| |||||||||||
Open Graphics schematics postedOpen Graphics schematics postedPosted Mar 7, 2006 17:13 UTC (Tue) by yodermk (subscriber, #3803)Parent article: Open Graphics schematics posted
Note that, as I understand it from the KernelTrap article last week, the "something" that will be out this year is just a development board with video hardware. So, that board will be useful for creating the graphics engine. Again, as I understand it. I'm suspecting that we won't see an actual card before 2008, but who knows ...
I really wish them luck. I'm holding out with my trusty Matrox G400 until something else compelling comes along with Free drivers.
Sounds like they're also having issues raising money. What I don't understand is why Red Hat and Novell and Linspire & such don't jump on this. Especially RH since they won't ship proprietary drivers. Can't they tell that the lack of Free drivers for modern hardware is a real hold-up for desktop Linux? The $2 million that the project needs is nothing for most of those companies!
(Log in to post comments)
Open Graphics schematics posted Posted Mar 7, 2006 17:30 UTC (Tue) by elanthis (subscriber, #6227) [Link] What real return on investment would those companies see? They don't sell hardware. They sell software. The existance of the OGC isn't going to at all be useful to a Linux vendor unless the hardware vendors start using it extensively, and the OGC has to compete against the likes of Intel and Via on the low end, and can't hope to compete against NVIDIA and ATI on the high end.
That's not to say that the OGC can't succeed. But the ROI is just not good looking for any Linux vendor that isn't also in the hardware business.
Open Graphics schematics posted Posted Mar 7, 2006 17:39 UTC (Tue) by yodermk (subscriber, #3803) [Link] Then perhaps Penguin Computing should help fund it.
An Open card does seem like a real boon to Red Hat though. As they get into pushing the Linux desktop more, such a card would really help their system to work better on more computers, given their refusal to ship proprietary drivers.
Even the for-pay distros like Xandros and Linspire could benefit. I suppose they have no practical problem shipping closed drivers at the moment, but their business is dependant on the video supplier, a similar problem to depending on any other proprietary software. What will they do if nVidia and ATI decide to quit supporting Linux?
Open Graphics schematics posted Posted Mar 7, 2006 20:41 UTC (Tue) by allesfresser (subscriber, #216) [Link] Except for the enormous goodwill it would create for that company, which shouldn't be overlooked. ROI can appear in more ways that just dollars. It also might function as a lever to encourage the majors (NVidia, ATI) to look more kindly on opening their specs.
In their best interests Posted Mar 8, 2006 9:25 UTC (Wed) by eru (subscriber, #2753) [Link] Besides if all the major Linux vendors sponsored the project, the cost per vendor would probably be small. Probably could be sunk into their PR budgets. Linux vendors often sponsor software projects, so there is no reason why they couldn't sponsor a hardware project that in the future could help reduce their headaches related to graphics software.
Open Graphics schematics posted Posted Mar 8, 2006 1:35 UTC (Wed) by drag (subscriber, #31333) [Link] Well... OpenHardware is cool.
Think about it, this will give people at schools or at home the ability to hack on hardware in such a way that would be very very expensive otherwise.
Some other items:
http://www.linuxdevices.com/articles/AT3888835064.html This guy built a networkable video camera that he designed himself and then GPL'd. It uses Theora hardware encoder that is capable of shovelling a 1280x1024 movie at 30fps down a normal ethernet wire. Again this uses a FPGA device.
Found this item:
http://www.rowetel.com/ucasterisk/hardware.html
Here is a open letter on why to open source hardware. http://blackfin.uclinux.org/forum/message.php?msg_id=211
And then last, but certainly not least by a long shot is Sun's OpenSparc project to create a fully open source proccessor!
I think some day it may be cool to have a completely open source hardware system..
Anyways, there is something besides 'being open source' that is very very cool about this project.
There is NO product like it. It's going to provide interested people with a fully documented PCI express card with a large reprogrammable FPGA chip onboard with 128 megs of fairly fast ram.
Nobody makes a product like that. It would be a boon to people doing research on CPU design on a small budget or students studying hardware programming languages and all sorts of stuff like that. Say if your using a Linux box for high end video editing.. you can use this large sized fpga as a graphics proccessor for encoding and decoding material at a high rate of speed with little cpu usage.
