Would A Kickstarter Open-Source GPU Work?

[teotwawki]

I would ..

What would be cool is a completely open-source computer (I'm talking desktop here, not a little Raspberry Pi), but just a GPU isn't going to do any good.[/QUOTE]

AFAIK, The GPU is the ONLY component of a computer remaining for which there isn't a fully open source choice.
There are already open source CPUs though not ARM or x86 compatible.

I liked the open-source graphics card a lot but the entry price was too high (very low volume). So perhaps with Kickstarters model of going big or going home might break the tipping point. Also open source (both hardware & software) have gained a lot of acceptance since then.

I'd particularly like the ability to re-program the FPGA at will, E.g. Litecoin mining at night.

[Shnatsel]

People don't seem to get that this is about open-sourcing the hardware and not just parts of software. This is unprecedented in the GPU market so far - GPUs have always been the tinkerer's agony. And having a GPU with open hardware might be epic - not only for education, but also for companies who need a modified GPU design and don't wanto to license someone's IP for insane amounts of money.

I don't get why they *ask* if it will be successful if they try and crowdfund it. Checking if there's enough interest and money to implement some crazy project is what kickstarter is about.

[dee.]

Open hardware? Sounds great. Let's hope they pull it through. The performance doesn't really matter at this point, this is proof-of-concept level stuff here so all they need to do is prove it's viable to create an open source GPU. After that, it's going to snowball from there.

[ssam]

There will certainly be plenty of people who don't care. If you look through these forums you see people who use the closed AMD and Nvidia drivers because frames per second matter more to them than open source.

There will also be people who care about open source, but are satisfied with current offerings from intel and AMD.

Then there will be people who follow the FSF hardline, and consider closed firmware to be a deal breaker. (see the very sort list of FSF recommended distros https://www.gnu.org/distros/free-distros.html ). The FSF refused to endorse the Openmoko GTA04 phone, because of a proprietary firmware, even though its the openest phone in production. So you should probably talk to them about your design.

Personally I'd throw some money at your project (like i did with the open graphics project). I think its great to see some open innovation in this area. Though I'm not convinced an FPGA is good for a final project.

What I'd really like to see would be a GPU based on the adapteva epiphany. Its already opensource, has small footprint and power requirements, looks like it can do the sort of calculations that graphics needs, scales to as many cores as you want, etc. I guess you just point llvmpipe at it and you have a graphics card.

[Drago]

There will certainly be plenty of people who don't care. If you look through these forums you see people who use the closed AMD and Nvidia drivers because frames per second matter more to them than open source.

There will also be people who care about open source, but are satisfied with current offerings from intel and AMD.

Then there will be people who follow the FSF hardline, and consider closed firmware to be a deal breaker. (see the very sort list of FSF recommended distros https://www.gnu.org/distros/free-distros.html ). The FSF refused to endorse the Openmoko GTA04 phone, because of a proprietary firmware, even though its the openest phone in production. So you should probably talk to them about your design.

Personally I'd throw some money at your project (like i did with the open graphics project). I think its great to see some open innovation in this area. Though I'm not convinced an FPGA is good for a final project.

What I'd really like to see would be a GPU based on the adapteva epiphany. Its already opensource, has small footprint and power requirements, looks like it can do the sort of calculations that graphics needs, scales to as many cores as you want, etc. I guess you just point llvmpipe at it and you have a graphics card.

Please, can you post link to HDL files, to actual Epiphany chip design. In their repo I found the board schematics, FPGA HDL files for IO and HDMI controller, but no actual chip design files. I tkink it is not as open as they claim it to be.

[ssam]

Please, can you post link to HDL files, to actual Epiphany chip design. In their repo I found the board schematics, FPGA HDL files for IO and HDMI controller, but no actual chip design files. I tkink it is not as open as they claim it to be.

I guess its just the parallela board that's open hardware. There is certainly a lot of documentation about the epiphany chip. But then again can you get the HDL for a Xilinx FPGA? Not the virtual chip being run on the FPGA, but the FPGA it self. I.E. if i owned a chip fab, could I build this open graphics card, or would I have to purchase an FPGA?

Still, if you put an epiphany on a PCIE board with a DVI/HDMI then you would have a fast GPU which would be easy to right all open drivers for with no binary firmware blobs.

[Mike Frett]

Having something that was Open enough to view all the Code, heck yes. The NSA can't hide back-doors in something like this. You guys sound like you don't care if some secret organization is taking away the rights that so many spilled blood and guts for you to have. That's sickening.

[timothyja]

Ok so I did so did some quick googling.

1 .For those (me) that dont fully understand what a FPGA is see: http://www.fpga4fun.com/FPGAinfo1.html
So the cool thing is they can be reprogrammed over and over meaning improvements in the design in theory would mean performance improvements without physical hardware upgrades (of course performance is still limited by the FPGA itself but still pretty cool.)

2. The only company I could find that made an OpenGL Graphics Accelerator was here: http://www.logicbricks.com/Product/D...&sifraCvor=409
Not saying this is the only one but its the only one I could find. They have a video demoing the technology here: http://www.logicbricks.com/logicBRIC...ideo-Clip.aspx

3. I found an interesting pdf from the Iowa State University a class where students implement OpenGL on a FPGA in order to run Quake. I wish this had been offered at my University back in the day.
See: http://home.engineering.iastate.edu/.../SteJon11A.pdf

[c117152]

Well thats exactly why crowdfunding has become so successful. Banks, Game Publishers, etc have no interest in development of ideas or outside the box thinking to the point of ignoring anything other than what fits their preconcived notion of what will make them money. The success of Kickstarter is a testament to the idea that people are willing to support ideas traditional financers are not. So I guess you are saying they are taking the right approach by considering Kickstarter.

This open GPU idea isn't "development of ideas or outside the box thinking". It's taking something we have on the shelf, from a half dozen companies, and reinventing it.
If this was an initiative to buy the patent rights of an established and successful GPU, and then release them under a copyleft* open hardware public license, I would be all for it. This way manufacturers would keep using it since it's already a success, but would also contribute improvements over time while keeping the tech accessible and affordable.
And still, the amount of money necessary to pull something like this off is quite substantial. AMD sold their own to Qualcomm for 64M (http://www.engadget.com/2009/01/20/a...lcomm-for-64m/) and it was considered a bargain at the time.
So, I stick to my previous statement, and add, "Hell No".

*With hardware there's no difficulty in taking and not giving back since you're not wasting time on maintaining a code base on your own, so a designe released under a permissive license would have no hope of seeing upstream contributions.

[yuumei]

I would back it just to attempt to persuade the other GPU companies to open source their drivers.

Would I be correct in thinking that once you got the fpga source you would be able to scale the performance of the gpu to the fpga it is running on. I.e if you had a larger fpga you would be able to fit more cores on it? That would be quite interesting as you could create low to high end gpu boards. Maybe that could be the tiers on kickstarter.