In October of 2004, Timothy Miller [interview] initiated discussion about the creation of a free and open source friendly graphics card [story]. Since that time, discussion has continued on the project's mailing list [archive]. In a more recent announcement, Timothy talked about the future of the Open Graphics Project. Originally funded by his employer, Tech Source, he explains, "Tech Source is now officially no longer involved in the OGP. Arrangements with Tech Source are being finalized that remove any potential IP issues," going on to add optimistically, "the upside of this is that we have more flexibility and freedom in the design."
Going forward, the project plans to focus on the creation of a prototype that will be available for purchase, "the success or failure of this project will be dependent only on how much the community wants it." 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." He adds, "it will include all the necessary graphics and memory hardware, plus other things that we deem are vital to its success as a product in its own right, even in the absence of the OGP graphics core." As for the current status of the project, Timothy explains:
"We will develop the board design and minimal IP (PCI controller, etc.), build prototypes (for a few thousand dollars out of our own pockets), and put the product on sale with a minimum delivery time of four weeks, and we will build devices in minimum lot sizes to keep production costs down. Revenue from the sale of this product will then be used to attract investors and fund the next stage, production of the ASIC. The NRE alone for the ASIC is about $1 million, which means either we'll need that much in investment money, or we'll need a partner who is willing to foot the bill and share in the revenues. (That partner could be the open source community as a whole, in some form or other.)"
From: Timothy Miller To: ogml ,ogml-announce Subject: [Open-graphics] ANNOUNCEMENT: OGP Status and Future Date: Tue, 05/03/2005 - 18:26 You may recall back in February, I announced that Tech Source had changed their stance on the OGP. At that time, management decided that OGP would not be profitable within the time they wanted. They decided to put the project on hold indefinitely and pursue other priorities instead. But the engineers at Tech Source still believed in this project and decided to make an effort to fund and develop a business plan to resurrect the project either within Tech Source or with outside funding. A group of engineers have been meeting after hours and contributing their personal time and money to this effort. In addition, I negotiated to cut back my hours at Tech Source so I could spend more time working on OGP on the side. Unfortunately, before we were able to come to a conclusive result, the professional, legal, marketing, and business planning costs had piled up to the point that we felt it was wisest to cut our losses and develop a better plan. We have decided that we would rather spend our time and money on building building real products. At present, those who have chosen to pursue this project in earnest include myself and two other hardware engineers, all current Tech Source employees. In our opinion, not only is this a worthwhile project, it is also a legacy we can help create to have a positive impact for the open source community. Tech Source is now officially no longer involved in the OGP. Arrangements with Tech Source are being finalized that remove any potential IP issues. We are, therefore, pushing forward unfettered, except for the fact that it isn't our day job. The upside of this is that we have more flexibility and freedom in the design. Now, we've decided that it's time we put aside the business approach, and tie up any legal and intellectual property loose ends, so that we can finally start to do what we do best, which is engineering. We will produce a product that we can produce, the prototype. The success or failure of this project will be dependent only on how much the community wants it. We have developed a work plan that we believe makes development of the OGP hardware feasible within our time and money constraints. Although the work plan includes development of the ASIC and final retail board, the immediate work focuses on the prototype board as a product. This may result in some changes and compromises that make the prototype somewhat of a more general-purpose device. It will include all the necessary graphics and memory hardware, plus other things that we deem are vital to its success as a product in its own right, even in the absence of the OGP graphics core. We will develop the board design and minimal IP (PCI controller, etc.), build prototypes (for a few thousand dollars out of our own pockets), and put the product on sale with a minimum delivery time of four weeks, and we will build devices in minimum lot sizes to keep production costs down. Revenue from the sale of this product will then be used to attract investors and fund the next stage, production of the ASIC. The NRE alone for the ASIC is about $1 million, which means either we'll need that much in investment money, or we'll need a partner who is willing to foot the bill and share in the revenues. (That partner could be the open source community as a whole, in some form or other.) There will be two models of prototype board: (1) A 3S4000-based board ready for the OGP core, and (2) a less expensive 3S1500-based (synthesizable with the free Xilinx Webpack) board that is purely for hobby and FPGA-lab projects, not suitable for OGP. Obviously, the work schedule has slipped. Rather than an expensive FPGA-based graphics board (in June), our first product will be an FPGA-based project board (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 have included a URL for the work plan (below). It is preliminary, conservative, and up for debate. I will later provide more explanation as to what parts we are doing and what parts we hope will be done by the community. We apologize for not letting the community know sooner but, since February, it has been a time of uncertainty. Things changed so much and so often that we feared that sharing it would do nothing but create fear and doubt that might jeopardize the whole project. The remaining members of our team share a united vision and a common goal. Hopefully, by November, this will result in the availabilty of hardware that the community can help develop into a workable Linux graphics solution. An Gantt chart of the work plan is hosted here: http://opengraphics.gitk.com/ogplan.jpg
Call me silly...
