One Million Dollars For A Shader-Based LGPLv3 GPU
Phoronix: One Million Dollars For A Shader-Based LGPLv3 GPU
A company is going to attempt to open-source their graphics accelerator 2D hardware design under the LGPLv3 license. Additionally, they claim for one million dollars they would be able to come up with a 3D shader-based open-source graphics accelerator...
$1 million is nothing for a project of this scale.
This. I don't see this project getting anywhere. The million would be far better spent hiring some full time devs to get the Gallium3D drivers caught up.
Originally Posted by birdie
I strongly agree - we don't need more devices and drivers, we need things to be completed and polished. With ARM platforms becoming more abundant and more games coming to x86, we've got some catching up to do.
Originally Posted by Kivada
From what I see... they *already* have similar GPUs they just need the money to roll it all together and release it as LGPLv3. The point of this is that for at least one device... the full stack would be open source. That IS definitly worth the effort. No legal teams holding back the specs it would all be out in the open.. and with continuing development in the open.
Originally Posted by Kivada
Nvidia drivers and many ARM drivers (zero interest in giving us docs) will never be as polished as this (fully documented) or the radeon drivers (they have docs at least) this open source GPU would probably be very ideal for many phone manufacturers. If it works its nothing short of a game changer.
Its also worth noting that the 2D drivers already exist... at the very least.
Also... depending on the projected performance $1million might be a steal.
Last edited by cb88; 10-08-2013 at 05:22 PM.
Just a thought ... but teaming up with humble bundle would be interesting. I wonder if it is possble to do that or if kickstarter would allow it.
I wanted to chime in. I'll be happy to answer questions or take abuse.
I'm one of the two co-founders of Silicon Spectrum. I was just coming over to post a link to FB and Michael beat me to it. I was wondering why our traffic spiked today.
We will launch tomorrow on 10/9. My partner Jim was the CTO of #9 and I was a lead architect/ designer, so we wanted to originally launch on 9/9, but we had to put it off due to a death in my family.
I think we have a plethora of advantages over previous open graphics attempts, but the main one is the 2D is 100% working in an FPGA and was used in a digital picture frame ASIC. It supports everything you need for a modern OS:
BLT/ Fill/ Pattern Fill
Lines (eline, line, patterned line)
text mode acceleration.
Display list processor for offloading the CPU.
Fully functional VGA option for backward compatibility.
It fits in a Cyclone II 25 or Spartan 3. For 2D speed isn't really an issue, but we run faster than we ran in the original 2D ASICs. >100Mhz for the Drawing Engine.
The 3D is functional, but has a few bugs. Also, as Michael pointed out, we intend to release everything LGPL, so it can be used in commercial or free projects. We have a stretch goal structure, so if we are successful at the lower levels, we will release something the community can run with. The 3D has beefier requirements, but can be implemented in a midrange FPGA (Cyclone V-110 or Arria V). We are currently prototyping on an Arria IIgx-95
The Unified Shader Model/ OpenCL is something we have looked at, but that is 100% new design. If we make the stretch, it will be implemented and released LGPL.
I hope you guys will take a look and spread the word. I'm working on a better website, but wanted to make sure it supported SVN/ Forums, etc, so I'm a little behind. I have been following up on Facebook and will be happy to follow up here.
Cofounder Silicon Spectrum
Last edited by fbruno; 10-08-2013 at 09:24 PM.
Reason: Really dumb spelling mistake
and whos gonna assemble the cards? whos gonna send to all over the world?? whos gonna update the architecture for modern shaders?? whats the performance of 2d and 3d?
its seems so wonderful having a opensource card, but its never be good for more than a year...
and for who? only poor manufuctures will be interested in it. guess who? nobody...
i think you need to contact manufactures and sell... make only the drivers opensource for fanatic developers correct the bugs...
The point of this is that you'd buy an off-the-shelf FPGA and program it with the GPU specs. And that strength is also it's greatest weakness - it's guaranteed to be a lot slower and more expensive than something you'd buy prebuilt.
Originally Posted by amagnoni
Although technically, if you wanted to start up another project manufacturing the cards this would be a good starting point.
If decent, it will end up in SoC designs to avoid paying per unit licensing fees to the likes of vivante or imgtec.
If the company builds its business on selling engineering consultation, it could be a good business.
Last edited by d4ddi0; 10-08-2013 at 10:56 PM.