Zink OpenGL-On-Vulkan Benchmarks Are Very Competitive To Radeon OpenGL Driver

[ET3D]

> When looking at the CPU usage over the span of all the OpenGL benchmarks carried out, to much surprise the Zink OpenGL-on-Vulkan run actually led to lower CPU usage than with the RadeonSI Gallium3D driver.

Exactly this!

I wonder why nobody else seems to regard this as the most important metric, especially if it means that Valve can improve Steam Deck's battery life even further by switching over to Zink for OpenGL games in the (near) future...

It will be necessary to use a frame rate limiter (matching Zink and RadeonSI frame rates) to see if Zink is really more CPU-efficient, as lower frame rates would naturally use less CPU if they are caused by a GPU or pipeline bottleneck.

I somewhat doubt that we'll see Zink really being more CPU-efficient. Also I don't think it's all that relevant to Steam Deck, as I think most of its users will not play OpenGL games.

[andrei_me]

Michael what about unigine superposition test? Isn't it OpenGL based?

[ET3D]

Compared to the May test, it looks like there are some improvements, both for Zink and RadeonSI, and there are also regressions. It would be nice to see the results of both tests on the same graphs.

[darkbasic]

I think when RadeonSi will work with ACO, the gap between Zink and RadeonSi will widen again.

ACO on RadeonSI is clearly not going to happen anytime soon, so...

[user1]

ACO on RadeonSI is clearly not going to happen anytime soon, so...

I know, but that doesn't mean Zink should start replacing a perfectly working, super optimized GL driver, like many here are excited about. I already said in another thread that if Zink doesn't have a clear performance advantage over RadeonSi, it's pointless to replace it. If you look at DXVK, which is already mature and exists for more than 4 years, in some games it can slightly outperform a native driver (also depending on which native driver is it), but in others it's a bit slower. That's why I don't see why would Zink be different in this regard.

Personally, I'm much more excited about RadeonSi finally adopting ACO than all the Zink improvements.

[darkbasic]

I know, but that doesn't mean Zink should start replacing a perfectly working, super optimized GL driver, like many here are excited about. I already said in another thread that if Zink doesn't have a clear performance advantage over RadeonSi, it's pointless to replace it. If you look at DXVK, which is already mature and exists for more than 4 years, in some games it can slightly outperform a native driver (also depending on which native driver is it), but in others it's a bit slower. That's why I don't see why would Zink be different in this regard.

Personally, I'm much more excited about RadeonSi finally adopting ACO than all the Zink improvements.

It's not a matter of replacing RadeonSI, rather about avoiding writing new OpenGL drivers from scratch.

[user1]

It's not a matter of replacing RadeonSI, rather about avoiding writing new OpenGL drivers from scratch.

Regarding new drivers, yes, I agree. Zink even in its current state is already mature enough to remove the need to write new GL drivers that will take a long time to mature anyway.

Still, after reading some comments it seems some people really want to see Zink replace RadeonSi.

[Linuxxx]

If you look at DXVK, which is already mature and exists for more than 4 years, in some games it can slightly outperform a native driver (also depending on which native driver is it), but in others it's a bit slower.

Actually, it can also far surpass any native driver on Windows, especially with demanding DX9 games; most prominent example: GTA IV

[user1]

Actually, it can also far surpass any native driver on Windows, especially with demanding DX9 games; most prominent example: GTA IV

That's because vendors don't give a lot of attention to their old DX9 drivers these days. Also, the only cases I've seen it far surpassing Windows drivers in some newer games, is when comparing it to AMD's DX11 driver, which may point to inefficiencies in AMD's driver. When comparing it to Nvidia's DX11 driver on the other hand, I've seen like only one case where DXVK outperforms it.

[R41N3R]

As soon as compositors work on Zink (gamescope, kwin, ...) I would definitly prefer it over radeonsi. Getting 90% of the performance of radeonsi is good enough for me when considering that it works on Vulkan. DXVK is the perfect example why I would choose Zink on the long therm.