This is something that benefits any distro, and OSDL as a whole, if a business plan has been gotten together there's a neverending amount of places this could be pitched to.
I'm still in - and you should be too!
I thought it was a great idea before, and I still do. If the card can be kept down to around $200 I would still buy one, or several over time, to be free of all proprietary encumberances. It's time to put up or shut up. Everyone talks a great game but who's willing to pledge $200US right now? Tim, would it help if we could start backing you today?
Hopefully in the near future we'll be able to begin a project to build a "trust free" computing platform. If I have to run Palladium or whatever the equivalent is I'll stop using my computers at home and turn them into floorlamps.
I'm there
In case the cards will be available in Europe, I'll be among the first ones to order one.
Wouldn't this be a great use for the...
... Cell Processor? Aren't the SPE's on the Cel processor somewhat similar to the 3D processing chips on nVidia's and ATI's graphics cards? Would it be possible to have an OpenGraphics Video Adapter that had a Cell as it's GPU?
remember blender ?
Well, the blender project managed to raise 100K Euro (~USD, at the time) in 7 weeks,
if I remember things correctly.
Maybe something similar will do.
Oh, and I'll buy at least one card, if it's $200 or less, plus shipping to Europe... :)
What does this mean for me?
I might support it, I might not - I can't tell. I'm just asking, why would I buy this?
Does this just mean that I can have an mostly-IP-unencumbered video card? How much $?
How is this different than getting a $50 card that is well supported?
I can get a proprietary card that "mostly just works" now at a decent price. Can OGP offer me anything tangibly better than that?
It's going to have to be a)cheaper b)more capable or c)more dependable to go anywhere, IMHO.
I can get a proprietary card
Yes, they can offer you a card that's "just free (as in freedom)": open specs, open drivers. Manufacturers will be able to ship drivers in which ever form, and people will be able to write drivers for any operating system.
Whether that is tangible/important or not, depends entirely on you..
It will have tangible effects
It will have tangible effects. More stability, I assume. nVidia drivers work okay for some people, but for others, they are somewhat unreliable, and can bring down the system. Open source drivers would be less likely do this, as bug fixing isn't being done behind closed doors.
From the FAQ: > 4. I heard
From the FAQ:
> 4. I heard this card may use a FPGA, are we going to get the 'source'
> for what is on the FPGA also?
>
> Do not count on this to happen. It may happen, but definitely not
> from the start, and is not part of the goal of this project.
What's with that? They don't plan on releasing the specs for how the device operates internally? Isn't the whole point of open hardware that the hardware is _open_, not just the interfaces to the hardware.
I was thinking that this was a good idea, that it would be great to get ourselves away from proprietary graphics cards makers. But this project won't allow us to do that. Once these people finish and release their card, we won't be able to improve on that and produce a better card. All we'll be able to do is improve on the drivers for the card. That isn't a long term solution at all. The aim of any open hardware project should be to remove our dependency on proprietary graphics cards to graphics cards that we can develop ourselves, openly...
What's with that? They don't
You've missed the point of the Open Graphics Project. It's aims to be a graphics card with an open interface specification that developers don't need to reverse engineer. It's goal isn't to produce an entirely open source design.
Wrong
We already have cards with open interface specifications. We want a card with an entirely open source design, which apparently, this one is not. So much for that idea...
Please point to a single card
Please point to a single card which has an open interface specification, meaning one which is not bound by NDA or copyright.
ATI Radeon 9200
I hear that the ATI Radeon 9200 works quite well with the open-source in-kernel driver ('radeon'). I think ATI provided the initial versions, and the implementation is definitely not bound by NDA or copyright apart from the GPL. What better open specification can you get than a working free driver?
troll or what?
The Open Graphics Card will have drivers that are 100% open source. But that doesn't mean that the card specification i open source! The inner working of the card can be closed, and only the interface open. I guess this is also the case with 9200, even if the drivers and interface are open source
Radeon 9200
I have a Radeon 9200 and two 9000's. As cards go that are compatible with Free Software, they're very nice. The trouble is that you won't be able to buy them for very long, and many people think the level of support from ATI isn't enough.
Open Architecture
The specifications are open. They are ALREADY. The details of the internals of the chip will be described in every bit of detail that anyone would ever need, bugs and all. You, like the other poster, need to clarify exactly what it is that you think we're holding back.
suuure...
