 consumption the only 580w? Are you sure? This is from mining software, right? did you measure before the complete miner machine?

This is only the consumption of video chips from 7 video cards. There are no such results. Such a farm consumes about 900 watts or more from the wall.
Each miner must have a wattmeter in order to check the correctness of the settings and not set lower consumption values, from which the power consumption from the wall does not change, but the video cards start to fail.

Yes, of course the 580W consumption is just taken off the software. Off the wall I haven't measured this rig yet.
But the same software reading for my 6 x 1070 rig is much lower mh/s and over 700W power.

7xrx580 will be 800-900watts on the wall depending on your psu effiecency. and 6x1070 will be more or less like this.