RadeonSI/RADV Mesa 17.3 + AMDGPU DC vs. NVIDIA 387.12 Linux Gaming Performance

[Leopard]

Sorry but these results shows that AMD is not good as Nvidia even a slight.

Look at min fps. On Nvidia side , it is more than double of Amd results.

There are still looongg way to go for Amd. These dips will return you as stuttering which is a bad thing for gaming.

[slacka]

Because if running most of the tests at 1080p, there is no scaling with the higher-end parts.

Hardware fillrate isn't the end all be all of benchmarking. Your intro sells this as "much improved RADV Radeon Vulkan performance" but to see driver performance increases best, you must be CPU limited, not GPU limited. Yet all these 4K tests are GPU limited. These benchmarks are valuable for those looking to buy hardware, not to see how Vulkan drivers are maturing.

[Niarbeht]

4k makes sense on 43'' screens, as there it has a similar PPI (pixel per square inch) of a normal fullhd PC monitor, so you actually have more space.

If you buy smaller screens then you need to upscale your applications or it's too small to be read, and this defies the point of a 4k screen.

For many, myself included, the point of ludicrously high resolutions like 4k is to not be able to see the pixels anymore. The potential for extra screen real-estate is a side effect.

[tomtomme]

4k makes sense on 43'' screens, as there it has a similar PPI (pixel per square inch) of a normal fullhd PC monitor, so you actually have more space.

If you buy smaller screens then you need to upscale your applications or it's too small to be read, and this defies the point of a 4k screen.

Spoken like someone who has no 4k screen. Bravo.
If you had you would know that you only need to scale up a bit because of the higher dpi. Everything is so much sharper, so it can bei smaller. You may Not double your space but you get like 1.5 Times the space.

[andre30correia]

not bad, they only need more optimizations and have the best driver in 12 years for amd/ati cards

[Del_]

This is really amazing. Looking aside from RADV and Vega, which both needs some further maturation, radeonsi is now highly competitive with the closed source nvidia driver in opengl gaming. This is a fantastic achievement, congratulations to all developers who made it happen!

[starshipeleven]

And to be frank, it is a benchmark so Michael is trying to push the cards to it's maximum potential too, same if 8K/16K is the norm (and implemented) and then that will be used etc. Not saying 4K is the norm because it's still a niche, my friend has a 4K screen not thinking he was sitting too close would screw up his eyes/neck/body in any way he reduced the resolution (the application not the desktop so he still utilizes the 4K resolution). Makes me think if a lot of gamers that has 4K screens thinks of the distance from your screen to your seat is good too.

4k makes sense on 43'' screens, as there it has a similar PPI (pixel per square inch) of a normal fullhd PC monitor, so you actually have more space.

If you buy smaller screens then you need to upscale your applications or it's too small to be read, and this defies the point of a 4k screen.

[Sethox]

All the vulkan tests were run at 4K, which somewhat hides the performance gap between RadV and the Nvidia vulkan driver...

Because if running most of the tests at 1080p, there is no scaling with the higher-end parts.

And to be frank, it is a benchmark so Michael is trying to push the cards to it's maximum potential too, same if 8K/16K is the norm (and implemented) and then that will be used etc. Not saying 4K is the norm because it's still a niche, my friend has a 4K screen not thinking he was sitting too close would screw up his eyes/neck/body in any way he reduced the resolution (the application not the desktop so he still utilizes the 4K resolution). Makes me think if a lot of gamers that has 4K screens thinks of the distance from your screen to your seat is good too.

[Kendji]

Really impressed the driver improvements that has happened and keeps happening

[Michael]

All the vulkan tests were run at 4K, which somewhat hides the performance gap between RadV and the Nvidia vulkan driver...

Because if running most of the tests at 1080p, there is no scaling with the higher-end parts.