Yeah, like, 20-years old VGA cards. :-)
It's worth owning one of these cards regardless
The initial prototype card isn't going to be as cheap or as fast as the proprietary cards... but that's not what it's for. This is a card meant for Open Source land, with FOSS hardware + drivers, and in that area it's very needed.
For instance -- let's say you run Linux and FreeBSD on both an x86 and a Mac. Which card would you try to put into the Mac to get 3D in Linux or FreeBSD? Don't be so quick to answer... look first. You'll find out that you can get an ATI card for a Mac for OS X, but you won't find drivers for Linux. Nvidia don't have 'em, either.
There's LOTS of reasons why this is a very needed project and why I'll buy one when they're available. We need a card like this -- right now there are places where there aren't enough alternatives, so this work is more than justified.
The Gantt chart looks promising and I look forward to tracking the work of the OGP. As an EE I'd just like to remind everybody to keep in mind that the triangles always move to the right. [I.E. some minor schedule slippage always happens.]
I'd like to thank the OGP team for their work, their support and resources, they're recognition of the need for this hardware on the market, and most of all for their precious time.
- Chris
the VHDL source
Back in Dec'04, Timothy Miller stated that the VHDL source would not be available (at least not immediately) because Tech Source (sole founder of OGP at that time) believed that it would compromise profitability of the project.
Now that Tech Source is no longer involved in the project, what's Timothy's thoughts on profitability/source code? Personally I believe the project will benefit from releasing source code under an open-source license.
Releasing the code under GPL
I'm going to go into more detail on this later, but here's my totally unofficial reply:
My partners and I believe that the risks associated with releasing under GPL are greater than the benefits. There various ways in which we could be legally ripped off or undermined, and we would run off certain kinds of investors. Hardware design costs money, remember! We also believe that since we're already offering more than any other graphics vendor in terms of openness, the community should be willing to give us a little slack at a time when we've just managing to figure out how to make this project even feasible.
People say they want this project to succeed, but so far, only I and my partners have been willing to spend any actual MONEY on it.
Speaking of money, there is a certain, yet-underdetermined, amount of money that would offset the risk. There are other precedents for this sort of "buy-out", such as Blender. The difference between us and Blender is that, dealing with hardware, the risk, and therefore the buyout, is much greater.
We are focusing on a prototype board right now, something that is largely unaffected by the GPL issue, since all the initial IP in it will be under LGPL. This prototype will serve as the foundation for testing and debugging the OGP graphics core. When we have something tangible, there is perhaps a chance that people will be willing to make the buy-out, but in the mean time, it I doubt it would happen. The community has a window of opportunity to make that buy-out that is dependent on the timing of our funding for the production of the ASIC.
Remember, when you ask for something, make sure that what you ask for isn't going to destroy the foundation of you're asking for. Lots of people demand that we release the RTL under GPL, but those same people offer us nothing when it comes to figuring out how to deal with the risks. "There are no risks" doesn't cut it, because it's just not true, and "the benefits will be greater" doesn't help when no evidence is provided to support that idea.
In any event, the RTL will eventually be released under GPL anyway. It's just a matter of WHEN. And there's no telling what our business model will be like should we find ourselves fortunate enough to work on a second chip.
How does releasing the GPL so
How does releasing the GPL source make you any more vulnerable to being ripped off than giving the Xilinx bitstream? In either case a competitor can clone your product by simply making the same card and then loading your FPGA configuration into it.
I do understand that the GPL offers little protection to hardware designs though, which is why a license similar in spirit but different in mechanism is necessary for hardware.
Startups need to get traction.
Going from expensive FPGA to high-volume ASIC requires the source. Making the source GPL just means that you also have to provide access to the VHDL to anyone who buys the chip.
Throwing together an ASIC has a bit of NRE cost up front. If two or more guys build essentially the same ASIC, the one with the lower manufacturing overhead and better distribution channels wins. The startup that wrote the VHDL is unlikely to be that candidate, by virtue of being a startup.
I, for one, support the decision to keep the VHDL closed for now. Compiling an FPGA or synthesizing an ASIC happens far less frequently than editing and compiling code. You're not shipping bits, you're shipping transistors (at least in the case of an ASIC). Perhaps do like they did w/ Ghostscript, Doom and Quake: When you release major version N, make the source for major version N-1 available.
Why is the GPL not sufficient
Why is the GPL is not suitable for protecting HDL source code? Surely the principle of the source code being available provided recipients of products using that source code are also provided with the source applies just the same?
It's not about GPL
The issue here doesn't have anything to do with the protection afforded by the GPL, but rather the issues that the GPL makes legal that could nevertheless hurt the business. There is a discussion on the mailing list right now where we're trying to resolve these problems. I strongly encourage you to read it.
