Slashdot Mirror
Open Source Graphics Card Available For Advance Orders
mollyhackit writes "The Open Graphics Project, which we've been following since it first started looking for experts four years ago, has just announced that the OGD1 is available for preorder now. The design features 2 DVI, 256MB RAM, PCI-X, and a Xilinx Spartan-3 FPGA along with a nonvolatile FPGA for programming on boot. FPGAs are reprogrammable hardware which means the graphics card can be optimized for specific tasks and execute them faster than a general purpose CPU. The card could be programmed for certain codecs to speed up encoding or decoding. An open hardware design means potential for better driver support. Of course you could always use the FPGA for something else... say crypto cracking."
The SPARTAN 3 is a hobbyist FPGA. Cheap, and a lot of gates, but slow. A Virtex 4 would've been nicer :)
They're probably going to own 100% of the high-price videocard market with that.
That's PCI-X, not PCI-E. The rest of the stats are also a retro-blast as well.
I'm not sure what kind of architectures you could really test with this thing. It has slower memory on it than is on my motherboard. I honestly believe you could write software renderers faster than this thing.
Done with slashdot, done with nerds, getting a life.
Pci and pci-x is dieing
Having recently taken a graduate class where I had to write my own shaders for OpenGL, it was neat to play with the video card on that level; however most cards are quite limited with what is open API.
This card, while too expensive for me, might spur some interesting projects - cypto stuff and Ray tracing come to mind. I hope someone does something great with this.
At this stage of their development.
This is not a finished product by any stretch of the imagination. These are prototypes. Back in the day prototypes were wirewrapped nightmares and they cost a lot more than $1500!
Well the original Doom (and engines like Chocolate Doom...) have a hard-coded 30fps limit.
Not the point. The processor is not designed for that. You stuff microcode and data into it from the normal CPU and turn it loose, and it does its thing WAY faster than the normal CPU can. This thing has no disk IO and would not run well at all as a general purpose CPU. It's like trying to go grocery shopping with a Formula 1 car.
crypto cracking
Really? I have friends who splash out $1000s on their hobbies, whether it is robots or R/C. This is a steal in comparison to some more expensive and consuming hobbies, especially considering the (underpowered but still excellent) FPGA.
If graphics programming was my thing, I so would get one. I am considering getting one regardless, if only to use it for ray tracing.
Flexible hardware + Good open source ideals = excellent product
Signature v3.0, now with 42% less memory usage.
you might be getting ripped off if you're paying $1500 for a Spartan-3 board.
I guess they don't really have the board volume to get low prices. But If you want a graphics card for $1500 that's probably less functional than an NVidia commodity card, I'm not gonna stop you.
OTOH, If you're interested in FPGA programming and a novice at it, you'll want to get a MUCH MUCH MUCH cheaper Spartan board (like 50 to 150). See http://digilentinc.com/ for good starter boards.
If you're serious about FPGA programming (or just willing to pay $1500 to $3000) you will definitely want to get a board with a Virtex or Stratix on board:
http://www.xilinx.com/products/devkits/HW-V5-ML501-UNI-G.htm
If you want to have it on PCIx:
http://www.xilinx.com/products/devkits/HW-V5-ML555-G.htm
You can also get FPGAs socketted for AMD's Hypertransport bus and Intel's FSB:
http://xtremedatainc.com/ (Altera FPGAs)
http://drccomputer.com/ (Xilinx FPGAs)
http://nallatech.com/
http://celoxica.com/
(some of these vendors also sell PCI solutions)
FPGA programming environments still mostly suck. it's a market impeded by proprietary standards and a whole lot of NP-Hard algorithms. We're working on it...
How are you going to emulate video cards that are undocumented enough to not even have existing open source drivers?
I don't have the kind of cash they're asking for, for a graphics specific FPGA. If they could tailor the board towards the FGPA market in general, I'm sure they'd find people interested in more than just it's rendering capability (me!!).
:D). If the card can do offline rendering efficiently enough to experiment with discrete pipelines (more gates = more fp precision!) I'd be a happy graphics geek.
I'm concerned about the shelf-life after I'm done tinkering.
I'd like an I2C bus, a few led connectors, and some magic so that I can connect a general purpose daughter board the FPGA's address bus (ie: implement USB, LAN, audio support that way). Every FPGA should be able to run as a Tanenbaum CPU by law!
As far as rendering goes I can't see an FPGA being as fast as an ASIC - propagation delay is going to hammer it, and syncing will be a bitch - but I'm still interested in what it can do offline (assuming I can get a vesa console
Good luck!
