News AMD's New Radeon Preview Driver Brings Double-Digit Performance Gains

[Admin]

AMD has released the AMD Software Preview Driver May 2022 with new features and performance improvements.

AMD's New Radeon Preview Driver Brings Double-Digit Performance Gains : Read more

 

[hotaru.hino]

Kind of annoyed AMD didn't provide the FPS values. For all I know, that 30% on WoW means it went from 40 FPS to 52 FPS

I mean sure I could go try to extrapolate what the performance is from reviews, but I shouldn't have to.

 

[Makaveli]

Kind of annoyed AMD didn't provide the FPS values. For all I know, that 30% on WoW means it went from 40 FPS to 52 FPS

I mean sure I could go try to extrapolate what the performance is from reviews, but I shouldn't have to.

If you are on RX6000 series hardware download the driver and check it yourself you don't need to wait for AMD.

 

[hotaru.hino]

If you are on RX6000 series hardware download the driver and check it yourself you don't need to wait for AMD.

I don't have an AMD graphics card. And even if I did, it may not be applicable to the card that I have. The point is that percentages without any sort of absolute value to compare against is meaningless. What if I had some performance requirement? I still can't tell if the driver update allows the cards to get the performance I want.

Also what's with reporting low percentage gains like 3-5%?

 

[SunMaster]

I don't have an AMD graphics card. And even if I did, it may not be applicable to the card that I have. The point is that percentages without any sort of absolute value to compare against is meaningless. What if I had some performance requirement? I still can't tell if the driver update allows the cards to get the performance I want.

Also what's with reporting low percentage gains like 3-5%?

It's a driver update, not a major piece of new hardware. Like all software updates it's probably safe to assume the performance gains for the most part are minor.

No reason to get upset over anything.

 

[hotaru.hino]

It's a driver update, not a major piece of new hardware. Like all software updates it's probably safe to assume the performance gains for the most part are minor.

So why report it at all?

 

[SunMaster]

So why report it at all?

Did you even check out the link provided in the article? While it's not like a press release causing your whining, they did provide some numbers.

If you're waiting for anyone (at all) to do a driver version test vs driver version test I think you're going to get very disappointed.

 

[Alvar "Miles" Udell]

Is it really a performance increase, or is it a quality reduction disguised as a performance increase?

 

[mitch074]

Is it really a performance increase, or is it a quality reduction disguised as a performance increase?

You can't really have that any more with newer cards : most of the performance is eaten away by the shader computations, and reducing quality would mean corrupted graphics in some places.
On the other hand, since shaders need to be compiled, a key optimization in shader compilation may yield huge performance improvements. If we use the AMD open source drivers as a reference, sometimes an optimization is left aside because it's buggy, but when enough workarounds actually allow it to run, it is then enabled - and we see such performance increases across the board.

I wouldn't be surprised if AMD's Windows driver team took a look at how the open source driver developers performed some operations and ported them inside the Windows driver. Since both use differing code bases there will necessarily be a delay in such a port, especially on DX11 that is far more complex than Vulkan, but considering how many games ran better on the Steam Deck with Steam OS than with Windows, it wouldn't be a stretch that they took note of what bottlenecks were found by Valve and the community and decided to implement them.

 

[Diceman_2037]

I wouldn't be surprised if AMD's Windows driver team took a look at how the open source driver developers performed some operations and ported them inside the Windows driver. Since both use differing code bases there will necessarily be a delay in such a port, especially on DX11 that is far more complex than Vulkan, but considering how many games ran better on the Steam Deck with Steam OS than with Windows, it wouldn't be a stretch that they took note of what bottlenecks were found by Valve and the community and decided to implement them.

This is exactly what happened, AMD has adopted PAL as an abstraction layer that sits between API and Kernel for D3D11, and possibly opengl.

 

[mitch074]

This is exactly what happened, AMD has adopted PAL as an abstraction layer that sits between API and Kernel for D3D11, and possibly opengl.

Then such a performance boost is no surprise, indeed - considering the 30-60% performance difference one can see between the Windows and Linux OpenGL driver for AMD cards (Doom 2016 openGL mode on a Polaris GPU), a 5-25% performance increase in DX11 is almost underwhelming, actually.

EDIT : someone on Reddit reported a 15% performance gain on Overwatch for a RX480, so it really does look like an overall driver optimization. Not that I care much about Windows drivers, but it really looks like "AMD gets better with age" verifies itself once again - it would be interesting to run a RX4/580 vs Geforce 1060 6Gb benchmark again and see how these still popular cards now scale.

 

[lothear]

Fine wine strikes again!