AMD Radeon RX 6800 Series Linux Performance

[ms178]

Didn't even know I had such package in the repo, lol. Or perhaps it's some package from Tumbleweed? I don't use Tumbleweed myself, so I'm not aware if something goes wrong there.

Yes, I am on Tumbleweed. It is not a huge deal, but these small things add up over time. I also need to do a zypper dup -r with your repo to get the complete Mesa stack updated with the versions you provide. The default route just installs four or so packages and doesn't update the vulkan drivers etc.

[ms178]

Once more a proof how good the mesa stack is.

It also comes down to the applications tested. As some applications make use of Nvidia's proprietary APIs (e.g. OPTIX), Nvidia gets a huge advantage in these. Of course not everyone cares about these applications though.

[angrypie]

Exactly where I expected Big Navi to be, which isn't a bad place at all.

For fuck's sake stop feeding the NVIDIA fanboy. Maybe you didn't notice but he has been repeating himself endlessly across every thread that mentions a new AMD product (a troll with Alzheimer? We'll never know).

[mibo]

Thank you Michael for the nice review.
And, thank you AMD for the great product with good linux support.

I will probably wait for the 6700 series cards and hope they have the same low idle power consumption.
The gaming tests of the 6800 (without XT) seem a bit weak compared to the Windows tests I read (6800 should be above the 2080Ti).

Also looking forward to the OpenCL results :-)

[Shevchen]

Alright. Mesa + ACO hasn't really been optimized for RDNA2 yet and the performance is already vergy good. I expect the performance values to be higher than the pro driver in the future and the results are for a rather fresh start quite impressive. Also, the Mesa devs may be like: "This isn't even my final form!" now.

Time to give them some coffee and... well, cards. AMD? Are you reading this?

[piotrj3]

Not bad at all, although most benchmarks on internet shows that at 4K AMD simply falls behind to Nvidia, while having really good results at 1440p and 1080p.

What makes me curious, does AMD only use 16GB of gddr6 vram for sake of populating memory bus? Because those 4k results smells like memory speed bottleneck.

[pete910]

Excellent option for gaming without RTRT and Linux users.

If you're looking for a future-proof card, you probably have to choose NVIDIA or wait for RDNA 3.0.

Good point on the future proofing.

Think AMD is the long term answer looking at this. It appears that NV sponsored titles are in fact better suited to NV where AMD sponsored titles is reversed, Likely due to AMD powering the consoles .

[blacknova]

Think AMD is the long term answer looking at this. It appears that NV sponsored titles are in fact better suited to NV where AMD sponsored titles is reversed, Likely due to AMD powering the consoles .

That is not exactly surprising, even before RT have been implemented on both AMD and NVIDIA GPUs both vendors had different strong points - AMD usually preferred more complex pixel shaders, while NVIDIA better handled geometry.

We as well might see something simillar in RT department.

[piotrj3]

Good point on the future proofing.

Think AMD is the long term answer looking at this. It appears that NV sponsored titles are in fact better suited to NV where AMD sponsored titles is reversed, Likely due to AMD powering the consoles .

Dirt 5 is a joke tho, literally racing game thar doesnt' have wheel support, has arcade racing style in dirt (that is not acceptable) has mixed rating on Steam, and 1440p gets 70 fps on 2080ti. Even graphics doesn't look impressive, Forza horizon 4 doesn't look worse yet scores easly more then double the FPS at such resolution. Forza horizon 4 although is console port that seems to be primary AMD focused, still doesn't have such giant performance penalties for Nvidia.

[HenryM]

Those DOTA numbers look absurd. are the wattage figures directly comparable? 3080 is only 6% faster than 6800XT with 82% higher average power draw...