NVIDIA 375.10 vs. Linux 4.8 + Mesa 13.1-dev AMD GPU Benchmarks

[geearf]

Sorry to say it like this, but If you don't see *anything unfair* there, then you should open your eyes Vulkan and OpenCL is missing there, but Pro driver has those

OpenCL is not unfair to me, it's just the current status of the driver. (Useless comparison sure, since it only compare one against nothing, but not unfair).
Vulkan is more arguable, because it's a PPA issue not a driver one.

[geearf]

If we somehow bypassing that, then we should bypass drirc for open driver so that Unigine could not run - that would be the fairest

I won't disagree on that

[dungeon]

Vulkan is more arguable, because it's a PPA issue not a driver one.

So you are talking about some issues in comparison to this Vulkan conformance passage

See on Khronos site conformance for Vulkan 1.0 done on Ubuntu for Polaris and put a stamp on it

Product Family: AMD GCN 1.3 AMD Radeon RX 480 AMD Radeon RX 460	CTS Version: 1.0.0.3 CPU: x86 OS: Ubuntu 16.04 64-bit

[dungeon]

There we see cts passing on Ubuntu and Polaris and here we present some driver issues, ppa issues, whatever ... in comparison to something that is regulary confirmed to not have issues

[Ouroboros]

The most is not really an argument, as i can say that most gamers use 1920x1080 resolution, but most test here were not done on that resolution

That way i can make it up and legally say how 4K and 2560x1600 are less used than 1920x1080 even combined, then difference is on usage of closed AMD stack vs open

Yes, most is an argument. It's whatever Michael feels is relevant and worth testing; he doesn't have an infinite amount of time to put into testing every driver with every game on every card with every configuration.

And yes, the fact that
1) No driver for Ubuntu 16.10
2) The driver is four months old while Mesa and NVIDIA have been moving. I'm sure Michael has done enough AMDGPU-PRO benchmarks since 16.30, too.
Are perfectly valid reasons not to bother with closed AMD right now.

[dungeon]

Well i agree on that *one* that there is no driver for Ubuntu 16.10 and that one is a valid reason to me.

Any else artifical reasons are really just BS on average

[dungeon]

he doesn't have an infinite amount of time to put into testing every driver with every game on every card with every configuration.

Well i don't have time too... but i read Vulkan/OpenCL conformance and spot that tramendios amount of hardware passed that - done by AMD itself and on Linux.

In contrast, what user here see that those things not even working - which if you can agree is not really true

[tomtomme]

Dungeon stop BSing yourself. You say Most benches were not done at 1080p? Look again and learn counting. And no, you cannot count synthetic gputest.

Really unfair how hard a competition nvidia gets by an open source driver at 4k...

[dungeon]

Dungeon stop BSing yourself. You say Most benches were not done at 1080p? Look again and learn counting.

Ah yup, seem you are right 8 out 15 are really 1080p

And no, you cannot count synthetic gputest.

Why those are there then, so that anybody can ignore them?

[zakhrov]

I don't blame Michael at all. He's getting the NVidia cards for free and paying for the AMD ones.

That's smart marketing by NVidia.

AMD used to send over review samples as well. They stopped doing that though