Sweatshop
I will only buy the card if it is known that the cards where not made in some third world sweatshop to keep the cost down!
It should be known that the cards are made by people who recieve normal, good, wages for their work. Not lik a dollar an hour or something, not working more than 8 hours a day.
If this project turns out as so (no sweatshop, decent wages, working hours for the involved), I will definetly buy one.
Sweat-shops
No, the cards will be made in a sweat-shop consisting of myself and my partners, working long hours and weekends, and making very little money. Wanna pull up a soldering iron and help out? :)
Open Source Build
Will you release the design and component list as open source, with instructions and allow users to build their own - then you would not have to take the build or inventory risk and just allow other users to build it.
(It would be like releasing only source code and allow the users to make the binaries)
Obviously to make a profit you could sell the fully built version, and for additionaly profit lock people into a long term support contract :).
Sounds like an exciting project.
Cheers
Christian
What will be released
There's tons of stuff that's going to be released. The whole objective is to have an open architecture. I don't have a list right off of what will be released, but mostly it comes down to there being chunks of drawing engine internals that will be released at a LATER DATE so that we can ensure that we don't have our knees cut off before we can recoup our expenses. Once a certain volume is sold, we'll probably release everything, including the stuff that's useless to most people.
What's released
Most things will be released right off, just not some of the chip internals, until we've sold sufficient volume to cover the expense of fabrication.
Ah cool, hehe :) Sure I il
Ah cool, hehe :)
Sure I ill help you out, now ill just invent a teleportation device and off I go!
Ill definetly buy one.
Hey, isnt it part of the idea
Hey, isnt it part of the idea that people will be able to learn how the innermot of graphic cards work, so they too could be able to make it if they had the right tools?
I think the fault with all technology today is that its all specialized, we are using things we woudlnt be able to make ourselvs.. So a computer which it is possible for a user to build if its necesery, thats great.
Open Source Politically Correct Card
You have to be kidding me.
A politically correct open source card.
Don't the engineers have enough to do already?
:-)
-gc
Verification of RTL?
Just out of interest, how are you planning to fully verify the RTL?
Verification is the most labour and compute intensive part of any design and I see no time set aside for this in your project plan.
I would assume this would be an ideal task for some kind of distributed effort (e.g. like SETI@Home, but using Icarus Verilog or something similar instead)-- it's not the sort of thing you can run out of hours at work (I know our LSF cluster runs 24/7 verifying/regressing the companies IP so there would be no time to spare for personal projects).
Verification
The whole point behind developing the FPGA-based board is to do rapid verification of the design. Mind you, there will be plenty of simulation done, but there are countless things that are just much easier to test in real-time.
volume and price
I was contemplating myself hardware design at some point in time (different market though) and the thing is, if you want to lower the price, you must deliver a lot of volume.
No doubt, i like the idea, but i can't see too many ways to keep the cost down, at least not for the first version(s). Once/if the project takes off, sure, a large volume will automatically provide the means to pull the price down, but until then?...
Well, i guess if you find a VC to pour money initially into the project, that might help, but that diminishes the founder's share later on.
Where are the hard guarantees of release when terms are met?
Not to sound negative, (I really like the idea of a fully community accessible graphics adapter), but I've heard the things Mr. Miller says somewhere else before.
We need to protect our income, so we won't release everything right away. We'll release when certain conditions are met.
This does sound earily similar to what Transgaming promised in their first year after opening doors. They offered subscriptions, with the promise to release it all back to Wine, when they would get a certain number of subscribers to fund the development.
Transgaming did contribute to Wine (we're thankful for that), but their promise to release everything back to Wine, after reaching the certain number of subscribers, was never made good on. Transgaming saw the money flowing in and they decided not to kill the Goose that lays the Golden Eggs and simply stayed proprietary.
My question to the OGP development team is this; What guarantees does the community have that you'll release the Goose with the Golden Eggs, after the money starts flowing in? Don't get me wrong, I do believe that at this time you are sincere about releasing the whole enchilada after recouping development costs.
My question is, what legal framework is being built to ensure that this release is going to take place after the recoup of amount X has taken place? Times change, people change, projects change. What is making this project different than say Transgaming with their abandonned promise?
Is there a poison pill like Trolltech has with KDE E.V. that they will release QT BSD-style after a hostile take-over or a shift in the priorities that Trolltech takes with QT? Where is the escrow to ensure the time and money the community pours in is not going to be in vain?