Or if your a computer scientist you could reprogram it to run calculations for something like protein decoding and setup a little cluster of these inside a workstation to do some very serious number crunching.
All sorts of nitch products could open up for it.
OF course for the rest of us it probably won't be worth looking at, aside from curiosity, until the ASIC version comes out. (It'll smaller, cheaper, less heat, and faster then the FPGA version)
If they sell enough of them to justify the cost of production plus a little bit of profit, who knows were the next version would go?
I mean there could be all sorts of weird stuff happenning...
For instance a return to software rendering. Imagine a Cell proccessor in your PCI slot with a small FPGA core designed to do some limited video processing and provide monitor output.
At 3.0ghz the Cell is suppose to offer around 200 Gflops of performance. That's about the same as your average Nvidia Geforce 7800 video card.
And seeing how the Cell is designed as a consumer-style item for embedded systems like the PS3, then the price on the item can't be all that expensive.. at least not when production gets fully underway. Especially compared to the 400-600 dollar price tag for nvidia's latest and greatest.
Plus you can just keep adding Cell processors.
Imagine running MESA libraries specificly designed and optimized to run on Cell on your native x86 or Sparc system. The performance probably won't be comparable to a 7800 card, but the quality of output and flexibility and compatability of the accelerated graphics would be much higher.
Also imagine if somebody wants to get realtime Raytraced graphics off of the ground? Porting Yafray to run on a system like that would make doing high end 3d graphic rendering a 'get up and get a cup of coffee' type situation instead of a week long ordeal.
Not that I think that this will ever happen.. It's a complete pipe dream. But its just me saying that were something like these open source hardware projects have lots and lots of potential.
Open Graphics schematics posted Posted Mar 8, 2006 15:50 UTC (Wed) by freeio (guest, #9622) [Link] >Found this item:>http://freeio.org/hardware.htm >Lots of little I/O boards that are GPl'd.
Yes I designed those and GPLed them. Yes they work as advertised, and are cost effective for what they do. But in the last six years, I have noticed that there has been nearly zero interest in any of it. Free hardware development is problematic, and there are several reasons:
1. There never will be free (as in beer) hardware. For any small project, the economies of scale virtually guarantee that a closed-source proprietary board that you buy will cost a fraction of what building a free board costs. The manufacturers are buying parts in huge lots and get good prices, but the individual builder pays full retail, when the parts are available at all.
2. Even given a full and complete GPLed design, building free hardware is not a task that most folks are really equipped to do. The circuit board is virtually always built by a PCB fab house, which is fine if you are in the business, but not so easy to arrange if you are not. The parts continue to shrink, and soldering them down is a challenge. Resistors and capacitors the size of grains of sand do not cooperate, and active components with hundreds of tiny leads are a mess to get down right.
3. No sooner is a part is designed in than it is discontinued. OK, a lot of parts are not that way, but most of my designs use programmable logic (CPLD or FPGA parts) and the commercial lifetime of these is short. I have several working designs for which the parts are no longer obtainable. This is not like software at all - when Cypress (for example) discontinues a part, it becomes unobtanium within a few months.
4. Most of the required development tools are proprietary, and only run well on proprietary operating systems. The GEDA tools have come a long way in the last six years, but they are of no help when dealing with the programmable logic which many projects require. You have to use the vendor supplied tools (Altera, Xilinx, Lattice - or the really expensive third party tools such as Mentor Graphics, Synplicity, etc.) in order to fit your design into the CPLD or FPGA which runs your logic. You can GPL the VHDL or Verilog code you used, but the tools are not free by any means. Having paid tens of thousands of dollars for tools sets over the last 20 years, and having these go obsolete/non-supported immediately sours ones taste for free hardware design.
5. The concept of free/libre hardware designs has not been an easy sell in the free software community. I remember corresponding with RMS six years ago about the concept of free/libre hardware designs, and at the time, he was not convinced that the GPL could even be used for hardware. My argument was that the GPL could be fully applied to any design document, and that the design was the important part, that the hardware is just like a binary file. He accepted that argument at the time, with his famous "Happy Hacking" blessing.
So sure enough I have a web site full of free/libre GPLed hardware designs, and I am proud of it. But quite frankly, if I had it to do over again, I would have stayed with the free software side of the house.
Cheers!
Marty
Open Graphics schematics posted Posted Mar 8, 2006 21:40 UTC (Wed) by drag (subscriber, #31333) [Link] Ah. Very good.