Matt
"I don't necessarily agree with everything I say." - Marshall McLuhan
At least, it doesn't appear to me that it would. The product page states that the thing is sold blank. Unprogrammed. Meaning it's not gonna run *anything* till someone programs the thing. Once someone codes it up to run OpenGL/Direct3D decently, maybe it could run 3D games OK; kinda hard to tell. The hardware in it appears to be top-notch, in terms of lots of high-speed ram. Can anyone give us any idea what kind of performance that FPGA can give?
It looks like, basically, this thing is a $1400 prototype that OEM's could use as the basis for a consumer video card.
Can someone out there who knows more about hardware design and fab than I do tell me - once someone has come up with decent programming for an FPGA, can non-programmable, cheaper, maybe even faster, chips be fabbed? I assume that is generally how the design process works - start with an expensive, programmable chip, get the firmware correct, then mass produce non-programmable chips that are much cheaper?
I'd love to see a graphics processor that could be programmed to create graphics that look like classical hand-drawn animation. I think you'd need to do some curve-fitting in X,Y,Z and T in order to achieve that. We all know there are cel-shaders and vector renderers that can render 2D stills that look authentic, but that's still a far cry from animating something that looks like a Disney classic, or like anime. Fitting 3D polys in X, Y, Z into curves in X and Y may be trivial, but figuring out how to turn data from X, Y, Z, T into X, Y, T is the real challenge for non-photorealistic cartoon rendering.
There is no way in hell these people can compete with ATI/NVIDIA. Have you ever been to NVIDIA? Do you have *any* idea how many really smart people they have working on these problems 60 hours a week?
This project would be so much better off reverse engineering Cuda to make an open source driver than trying to make their own graphics chip. Hell, even Intel is having a very hard time getting a high-end graphics chip to work, and they've got so many more resources than this project.
Open source software works because anyone can hack on it and produce comparable stuff with zero initial investment. Hardware does not work that way. There is just way too much of an initial investment required. Even with FPGAs it's too expensive, and you're way too far behind to start with.
These people are idiots to think they can succeed here unless one of them has a 90nm fab in his or her backyard. (Sorry -- this is qualitatively different than trying to write your own OS, which is done all the time in undergrad classes.)
There are literally thousands of FPGA prototyping boards on the market, many of which cost much less than this one. So while you could use this for other things, I can't imagine why anyone would spend the extra money unless they wanted to use the video specific features like the dual DVI interface. Furthermore, the purpose of the project is to develop an open source video card, and this card was created as a tool for those developers to experiment with.
So, it was created to prototype a video card, and it's only practical uses are real-time video (output) processing, thus it is a video card.
This sounds way more practical than the OpenGraphics thing. $1500 on top of having to find a PCI-X board? No thanks.
If you consider "ricejob" offensive, you need to step off the meta-racism bus. Would you prefer "horribly mutilated japanese-origin vehicle with weakly-bred owner" ?
Ricejob it is.
-Billco, Fnarg.com
If you can find an FPGA for which there are open-source development tools, by all means please let us know about it. Meanwhile those of us that want to get actual work done with FPGAs will make do with the closed-source tools.
People routinely appear in comp.arch.fpga saying that they will write open-source FPGA development software, but none have succeeded at that yet. Perhaps the underestimate the magnitude of the problem. Xilinx has literally thousands of man years invested into developing their tools; it's not something for which one person or a small team can knock out a functional replacement in a year or two.
I try to use open-source software as much as possible, and I release much of the software I write in my spare time under the GPL, but for certain problems, open-source software just isn't going to be practical in the near term.
If you really want totally open source drivers they'll be available for ATI cards in the not too distant future.
This graphics card when it finishes won't be "open source" because you won't be able to change it, it might have open specifications, and it might have a good relationship with the open source community, but the open source community is just as bad at maintaining a relationship with hardware vendors as hardware vendors are at maintaining a relationship with open source.
If you're willing to pay $1500 for your ideology that's your call, but I'd rather pay $500 and get a card that's substantially faster, and is actually programmed to do something other than diagnose itself and I don't really give a rats if the drivers are open or closed source.
Don't! and I'd say that to anyone. What they are offering is a FPGA dev kit, with nothing to put on the FPGA. Yes, they've done a board design, but that's really one of the easiest bits, especially as most firms that sell the chips give you sample designs that you can stitch together.
The HDL is the key to this project, and as far as I can see they haven't produced anything beyond very basic PCI and Memory Controllers (which I'd expect to be very low performance). I looked at the same code about 2 years ago (maybe more) and it's in exactly the same state now as it was then. I say this as someone who writes VHDL / Verilog for a living and was wondering if I should contribute, but I'm not interested in carrying the whole thing myself.
If this projects manages to get a framebuffer device up and running within 5 years I'll be impressed. I think the whole project is incredibly naive, and doesn't understand the scale of the project they're trying to do