GPU Board with Vulkan Support


Nullifer

Still Fresh
Joined
May 27, 2016
Messages
7
Age
24
Are there any chances of getting a board that has vulkan support in the future? Vulkan really excites me for platforms like this as a developer. Also i'm curious about what factors went into the decision of the gpu if anyone knows.
 
The factors that went into the GPU selection were that it came with the CPU. Very few mobile processors have discrete graphics chips. It wasn't ruled out because it had some chance of decent linux support, and performs reasonably well without destroying the battery life.

There's no reason why Vulkan couldn't come to a Pyra CPU+GPU+RAM board, but it would depend on someone making a chip with suitable power consumption and being willing to sell it to ED. Your guess is as good as mine as to whether that will happen in the lifetime of the Pyra chassis.
 
The current CPU board is equipped with an OMAP 5 CPU, which comes with an integrated PowerVR GPU. There are several discussions about future CPU boards and why the OMAP 5 was selected. Long story short, it was the only one which could be purchased (due to minimum volume or other constraints).

It seems likely that future SoCs will have Vulkan support (a quick google suggests that PowerVR seems to support it with later GPUs). It will likely take a while though. The current units still need to be produced and shipped.

<edit>
1) Lest I forget, welcome to the boards Nullifer.
2) Damn ninja'ed by Levi :)
</edit>
 
The current CPU board is equipped with an OMAP 5 CPU, which comes with an integrated PowerVR GPU. There are several discussions about future CPU boards and why the OMAP 5 was selected. Long story short, it was the only one which could be purchased (due to minimum volume or other constraints).

It seems likely that future SoCs will have Vulkan support (a quick google suggests that PowerVR seems to support it with later GPUs). It will likely take a while though. The current units still need to be produced and shipped.

<edit>
1) Lest I forget, welcome to the boards Nullifer.
2) Damn ninja'ed by Levi :)
</edit>
Thanks, i've been a long time lurker.
That makes sense. I will probably hold out until there is a better option for a cpu board then.
 
That makes sense. I will probably hold out until there is a better option for a cpu board then.
I wouldn't count on that unless interest improves on the current version.
 
I wouldn't count on that unless interest improves on the current version.
I'm not financially able to get one atm, so I'm just going to hope the project does well enough to be able to lower the price point which in turn allows more people to get it and have the project become sustainable.
 
there won't be a lower price point unless there's substantially more interest (x10 or x100). (or if you can get a pyra second-hand for cheaper.)

but it's definitely important to keep the bank in the black; i understand that. i actually started saving for the Pyra about the time that i joined these boards. i don't have a huge "discretionary fund" each month, but over time it adds up.
 
I'm not financially able to get one atm, so I'm just going to hope the project does well enough to be able to lower the price point which in turn allows more people to get it and have the project become sustainable.
That's quite understandable. Rather than hoping the price becomes lower, I hope your financial situation improves in some way that you can afford to treat yourself to such a toy in the future. :)
 
That's quite understandable. Rather than hoping the price becomes lower, I hope your financial situation improves in some way that you can afford to treat yourself to such a toy in the future. :)
I'm just hoping these type of open source handhelds become more mainstream. I would love to develop crossplatform pc/handheld multiplayer games. I probably still will anyways when I can get one. I was hoping for vulkan, but i guess i'll just have to make sure my rendering is very separate from my gamecode, then switch over from opengl and learn vulkan when its accessible on more devices.
 
Vulkan? Really?

I'm not well up on the API, but it seems to be a bit... long winded. This is how much code is needed to draw a triangle, apparently:

https://github.com/SaschaWillems/Vulkan/blob/master/triangle/triangle.cpp
That seems like a bit much. Is it at least a 3D triangle?

Well its a 2d triangle but you view it from a 3d camera meaning you can look at the other side etc. Yeah vulkan is really low level but people will build libraries on top of it to make things easier. Also a lot of that code is probably stuff that you will abstract ontop of and only have to write once.
 
Well its a 2d triangle but you view it from a 3d camera meaning you can look at the other side etc. Yeah vulkan is really low level but people will build libraries on top of it to make things easier. Also a lot of that code is probably stuff that you will abstract ontop of and only have to write once.

God, I hope so. That's a ridiculous amount of code. I'm fairly certain I can draw a triangle in software in a smaller space. Note that the example I provided is already using helper libraries to manage things like flipping.
 
God, I hope so. That's a ridiculous amount of code. I'm fairly certain I can draw a triangle in software in a smaller space. Note that the example I provided is already using helper libraries to manage things like flipping.
As far as I've understood, there's quite a lot of boilerplate there. With a smallish library you can get to a much smaller amount of code for most use cases. For example, drawing a triangle using vulkano in Rust, without sacrificing (as far as I know) any of the flexibility: https://github.com/tomaka/vulkano/blob/master/examples/src/bin/triangle.rs

It's still a lot, but mostly initialization.
 
AIUI, the NVIDIA Tegra stuff would support Vulkan, but apparently they don't release specs so people can actually figure out how to use it on the hardware end. :(

Unusual for NVIDIA though, apparently they're doing upstreamed free software GPU drivers, and the MOQ is apparently only 60, so maybe there's hope for the future?
 
Oh, that's much shorter, yes. I wouldn't mind having to use that amount of code to render a mesh. Not at all.
 
Sarcasm.jpg
 
Vulkan has been announced to be standard in the next version of Android, so expect pretty much every 'new' SoC to support it.

And yes, it's a lot of boilerplate to get a single triangle drawn, but thankfully it tends to scale so that drawing a million triangles isn't much more work :)

Generally, Vulkan's advantage is to allow more control over what the GPU is doing, and lower CPU usage. If you're not CPU bound in the render thread, it'll be unlikely to give much in the way of advantages (though in mobile SoCs it's a little more complicated, as they're nearly always power limited these days, so lower CPU load can normally allow the GPU to clock higher, so 'saving' CPU cycles may be good even if you're already GPU bound)
 
Yeah, but all the extra GPU efficiency is for nothing if the CPU spends more time sending crap to the GPU than it does on game logic and audio combined...

Nah, I'm sure it's great, it really is.
 
Back
Top