12-Way RadeonSI OpenGL Comparison vs. NVIDIA On Ubuntu Linux

[phoronix]

Phoronix: 12-Way RadeonSI OpenGL Comparison vs. NVIDIA On Ubuntu Linux

After posting a number of NVIDIA GPU Linux benchmarks this week using their latest drivers, here is similar treatment on the Radeon side using their newest open-source driver code.

http://www.phoronix.com/vr.php?view=24720

[theriddick]

Is the point of these at 1080p to show which cards/drivers are bottlenecking the most at 1080p? There is some interesting situations where the GTX 980 beats the 1080 cards, not sure whats going on there?!

[Michael]

Is the point of these at 1080p to show which cards/drivers are bottlenecking the most at 1080p?

Yes and no. Main reason is due to going back and testing older cards (like with the NVIDIA tests on Fermi/Tesla and the upcoming R600g tests with the same result file). Obviously the old cards get killed at 4K.

[DanglingPointer]

Good to see the 290 in the mix! She's getting old but she's up there with the 580s and Fury!

290X on water is even better!

[Michael]

Good to see the 290 in the mix! She's getting old but she's up there with the 580s and Fury!

290X on water is even better!

My 290 finally started working properly again on 4.10~4.11.

Unfortunately I have no 290X.

[schmidtbag]

Good to see the 290 in the mix! She's getting old but she's up there with the 580s and Fury!

Agreed - as someone who owns a 290, it's nice to see where it stands in such a wide variety of tests. Though, it is a little cringey to see how power-hungry it is. I've sometimes considered getting a 480 due to power consumption, and because of stuff like Vulkan support. But, seeing as the 290 often outperforms the 480, currently it just doesn't make sense to make such a purchase. The new Vega chips look promising in terms of performance-per-watt.

On a side note, I also like having my 290 for novelty reasons - my rig involves an ITX case and the GPU is literally the maximum size that will fit inside this case. If it were 1mm bigger, it wouldn't fit.

[sonnet]

Personally I'd find more interesting if rather than using similar cards, were tested different architectures.
I mean, by just having a look at the cards selected for nvidia:
- GeForce GTX 950
- GeForce GTX 960
- GeForce GTX 970
- GeForce GTX 980
- GeForce GTX 980 Ti
- GeForce GT 1030
- GeForce GTX 1050
- GeForce GTX 1050 Ti
- GeForce GTX 1060
- GeForce GTX 1070
- GeForce GTX 1080
- GeForce GTX 1080 Ti

I mean: wouldn't have been more interesting if only one between the 970 and the 980 was selected (and the same to be applied for the other cards I underscored, thus discarding 4 cards in total from the ones selected) and rather having 2 cards from kepler and 2 from fermi generation selected?

I also don't get the R9 290: in the latest benchmark-review it seemed to be still plagued by some bug that heavily slashed its performances (at least on Ubuntu). Should I assume it's been fixed?

[jrch2k8]

BTW michael TAHITI's GPUs with AMDGPU driver/kernel runs vulkan just fine, At least all the examples i have found.

So, maybe you can test the 7950 at least on vulkan for the next article XD

nice job, tyvm

[schmidtbag]

I mean: wouldn't have been more interesting if only one between the 970 and the 980 was selected (and the same to be applied for the other cards I underscored, thus discarding 4 cards in total from the ones selected) and rather having 2 cards from kepler and 2 from fermi generation selected?

Michael tests whatever he has. I agree it's not really interesting to see benchmarks of similar architectures (particularly rebranded GPUs) but he might as well test with whatever he's got. He gets most of his Nvidia GPUs for free.

I also don't get the R9 290: in the latest benchmark-review it seemed to be still plagued by some bug that heavily slashed its performances (at least on Ubuntu). Should I assume it's been fixed?

It seems only some GPUs are/were affected by whatever the problem was. As a 290 owner, my GPU ran flawlessly since I've owned it.

[humbug]

Good article thanks, looks like AMD's openGL is finally where it needs to be.

Look forward to the Vulkan article. There is an increasing number of Vulkan games to benchmark. Dota2, Talos Principle, Serious Sam Fusion 2017, Mad max...

My 290 finally started working properly again on 4.10~4.11.

Please include the R9 290 in the vulkan article too. It's important to measure vulkan performance across all GCN iterations. Both on radV and amdgpu-pro.