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.
Works just fine:
Based on their current work plan, an FPGA-based project board will be available in November "that serves as the development platform for a much less expensive ASIC-based solution (second quarter of 2006), contingent on available funding."
I don't know if they've been paying any attention (I presume they have), but FPGAs have gotten extremely cheap as of late. AVNet lists the Xilinx XC3S200-4VQ100C with the following rates:
1 - $14.7950
25 - $12.8700
100+ - $11.2200
While I don't like assuming, in this case it's fairly safe to say that the price would be even lower for quantities of 1000 or more. I see little difficulty with them being able to mass produce an FPGA card for ~$50 US. (Something of a sweet spot price point in computer the computer industry.) The only real reason I could see for going to ASICs is to reduce the cost of very large runs, and/or increasing the performance of the onboard chip.
Javascript + Nintendo DSi = DSiCade
Just to make it clear:
(1) The OGP product is OPEN ARCHITECTURE. It's intended to be compatible with open source SOFTWARE.
(2) There is a specific plan to make the "blueprints" to the hardware also available under GPL and LGPL at various points. ALL of the IP and schematics for the first product (the prototype board) will be open source.
(3) Hardware always costs money.
(4) This is a real product, being designed by experienced hardware engineers who have all the expertise necessary to do it. To the hardware designers it is not a "hobby".
Hardware is quite a bit different then software, being a physical tangible item that isn't easily copied/manufactured.
Tell that to Pad2Pad. I can send them a computer file, and they can send me back a complete board (or run of boards).
In fact, hardware has become closer to software than you think. Thanks to languages such as VHDL and Verilog, you can *code* a chip and test it without ever pressing a piece of hardware. And if you use an FPGA, you can litterally download the chip design into the processor and have a working model of your design.
If you ever hear about "chip IP", they're referring to the practice of developing a chip design and then selling the design to hardware manufacturers. ARM is a particularly well known exmaple of this.
Javascript + Nintendo DSi = DSiCade
yeah, but, just like open source, you can still change for the boards and open up the source, or in this case, building specs, programming code etc.
It would definately be interesting to have an fpga based board with the board programming code source available and the hardware specs available. That way, you could fiddle with your board and get it to do what you want, just like open source. It could be a viable business if they were charging for the boards themselves, but letting people play with the internal components a bit more than with proprietary. I can see lots of hardware geeks / hobbyists buying them just for the experience of playing.
...no two people are not on fire.
Actually, the product is targeted at the mass market. This includes Linux desktops, heavy workstations, and embedded systems. Being open architecture, it can be supported on all other platforms as well. Certainly, this market isn't as large as, say, the Windows market, but lacking another product as OPEN as this one, open source users are likely to prefer this, because the device will be fully supported by open source drivers, and it won't be a stability concern.
Also, a memory bandwidth of 6.4 GB/sec isn't all that slow.
The OGP is NOT a hobbyist project.
Do you have any idea how complex a GPU is?
Actually, they're not to bad on complexity. Most of the chip complexity comes from constantly pushing the boundaries of performance. Even then, a majority of the tricky work is actually done in the software drivers.
You're quoting prices for very SMALL FPGAs. What makes you think we could fit something as complex as a GPU into a 3S200?
A 3S200 is not that small of a chip. Fairly good sized processors can be written on it, often with quite a bit of space left over. Even if they do need a larger chip (e.g. a Virtex III) they should still check the prices. Xilinx has been making sure that their chips are extremely affordable in large quantities.
In quantites smaller than 1000? Well, it's difficult to get a good price out of ASICs as well.
Javascript + Nintendo DSi = DSiCade
It is highly unlikely that you'll be able to
program an FPGA to do something faster than a
modern computer can do.
Now that's just nonsense. This is the thinking of "More MHz is better". The truth is that a custom chip design targetted at a specific task can easily out-perform a more generic chip. For example, the SaarCor can render a raytraced scene many times faster than a Pentium IV, using nothing more than off-the-shelf FPGA hardware running at 1/300th the MHz.
That being said, it's doubtful that the OGP will outperform someone like NVidia or ATI who already build custom chips. But it might be able to give them a decent run for their money.
Javascript + Nintendo DSi = DSiCade
I am not involved with the project, but I would like to offer counterpoints.
1. The hardware will be underpowered because this group has little experience (if any) designing bleeding edge graphics hardware
It is designed to be underpowered. It is not going to support a billion triangles per second. It is designed to provide many basic features of the video card, and provide them well, in an opensource way.