Specious argument
The fallacy in your argument is that you are equating hardware with software. You can't do that. The sorts of things that Transgaming promised to release are SOFTWARE. For the Open Graphics Project, not even the tiniest scrap of software will be held back, EVER.
In addition, every shred of information on the internal workings of the chip will be fully documented, IN ENGLISH, from the very beginning.
So, to be clear, what is it that you think we are holding back? Be specific. I'll address it as soon as you can prove to me that you understand what you're asking for. And throwing technical terms at me isn't going to cut it. This idea that we're holding something back that should belong to the community or isn't open source enough, I believe, stems from a lack of understanding on exactly what is and is not being released.
I'll tell you something that is never going to be released: The circuit schematics for the RAM chips. I don't have them. I will never have them. Micron or Hynix or whomever we buy from is not going to give them to us. Do I care? No, because every detail of the operation of those chips that I need to know is clearly and painfully-well documented by the manufacturers. It's a standard DDR interface, and any variations from the standard are clearly explained. They do a very good job.
So, again, I ask, what is it that you think you want that you think we're not going to give to you? And why do you want it?
The code!
The code!
You are holding back _code_ that describes how those chips work. It does NOT allow them to be improved by anyone. It allows them to work. Hardware made by you - I can improve drivers for it. But I can't improve hardware itself, as you are holding code that describe it.
To compare to software:
You are giving me interface to interact with your program. I can write and improve those things that interact with your program, but I can't improve your program. Because you hold source to it.
Does that answer your question?
It answers part of my question
But, I ask you, if you have the code to an ASIC, how does that let you improve it?
And I also ask, of the sorts of hardware that you can get documentation for, like network cards, who gives you the code so that you can improve it? Can you get Linksys to give you the RTL for their network chips?
If not, why are you flaming me for not doing something that no other hardware vendor does either? Have we not already done more than any other vendor by providing incredibly detailed documentation on internal structure and logic?
Anyhow, the point is moot. We've come up with an approach that should make everyone happy, and one of the list members is writing a paper on it.
Afraid I want to take something?
I don't want any of your code, nor schematics, nor a fab to produce it. My question is, what do you offer in writing that lifts you above a standard B graphics card manufacturer that releases interface specs?
I just hear a lot of talk about protection of the developers, a lot of fuzzy talk about releasing "whatever it is you are now planning to release". Specifics in writing about what will and will not be done with the OGP knowledge (I dislike the term IP) is something that would offer clarity and something concrete.
I have this funny sensation that we are seeing the birth of just another Graphics card manufacturer. One that offers a little more info on their secret sauce, but also one that ultimately is proprietary and just targets the FOSS community because the Windows market is already occupied by nearly unbeatable behemoths.
My question really is one about clarity. Just state what you will do in writing. No more, no less. No elongated "we're in the process of fleshing things out, we'll see how far this will go, we are planning to..." Just set it in stone what you are all about, no frills, no blurry grandeur pictures of a golden future. Just what you are going to do and what makes OGP really different from other graphics card manufacturers.
Since you already stated: This idea that we're holding something back that should belong to the community or isn't open source enough, I believe, stems from a lack of understanding on exactly what is and is not being released. It shouldn't be much of a hassle to get us to quiet down. Just say what you are up to, without the fuzzy BS.
Why do I ask? Well:
In any event, the RTL will eventually be released under GPL anyway. It's just a matter of WHEN. And there's no telling what our business model will be like should we find ourselves fortunate enough to work on a second chip.
"WHEN" could be now or could be never, no specifics given as to what would make release happen. And there's no telling what our business model will be Precisely, you could be the next Nvidia.
Get it all spelled out in writing, hurl it on a public accessible webpage. Kill the mystery and get the fact out in the open. When you make a promise, back it with a contract. When you won't release XYZ, just state it upfront.
Yes it would kill the buzz of a Mythical Open Source Graphics Card, but I think the community doesn't need another Eazel. If the dry facts kill OGP, then it simply isn't worth it and it's a good thing it dies. If the dry facts support OGP then it just enhances it chances of succes.
Implementation question
Hello,
Will this card have the FPGA directly connected to the PCI bus, and will the PCI IP be opensource ?
I'm on the lookout for such a card, which would be terrific for reverse-engeneering other PCI boards (USB snooping is simple and can be done entirely in software even on windows; PCI snooping is impossible to ensure at the software level).
Please keep the card open enough so as to allow unforeseen uses !
JB
PCI core
(1) The Verilog code for the PCI interface is being developed in the open and will be released under the LGPL license.
(2) The prototype board separates the PCI controller into a separate chip (Lattice) from the 3S4000 (Xilinx). This way, you can reprogram the Xilinx all you want without affecting PCI state information or connectivity.