The Most Power Efficient & Best Value Of AMD GPUs For Linux Gamers

[Luke_Wolf]

Have you seen the benchmark results?

Even considering the benchmarks if you care exclusively about performance and want AMD, yeah you're going to go with a fury. Things may not look wonderful now, but Catalyst needs quite a bit more work to work appropriately with that HBM memory (both on Windows and Linux) after which it should really shine, and the card won't really show its legs until Vulkan hits. However assuming you see a card as a 4-5 year investment it'll provide the best performance of current AMD cards in the long run. Now if you're the kind of person who replaces their card every year, then of course the results as they stand today are what you're going to care about and you should probably go with the 290 or 390.

[Jonathan Silverblood]

What I miss here is some APUs. When looking at power efficiency, they should at least be contenders. When looking at price, they should at least be contenders.

FPS though, probably not so much

[Ploplop]

AFAIK (and if I understood everything correctly), the estimation of the power consumption as done today is flawed...

If a device consume 100W during 2 seconds and another 50W during 20 seconds, current test will say that the second card is the most efficient. However it isn't because the second one takes much more time to perform the same work.

That's plain obvious for CPUs: the average power is not the good value.

For GPUs, it could be much more complicated: the benchmarks need to be

1) either at constant fps (to compare the power consumption displaying the same thing, not at 30 fps for card X and one at 180 fps for card Y with card Y being obviously much more power hungry);
2) or at constant frame number: this way, you measure the time it takes to get the job done (i.e. displaying X frames in total on either GPU).

But in both case (even in the first one), it is far better to intergrate the power consuption over time instead of calculating an average.

To the bystander: if you really want to consume less power, either try to play your game on a raspberry pi OR choose any card you want but limit the FPS number. You can't possibly compare the power consumption of one GPU and the other that does twice as much work and saying "OMG, I won't by that second card that consumes twice as much power".

[drSeehas]

A R7 370 faster than a R9 285?
Something is wrong!