I am aware that there has been alot of questions about using dedicated Nvidia or AMD GPUs with the Pi (I have asked a few of them). But ARM (mali core ) and Vivante among other companies appear to offer low-watt mobile GPUs that are equal to or slightly faster that the VideoCore 4 (VC4), and support OpenGL/OpenGL ES, Vulkan, and DirectX. Instead of gripping about the relativly poor performance of the VC4, is it theoreticaly possible to find a new working GPU and lisence the IP (Recompiled drivers or source code) to the Raspberry Pi Foundation to use in, say, Pi 4?

What are the actual issues with this? Other than drivers, will it theoreticaly be possible? For example, does a ARM Mali need a special port like PCIe x16 that is not on the Pi's CPU? (yes, yes, I know that the pi does not have any form of PCI on it, but it must have something like that, because it is connected to the VC4) Is it latency? Does the Pi not have enough RAM? Or Is it simply too expensive for the Pi Foundation?

I now have quite a bit of free time to research this and work on this, but I am only a hobby programmer with no technical training, so before I try to contact ARM or Vivante and waste my and their time, I want to see if there is one big show stopper before I start.

To be clear, my question is: Issues with PI and faster VC4 alternatives, just so that those people who bash me over the head for "not writing a prefect, cookie-cutter Stack-Exchange Question" might lay off this time, because somehow there is something wron with every single question I ask.

  • 3
    We are the wrong people to ask about licensing a product and the development of future versions of the raspberry pi, as we are neither the people who you would obtain a license from nor the people who develop the Pi. – Steve Robillard Dec 15 '18 at 7:52

The Pi has always been more about educational value rather than maximum performance. My pretty wild guess is that what you are proposing is possible but among the less likely options going forward. Eben Upton who started the Pi worked(works?) at Broadcom so the number of connections both social and technical involved there make it unlikely to switch to something else unless it offered such an unusual cost/power/performance benefit that there was little other option. Like any other change the benefit would have to be great enough to justify the cost which is someone doing the entire board design and integration to make the change possible.

  • thanks, that cleares some things up. I agree with you about the educational value aspect and the close connections with broadcom, but i would think that having a faster GPU would improve the overall usage experiance and such. if the GPU supported it, people could learn more advanced coding on it, like writing in OpenGL or Vulkan and using OpenCL. So if we are stuck with Broadcom, do they have a faster, compatable GPU that could be used? – lockheed silverman Dec 15 '18 at 16:43

Not the answer you're looking for? Browse other questions tagged or ask your own question.