Interview: Timothy Miller, Open Graphics Project
Timothy Miller is the founder of the Open Graphics Project. He has a Bachelor's degree in Computer Engineering, and has worked since 1996 as a software engineer, developing graphics drivers and application software. In 1999, he started learning chip design on the job, because the kinds of graphics chips that his employer needed couldn't be bought from 3rd parties anymore. He has been designing graphics chips and graphics-related chips ever since. He is currently working on a Ph.D. in Computer Science, specializing in Artificial Intelligence.
When was the Open Graphics Project started, and whose idea was was it?
I guess you might say that the OGP was started in October of 2004, when I made my first post to the Linux Kernel Mailing List about the general idea. I'd had the idea in my head for a while before that, but only then did I get permission from management to start it as a company project. Since then, the project has amicably separated from the company where it started.
How many people were working on it in the beginning?
At the beginning, the OGP was only me, but many people in the open source community became very active in the project. Two other hardware engineers and I have created a company, Traversal Technology, to handle the business and hardware fabrication aspects of the OGP.
How has it grown since then?
The OGP grew rapidly after I first started it, and it's continued to grow gradually since then. Now, we have Traversal Technology and also the Open Hardware Foundation that's dedicated to the non-profit, community, and democratic aspects of the project. Our mailing list has over 500 members, including well-known Linux kernel hackers, 3D graphics experts, college professors, representatives of many other open source projects, hardware designers, and people who want to become hardware designers.
Is the design complete?
The first product of the OGP, the OGD1 board Open Graphics Development 1 (http://wiki.duskglow.com/tiki-index.php?page=OGD1), should have prototypes available in September. Those aren't ones we can sell - they're for debugging - but we can show them. The OGD1 board is an FPGA-based development platform. We'll use it to develop a graphics chip, but it's a product in its own right; many people will buy it to develop their own hardware.
Is there a market for it?
I'm not sure exactly how large that market is, but we seem to be creating a movement. Individuals and businesses interested in open hardware are coming to me all the time, asking about it. On a regular basis, we get a new person joining our mailing list asking about chip design and reading the Verilog (the hardware description language we use) lessons on our wiki (http://www.opengraphics.org). That movement may create its own market. There is also a pre-existing market for FPGA-based prototyping boards. In that market, we're priced dirt-cheap compared to comparable alternatives, so we should sell quite well there as well.
Can the OGP and OHF sustain themselves?
Traversal Technology should be able to sustain itself... It's all a matter of making enough profit from our products. We don't know if OGD1 will do it, but if we can produce a graphics ASIC (Application Specific Integrated Circuit, a fixed-function device), we can sell that into the so-called "high-end embedded" market (single-board computers, kiosks, etc.) and we can be very successful. The OHF will have to sustain itself on donations (some of which will come from Traversal).
How have you adapted to the constant changes in graphics technology?
Graphics technology has changed and it hasn't. Today, you don't find many new GPUs that don't have programmable shaders. But you mostly only benefit from programmable shaders when playing high-end games. We're absolutely not after the gamers. We're after the desktop, where many common 3D features are overkill.
We're "keeping up" by not having yet fully designed a graphics chip, although we do have a detailed spec and software model for one (http://wiki.duskglow.com/tiki-index.php?page=OpengraphicsSimulator). We've focused on the development platform, OGD1, and using future revenue from that to pay for development and fabrication of an ASIC. When it comes time to design the graphics chip, we'll review our past design decisions and make sure that our design will be adequate in features and performance.
How will it be priced and manufactured?
Right now, the only thing we are concerned about pricing is OGD1. We have some ideas for the ASIC, but we'll have to reevaluate our markets when we are ready. OGD1 is being priced based on competing products and by how much revenue we think we need to make to fund further development. For some people, OGD1 is a prototyping board. For others, it's a development platform for Open Graphics. For some of us, it's a fund-raiser. Fabricating ASICs is very expensive, and we need millions of dollars to do it. OGD1 boards will be sold in the $600-$700 range to OGP contributors, with subsidies from OHF donations for the needy. On the open market, it'll be sold for at least $1000 for single-unit retail, with bulk discounts.
What still needs to be done?
At the moment, we are finishing up the artwork for the OGD1 PCB (Printed Circuit Board), and we're about ready to get PCBs made. We also have to finish some basic bits of logic for end-users (controller logic for video, memory, PCI, etc.).
What do you want to accomplish, or what is the ultimate goal for you?
Originally, the ultimate goal was to develop graphics hardware that was friendly to free software, full documentation, and whatever else people need to know to write open source drivers. That has shifted to become the primary goal, with the long-term goal to be open source hardware in general.
What are your thoughts on the impact that the Libre license will have on competition?
It's hard to say what will happen with the Libre license. We'll be openly publishing the "source code" to our hardware. That means that anyone who wants to conform to the license (GPL, ultimately) can copy our design. If we're undercut on price, and we're put out of business, then there will be no future designs from us. One defense we have against that is to start with an open source hardware license that is NOT the GPL. This license would allow viewing and modification but not mass-production by others. Once we've paid for the ASIC development, we can convert the license to the GPL.
Other defenses we have are trademarks and official stamps that say things like "genuine Open Graphics hardware" or some such, which you can get only if you are a member of the project and contributing to its growth.
So, what's next? What will we see coming from the OGP?
What's next is the unveiling of OGD1, with luck in September. Then, we'll have something to show at trade shows and online. Soon after, we'll start taking orders. How many orders we get will determine subsequent courses of action. Another way people can help is to help us market OGD1. We also have other sorts of open-source hardware ideas, but, for the moment, we need to focus on immediate practical issues.
What about future technology and its impact the Open Source Graphics Project?
In the 90s, 3D was fixed-function. Today, it's programmable shaders (which the OGP will get around to doing, when it's practical). Tomorrow... probably real-time ray tracing. Fortunately, we know something about that too. If the OGP takes off enough to become a self-sustaining force, who knows what we can develop. I'm sure open source ray-tracer hardware will be something to be reckoned with.
How can people help?
Two ways. One is to join the project and help us develop and debug the things we're working on. The other is to donate money to the OHF (Open Hardware Foundation), so as to help developers acquire hardware they'll need to help with some of the work.
From an idea in Tim's head to an Open Graphics Project, much has been done to Tim's credit, but there is more to do. Having the OGP ensures that we will have an Open Source option for the graphics hardware and software in our computers. If you're interested in Open Source Graphics, then go to the OGP page and sign up, as I did. Get information, see what's going on, and find out if it's something you're interested in. An Open Source Graphics Card? That's pretty cool, don't you think?
Talkback: Discuss this article with The Answer Gang
Scott Ruecker lives in Phoenix, Arizona and is currently working on getting his A+ and Linux+ certifications. He works for a "Large Internet Service Provider" and plays drums in a rock-n-roll band every Saturday night. First exposed to FOSS when he heard about "This Linux Thing" in 2002. Got his start on the Fedora Cores. Has used SuSE since 9.1 and thinks he likes it.