I think that if tools to create hardware become more ubiquitous then it could be possible.
Like if you could run down to the store and buy or maybe rent time on some sort of a inkjet printer that could print out layers of silicon and metal to build up a circuit board in a relatively inexpensive way based on designs you could download from the internet.
Of course then you'd have to figure out a sane way to sodier all the tiny things on that. And then if the tools to program FPGAs were relatively inexpensive and accessable..
Then maybe you'd start to see more interest.
The nice thing about software, I guess, is that the tools are cheap and acccessable. Also software is a community thing were you can take a existing code and apply it to lots of different purposes. Also software is modifiable and if you find bugs they can be fixed and then patches redistributed.
With hardware pretty much none of that is true.
Oh well. I would be 100% perfectly happy if hardware companies would simply release specs, for the software accessable portions, that would provide the nessicary documentation for writing drivers or whatnot.
Open Graphics schematics posted Posted Mar 10, 2006 13:26 UTC (Fri) by gyles (subscriber, #1600) [Link]
The tools to program FPGAs are often free-as-in-beer for the smaller devices - which is all you'd ever be able to fill at a hobbyist level.
e.g.
There are also low priced development boards available:
http://www.xilinx.com/xlnx/xebiz/designResources/ip_produ...
The barriers to entry are low.
Open Hardware still a minority interest Posted Mar 10, 2006 17:00 UTC (Fri) by wookey (subscriber, #5501) [Link] Having been playing this game myself for the last few years, originally with the Lart (site currently down?) and now Balloon. I have to agree with a lot of what FreeIO has said. It's a bit of an uphill struggle, and the compelling advantages of Free Software are a lot harder to see in Open Hardware. There _are_ advantages, but you have to care about them enough to counter the higher cost.An example of a group who do are educational institutions like universities, which care more about complete openness than most. So the Balloon has been taken up by The Cambridge/MIT Multidisciplinary Design Project, for example, who are promoting the whole concept of Free Software, Open Hardware, and Open Design tools in order to learn about computers and robotics and do 'cool stuff'. But in our case it is enthusiastic local companies that have made the thing possible. The Lart design was paid for by TU Delft research money, whereas Balloon2 and now 3 (see cool pics from last week's brand new hardware!) have been paid for largely by Toby Churchill Ltd (who use Balloons as the basis of their commercial products) with help from the University. Balloon3 has probably cost GBP 60,000 (100,000 $/euro) so far (prototype build), which isn't a huge amount of money, but it's beyond the means of individuals and very small companies. And you have to spend that sort of money every 3 years because the silicon changes. A large part of the problem is that the costs of making software versatile are low, and compile-time options to cut out the bits you don't need are not a big deal. In hardware, in order to make a board useful to enough people to make it viable a certain amount of generality and versatility is needed. But this costs in various ways - it makes the design more complicated, and thus slower and more expensive. And although you can do builds with different BOMs, the smaller each run is the more the boards cost. And the difference between a GBP 300 Balloon and GBP 100 proprietary board is enough to put most people off. We're hoping that Balloon3 will be More like GBP 200 and thus vaguely competitive, but it remains to be seen how the pricing works out in practice. The Lart was one of the last modern computers you could build by hand at home. Now that everything has gone BGA and the passives are almost invisible, self-building is simply not practical anymore, so Open Hardware building can only work through companies or institutions with enough money to get hardware built in the normal way. (Obviously for sufficiently simple designs it is still possible to build your own, but that's not the sort of design I am talking about here). And even then we are building batches of 100 (in the UK) rather than 1,000,000 (in China) which is why an Ipaq gets you a lot more hardware for your money. That shows no sign of changing in the immediate future. The other thing FreeIO mentioned is tools. Things have been steadily improving but hardware companies are still very jealous about their software and interfaces and don't really want to provide Free tools. Groups like OpenTech, Open Collector and Open JTAG (just the ones I have come into contact with - I am sure there are others) are all doing their bits to improve matters, but we need more people who understand this stuff involved, and probably a slightly higher marketing (spit!) to engineering ratio to get a bit more cash into the system. Just to end on a positive note, we _have_ managed to make a business out of this: not much of one, but we're still here, filling our niche. And being an optimistic kind of guy I'm still hopefull that Open Hardware will become more popular. We shall see...
|
|||||||||||