Slashdot Mirror
Open Graphics Project Looking For Funding
An anonymous reader writes "The Open Graphics Project was formed last year to create a free and open source friendly graphics card. According to this article on KernelTrap, the project lost their company backing a couple of months ago, but has decided to go forward with the effort with money from the developer's own pockets. The team plans to release a prototype card to the public in November, at which time they'll need to find $1 million dollars for the effort to continue." I continue to wonder about the Open Hardware projects but call me skeptical- people contribute to Open Source because it typically costs little more than time.
Hardware is quite a bit different then software, being a physical tangible item that isn't easily copied/manufactured.
While I do wish them well, I still have trouble seeing how this will really make headway.
I do know that if what they come up with is capable and affordable, as in the hardware won't cost me more then my current PC cost to build, I will give their resulting product a go.
If you ignore the other uses of a tool, does that make the tool less useful, or you less useful?
The market for this card is geeks, hackers and open source die hards.
Most will already have the latest kickass graphics card in a machine, so will NOT be interested in a lower performing graphics card simply because they can get all the hardware specs for it.
What they will be interested in is if it has something cool or kinky about it.
Such things would be... do the whole lot on reprogrammable fpga so people can really customise... provide some interesting DSP like four AL3101 chips or a sharc so it can do audio processing too.... make a low power version for tiny/embedded computers (put it on a gumstix board!).... put a xscale on the card so it's a computer.... provide interesting buffered IO so you can use it as a video signal generator...
It has to have a unique selling point over and above being open source!
Merely open-sourcing the drivers isn't enough. XGI and VIA have done that. In order to make open source drivers truly valiable, hackers have to be able to FIX them when there are bugs. That's very difficulty when the vendor doesn't release full specs on their hardware.
The OGP is based around open specs.
GPUs are not complex? Then why do we only have a very small number of companies making them? And, what tricky work is done in software? Shading? Bump mapping? All of the big functions are performed in hardware.
A 3S200 is not that small of a chip.
It is a small chip when you're talking about GPUs. Xilinx states that it contains 200,000 system gates. If you have ever worked with FPGAs, you'll know that typically only a max of 75% of the resources can be used if you would like to be able to route your FPGA and still maintain decent clock speeds. This leaves around 150,000 gates. At an average of 4 transistors/gate, this is equivalent to ~600,000 transistors. Compare this with the latest offering from NVidia and ATI, which are pushing the 300 million transistor mark. So, you need 500 FPGAs to get the equivalent resources (at a reduced horse power).
GPUs can NOT be programmed onto FPGAs. At least, not in an economically feasible fashion.
Nvidia and ATI have yet to really address the MythTV crowd with a passively cooled, inexpensive (who cares about 3D specs for their myth box?) AGP card that can do all the heavy lifting of decoding HD MPEGs.
pchdtv.com amd mythtv.org are pretty much the only places you'd need to "advertise".
You've got a community of enthusiasts that understand the point of open specs, are willing to experiment with hardware to "get it right" and aren't being well served by the incumbents. Sounds like a match to me...
In a rather funny coincidence, I and another student happened to program a GPU onto the exact FPGA you're talking about for a graduate seminar at UNC.
We got a very simple rasterizer and framebuffer, and that's it. We spent weeks optimizing to get it to fit on there and run at 50 Mhz. Had we added hardware division so that you didn't have to send actual plane equations, or crazy complicated things like matrix transforms, we would have had to have a whole 'nother chip.
Small Xilinxes are great for prototyping little designs or small modules but they're useless even as a full prototype chip and certainly aren't production chips by any means.
(P.S. Graphics chips are the second most complicated beast in your computer, after the processor (and if you've got an older processor and a newer graphics chip, it's probably not even second).)
...then you just don't "get it" at all--not what is possible in hardware engineering today, nor the philospohy behind Free (libre) and open systems.
Up to a certain complexity, fab services are available even to home hobbyists for a reasonable cost, and for large runs it is quite inexpensive. The REALLY big cost is in SET-UP costs to produce ASICs. Besides, fabrication costs are no different than for proprietary hardware--the licensing model for the intellectual property has nothing to do with how hard it is to physically build it.
Furthermore, even if the production model will be expensive to get going, these days hardware engineering is like programming--you don't sit at a desk taping out masks and such like they did when they made the 6502 processor. Its all source code in Verilog or VHDL these days. Therefore, if Linux can be successful then why not open hardware?
It is in the development/engineering where these cards can have an edge over ATI and NVidia--they pay massive dollars to hire people to design the hardware and drivers and lawyers to keep it all secret. This project has no monetary design costs. I for one don't even care if they don't ever produce a single card themselves, as long as they get the evaluation FPGA board and all the source designs/code complete. THAT is what is most important, besides having some manufacurers pick up the design.
Money is the least important part of this project. The industry is going to start stagnating now becasue the players are much too proprietary--by hoarding information and research they duplicate efforts and slow or stop development of interoperability standards. Insistence on keeping drivers proprietary hurts the software industry (particularly open projects and smaller proprietary competitors) and props up Microsoft.
Last but not least, an open design lowers the barrier of entry for smaller players and others who do not have graphics IP--right now card makers are at the mercy of two major players who design and make chips. If this project succeeds, many other chip makers can make graphics cards AND chips. Also, since the design is open, even if a chip maker discontinues or goes bankrupt others can use the design themselves. Widely licensing to many chipmakers is the biggest reason why the 6502 CPU was so successful--it was produced by MOSTek/Commodore, Rockwell, NCR, GTE, WDC, Synertek and many more. If Commodore hoarded its design and made all the chips themselves, do you really think so many computer makers, including arch-rivals Apple and Atari, would've stuck with the 6502 for so long if they only had one company--a sometimes competitor--to depend on for their CPU? Even if the 6502 was the cheaper option I doubt they would be comfortable with that. WDC and Rockwell also kept that design alive lonnger and improved it where Commodore wouldn't (CMOS version, added more defined opcodes, 16-bit extensions...).
If these guys play their cards right--especially if they can put out a few thousand GPU chips and get the ball rolling for others to jump on board it could revolutionise the industry and level the playing field for Linux and others on the desktop--and the more people on board the more rapidly the design could be improved. And unlike the case with the 6502, these improvements could be shared and standardised--and chip makers who contribute these enhancements can still have "first mover" advantage as an incentive to innovate.
If I was a well-to-do player in the Linux/open source community like Bob Young I'd certainly throw a few million their way...