Khronos Debuts OpenGL ES 3.2 & New GL Extensions, But No Vulkan This Week

[johnc]

OpenGL isn't gaining ground because it is tied to anything. It is gaining because it is available on all the platforms. Metal is available only on OSX and iOS. DirectX 12 is available only on XB1 and Windows 10. Why do those have futures but OpenGL doesn't?

It is available on all platforms. But it is not the preferred API for all platforms. Going forward it will only be used on Android and Linux.

Hardly anybody is going to release Vulkan games for Windows. And it looks like Vulkan won't even be supported on Mac, nor any other modern version of OpenGL.

[zanny]

Why do those have futures but OpenGL doesn't?

Because despite the platform lockin, and in the case of DX11 subpar features versus GL 4.5, developers are still by and large using them.

If we had developers writing once in well optimized modern OpenGL and criticizing platforms with subpar GL support we would be winning. As it is, as long as people don't care that their graphics APIs are locking them to one platform and imposing tremendous porting overhead to ever ship elsewhere, we have no room to grow. MS and Apple can bribe developers to use their shit and gain mindshare, but there is no big corporate entity with a vested profit motive in Vulkan adoption the way MS / Apple want developers stuck with them.

[Deavir]

Although OpenGL / Vulkan are cross-platform, the big issue is that it is essentially tied to only Android / Linux. It has no future on any other platform.

OpenGL isn't gaining ground because it is tied to anything. It is gaining because it is available on all the platforms. Metal is available only on OSX and iOS. DirectX 12 is available only on XB1 and Windows 10. Why do those have futures but OpenGL doesn't?

[johnc]

Although OpenGL / Vulkan are cross-platform, the big issue is that it is essentially tied to only Android / Linux. It has no future on any other platform.

[Ancurio]

I wonder whether it's worth it adding tessellation to Android. I remember when they were announcing DX10/D3D 10, it was being sold as a method of creating geometry almost for free, or at least with using a disproportionally small amount of resources compared to rendering using polygons and other <= DX9 methods.

But as far as I can tell that sales pitch has not been realised. Look at the clusterfunk that was the DX10 patch for Crysis 2, performance took a HUGE hit in implementing tessellation. And while I appreciate that may have been an extreme case, I can't see other implementations being good. I'm pretty sure Arkham Knight uses tessellation quite a lot, another mess on PC.

I'd like to know how phones, which are roughly between Xbox and 360 power right now for flagship handsets, are expected to take advantage of a feature that, so far, is bringing modern GFX cards to their knees. Correct me if I'm wrong. Maybe they're just laying the groundwork for the future, we all know how long it takes for ES versions to be utilised, most things are still using ES 2 as far as I know.

Are you sure you didn't meant Direct3D 11? D3D 10 only introduced geometry shaders, tessellation came with 11. But yeah, these geometry-generating shaders are often a huge performance drain. I've heard that what many high-end engines do is run a tessellation/geometry shader once, capture the output via transform feedback (streamout), and then just draw that over and over again with a plain verg/frag pipeline.

[Ancurio]

Is it now sure that Vulkan will come to the PS4? Or are they talking about the Steam Machines?

No, nothing is confirmed. When they say "capable of running on OSX" or "on consoles", that's it. It means there's nothing in the specification itself which would inherently stop or make an implementation unfeasible on those platforms. Whether the vendors of those platforms decide to implement it is up to them.

For now, the only platforms which are confirmed to get implementations are Linux/SteamOS (Mir, Wayland, X11), Windows, Android 5.0 and Tizen.

[Ancurio]

Who are we? Anandtech article mentioned everybody to define their feature set... that sounds to me like Ancurio's Vulkan or Dungeon's Vulkan, etc...

"We" as in Linux gamers I guess. Direct3D12 is already in full marketing force, holding dev sessions and introductions and such. Anyone with a previous Win version can download Windows 10 and start hacking away with it. I'd love for Vulkan to become the dominant high perf 3D API on Windows, but the more Windows devs get comfortable with D3D12 first, the less will ever care about trying out Vulkan too.

We'll basically get back to the situation of "well, everybody on Windows is using D3D, so I guess that's what I'm gonna learn and use as well", treating OpenGL/Vulkan as that exotic thing you might have to learn for your Android ports.

[duby229]

Anandtech article mentioned Google to define feature set for Android, Valve will define it for SteamOS and who will do it for Linux?

Microsoft does not care, so probably GPU vendors will define theirs... probably will be the same for Linux

Whatever game engine designers implement. It's their choice. The best googl can do is define standards that they expect to be met, but game engine designers can implement whatever they are capable of coding.

[duby229]

Engines will implement Vulkan like they do OpenGL now. That in itself won't mean much. Developers would still have to build their games around that, and, as of right now, very few probably will. Most of the interest from game developers is in DX12. I think Vulkan games will be a lot like we see OpenGL games today... sort of after-thought ported versions.

It really comes down to the target platform. The only target platform of interest for OpenGL / Vulkan right now is Android. And the future of Android gaming is sort of... unknown.

Lets take an example of c compilers, they can compile binaries for x86 or arm, or whatever.... A similar condition will exist for dx12, vulkan, metal or whatever else comes out. They are all similarly low level.

[kaprikawn]

I wonder whether it's worth it adding tessellation to Android. I remember when they were announcing DX10/D3D 10, it was being sold as a method of creating geometry almost for free, or at least with using a disproportionally small amount of resources compared to rendering using polygons and other <= DX9 methods.

But as far as I can tell that sales pitch has not been realised. Look at the clusterfunk that was the DX10 patch for Crysis 2, performance took a HUGE hit in implementing tessellation. And while I appreciate that may have been an extreme case, I can't see other implementations being good. I'm pretty sure Arkham Knight uses tessellation quite a lot, another mess on PC.

I'd like to know how phones, which are roughly between Xbox and 360 power right now for flagship handsets, are expected to take advantage of a feature that, so far, is bringing modern GFX cards to their knees. Correct me if I'm wrong. Maybe they're just laying the groundwork for the future, we all know how long it takes for ES versions to be utilised, most things are still using ES 2 as far as I know.

With regards to Vulkan, my big wish is for an open-source radeon driver asap, would be cool to see. Hopefully we'll start to see uptake in Vulkan useage at a far accelerated rate to OpenGL / DirectX etc. It seems like an investment in the future to use Vulkan.