Does anyone else run a AMD Radeon Pro WX 5100 with dual monitors plugged in, and notice that it runs really hot? With a single monitor it idles in the low to mid 50's C (which still seems hot for an idle GPU, but manageable). With two monitors plugged in, mine idles in the upper 60's C while my PCI fan revs up to around 2000 RPM, then after a while the PCI fan usually returns to its normal 1300 RPM, and the GPU runs up to the low 70's C and stays there at idle (goes up to the low 80's C under load).
I already repasted the GPU recently. Also, the GPU fan itself doesn't seem to increase its speed very much as the temperature increases, which all seems very odd...
Most issues sounds quite normal to me for a modern AMD GPU.
1) Low 50s is normal for single monitor idle especially if that's the hotspot temperature, but not the die edge temperature. Most modern graphic card's fan will just stay at a very low RPM until reach the target temperature, then spin faster and faster to keep the GPU at target temperature. In the case of WX5100, the target temperature is 85°C. Therefore, low fan speed when the temperature is still below 85°C is absolutely normal.
2) For dual monitors setup. The VRAM speed most likely will stay at 1000MHz or even straight to 1250MHz, rather than 300MHz (single monitor). This is why the graphic card will stay at a higher idle temperature.
3) High PCIe fan, this is the abnormal part. This is a known bug on cMP. There is no permanent fit for this, but quite a few easy work around. e.g. You can simply use AirOut (credit to cdf) to stress the GPU a bit every time after login, then the SMC will able to read the graphic card's reading correctly, and the PCIe fan will also behave normally.
4) When under high stress, you can expect the temperature fluctuate between 85°C to 95°C. This is how the card programmed to do so. Target 85°C means the card will let the GPU keep warming up until touch 85°C, then the spin should spin faster and faster to keep at that temperature. However, if the stress is really high, the temperature can keep increasing, until reach the Mac temperature 95°C. Then thermal throttling should kicks in, and keep the temperature won't go over 95°C. If fail to do that, then it's a sign that the cooling system (e.g. thermal paste) is failing.
A quick summary, "Keeping the GPU as cool as possible" is an outdated mindset. Modern graphic card simply not programmed to do that. One of the reason most likely because "thermal cycle" actually can actually shorten the die's life span. But keeping it at a higher temperature for long time won't. Therefore, modern graphic card's fan profile tends to let the GPU work at a higher (but narrower) temperature (range).