There is another argument towards Polaris GPU for Linux OS users :) for those who know.

Weird way to say "I lost an argument", or "I didn't have a point to begin with", but okay. Enjoy all your 7 FPS (delivered @ near-room temperature), Mikita!

You still didn't get the idea. Anyway, being a math graduate doesn't help to have logic. Let's just stop at this point.

Quote from: Dorby on January 13, 2021, 09:53:05
But doesn't alptop dGPU run through iGPU unless it has a dedicated MUX switch? So on performance mode, you'd be running CPU+iGPU+dGPU anyways.

No, iGPU can be switched off when required. Or, in some cases, they can even work simultaneously on one goal.

Go ahead and cool your RX 640 with LN2 if you want. Still won't manage even a measily 25% gain over our poor, cooling-starved RX Vega IGPs. "More efficient" my a*s. An entry-level dGPU is NEVER more efficient than the integrated ones, period. Not one such case.
P. S.: I'm an applied maths (CS) graduate. I "get the idea", OK?
P. P. S.: If you feel so bad about the thermal allowance on these Vega's, I can hold my PC open for you so you can blow (like with your mouth) on it? That will make you happy? :p

But doesn't alptop dGPU run through iGPU unless it has a dedicated MUX switch? So on performance mode, you'd be running CPU+iGPU+dGPU anyways.

You didn't get the idea. When you have separate chips to cool, you can make separate heat pipes + separate little radiators, this is more efficient than trying to cool one chip, allows to expand the thermal output just a bit more. For example, cooling one 70W chip is harder than cooling two 45W chips separately.

 ??? But it IS "that simple". Polaris 12 is a tiny piece of silicon, especially in the world of GPUs (which are "MASSIVELY PARALLEL" architectures, need I remind you).
No one is going to argue that such an absurdly small GPU is a waste of GDDR chips and should only be leveraged when a) no IGP is present; b) the IGP is Ice Lake G1 or Vega 3.
In the words of a record-setting US government official, and a real giant of public speaking, "Come on, man? Are you a junkie?"

This is not that simple, guys. Cooling two crystals (separate for CPU part and GPU part) to some degree is easier in some cases than cooling single powerful APU (say CPU+GPU in one crystal). Moreover, discrete GPU doesn't depend on crappy RAM. Even if it's crappy single-channel RAM with crappy timings (which manufacturers are so fond of installing), GPU has it's own memory.

You know you're dumber than Trump when you add A FUCKING GPU THAT IS SLOWER THAN INTEGRATED GRAPHICS

RX640: Waste of money, space and cooling

if it's slower than MX150, than it's slower than the iGPU

oh, Acer