2. The card will be overpriced because this group doesn't have the manufacturing clout of NVidia or ATI
The card is planned to be overpriced. They believe that the openness of the video card is valuable enough that people will pay for it.
3. The drivers will suck because nobody's going to buy this card and nobody will develop for it.
The openness of this video card means that if there is one kernel dev, and one X11 dev will have this card with time on their hands, it will be 100% supported.
4. The drivers will suck MORE because of all the trans-gamers out there who dual boot, they won't get the card because it won't be supported in Windows (or just very weakly).
The card is not aimed at gamers who play the latest games. It will not have the 3d performance necessary. It will probably have enough to run Tux Racer though
5. The company has no financial backing, so they will crash and burn early on and we will be stuck with abandoned hardware.
No the case, as the entire card spec is open. Even if all the original developers vanish, we will still have the specs for the card.
6. This time, effort and money would be better spent harassing the existing graphics card manufacturers into opening up their drivers, as least the non-trade-secret parts so we can do our magic on it.
It has been tried. A few came close: See Matrox around year 2000. The trend is that although linux is getting more support, more and more of the video card becomes closed an non-reverse engineered. We are currently lucky that nvidia keeps updating their drivers for the older video-cards, as the nv drivers (which are mostly reverse engineered / developed from tiny amounts of nvidia released specifications) suck badly. We are talking mach64 sucking less than nv. Sadly matrox g200 is probably still the most supported and reliable video card in linux. Yes it is better than nvidia, as nvidia drivers are quite famous for freezing X or kernel for no apparent reason.
(RenderAccel being the main culprit, but it happend with it turned off as well. Also without RenderAccel, 2d performance sucks)
7. (asbestos ON) I still don't think any Linux Distro in its current state should even be considered for desktop or gaming. But that's me being an elitist prick. Come up with a cleaner development model, make it "just work", and redo the whole windowing system into something that is NOT X, and maybe then we can start talking. The reason OSX works so well is because it does fifty backflips to almost completely hide the underlying Unix layer. It's not because I know Linux that I want to put up with its PMS all the time, sometimes it's nice to just click things with your brain switched off.
Now this is not a flame, it is just stupid.
Linux gaming is possible. Nvidia drivers provide a fast enough direct gl interface, that is adequate for pretty much anything.
Come up with a cleaner development model, make it "just work"
GL development already just works. Direct3D is not an inherently better system.
and redo the whole windowing system into something that is NOT X
X has very little to do with gaming, as most games use direct rendering or GL layers, and thus bypass X rendering anyway.
The reason OSX works so well is because it does fifty backflips to almost completely hide the underlying Unix layer.
The unix layer has absolutely nothing to do with graphics. All it needs is a basic terminal, and the darwins kernel just simulates the framebu
badness 10000
Open != Free
:P
Free != Open
GPL == (Open && Free)
Open Hardware == Open
Open Hardware != Free
In other words, a manufacturer could in theory create a card from an open hardware spec and charge for it. The idea being that said hardware would have specifications fully available. Further I would assume the hardware designer would require modifications to be made available.
If you've dealt with various Linux binary-only drivers in the past few years you'd know what the coolness was.
Hell, the coolness extends to Windows, too, as hackers could then modify windows drivers or create their own.
Yeah, I know, you were lookin for the +4 Funny, but some folks are going to read it and take it seriously
It is more productive to voice thoughtful opinions (reply) than to judge (moderate) others.
The full OGP core will fill most of the XC3S4000 (that's the second largest FPGA in the Spartan-3 product line).
:-) Some of the more interesting designs I've seen is when chip makers use microcode routines from memory to cover advanced instructions. This allows the chip design to use far less real estate than previously necessary at a slight performance expense.
Interesting. So when did they decide to switch from the XCS2000? (The chip listed in the spec sheet.)
Sure, you'd probably be able to make a *2D* core fit into a tiny FPGA, but a full OpenGL shader pipeline? Not likely.
Fully 1.3 compatible? No, but you could fake it in software.
If they wanted to keep it small, I see no reason why they couldn't use a similar design here. After all, at the end of the day a GPU looks a lot like a DSP. Add a few support chips (usually quite cheap) and they should be able to produce everything they need in a very small core.
Not that I'm suggesting that's what they do. Performance wise they'd get killed on this 3DFXish approach. Unfortunately, I don't have any pricing data on the 2000 and 4000 series (not that it means much anyway), but I'm thinking they could probably negotiate a pretty good deal.
Javascript + Nintendo DSi = DSiCade