On Sunday 10 April 2005 23:49, Timothy Miller wrote: > On Apr 10, 2005 9:40 PM, Daniel Phillips <[EMAIL PROTECTED]> wrote: > > On Sunday 10 April 2005 20:44, Timothy Miller wrote: > > > What you get, when you buy the prototype board, is the same logic > > > as what you get from the consumer board, except that it's slower > > > and more expensive. Mind you, there'll probably be plenty of > > > hobbyist things you can do with either the OGP logic on it, as > > > well as your own. > > > > I completely grok the "more expensive" part. However, I will stick > > to my guns on the point that if it is not _too much_ more > > expensive, we will be wildly successful. Whereas if it is _really_ > > expensive, well, I have my doubts about the "wildly" part. > > The prototype board doesn't need to be wildly successful. That is > not its purpose. Its purpose is to be a testing platform for the > ASIC. However, since we have to produce it anyway, we might as well > sell it as a product. It's also a good thing to show to investors > who want to see a prototype. If there's a good opportunity to > redesign it later to make it less expensive, then we will consider > that, but that isn't in the plan, and it very well should not be.
OK, I've done quite a bit of informal research by now, and I can say this with some confidence: the _primary_ thing that is interesting about this card is that there is to be a hackable version. The open spec aspect is great, and it means a lot to you and me, but it just does not have enough pull by itself to move large quantities of a somewhat anachronistic design. The only thing that is going to really sell this card is to get _lots_ of the fpga version into the hands of people who will then help in various ways, including generating demand for the ASIC version. I think it's time for another poll. Prove me wrong with real numbers! Until then, I'll keep pulling in the direction of keeping the design cheap enough to realistically distribute several thousand FPGA cards. > We're going to TRY to fit it into the 2000, but if we cannot, we will > go with the 4000, because we now have the luxury of saving some time > by just letting the design be bigger. Well, based on my efforts to understand the whole design and resource requirements over the last several months, it looks like it will fit in the 2000, but at the expense of dropping down to one pixel/clock for some important pipeline configurations. I really do want to have the 3S4000 for this, make no mistake about it. But not if the board is 3 times more expensive and the Xilinx tools cost money. That would totally kill the enthusiast segment that will want the card for completely different designs than the 3D pipeline. How can I say it more clearly? The enthusiast segment is the segment that is going to make this project fly. > The primary objective is to > produce an ASIC for the embedded and open source workstation markets > that is affordable and which has open source drivers. We still don't see eye to eye there. You see the ASIC as where the real sales are going to be, and you're probably right, but not without the enthusiasts promoting and generating demand for it. > The FPGA board > is nothing more than an elaborate simulation and testing environment > so that we can do a much better job of debugging before we sink $1 > million into ASIC NRE. OK, _strong_ disagreement on that point. The FPGA board is much more than that. > However, there's nothing stopping us from slapping a 2000 on the > board (same package as the 4000) and selling it to you, even if you > can't fit the OGP design into it. Well, except volumes. Exactly. Now, I think the fpga card can sell about 5,000 copies if it ends up somewhat close to $200. Those sales will come because we can really tell a good story, and hey, it does display video. With a decent sized FPGA production run, we get into the market in a nontrivial way 6 months sooner, and with an army of developers instead of a small squad. (Sure, they won't have the Verilog, but they will be perfectly caplable of developing useful core logic for, e.g., shader units in the next design. And you are going to need _massive_ amounts of help to get shaders happening. That is just not a one man project.) > But then > again, we're expecting low volumes anyhow, which is the primary > reason why this board will be so awfully expensive. "So awfully expensive" will kill the project, Timothy. It has to be reasonable, and we can do it. > PCB's aren't all that expensive. There isn't much to strip off the > prototype, but putting on a smaller FPGA would help, and if we get > volume orders, that'll reduce the price too. Universities ordering > lab-fulls of these things will get a good deal. It won't happen unless the community is solidly behind it. > Oh, and believe me, I've always wanted to have something like this > prototype board to play with. When I've come up with weirdo designs, > I've had to make due with what was on our medical products. One > time, just on a lark, I coded up an ASCII dumb terminal, with font > and everything, and stuck it into the video post-processing FPGA on > one of the medical cards. Some people were also amused by my > flashing "TIM RULZ" overlays I inserted into the pixel processing > pipeline. Also, some time, I'm going to need a place to run the CPU > design I've been working on. I've borrowed a number of ideas from > Alpha, Itanium, and a number of other RISC designs. I figured when I > finished it, I'd release the RTL to the design under GPL. Well, how come you're having so much trouble extending your own personal experience to the rest of the world? There are thousands of people _just like you_ in that regard. We'll be able to get our message to nearly all of them, and they will respond by pre-ordering the FPGA card. I think we ought to try to pre-sell 5,000 FPGA cards (half the original target). I bet you large quantities of beer that we can pre-sell 5,000 FPGA cards, and without denting the ASIC market at all. It will enhance the ASIC market, actually. I described the details to some of the Red Hat folks here in Toronto, and a bunch of them wanted to buy the FPGA card right away. Not the faster, cheaper, cooler ASCI, but the FPGA. I think if you talk to some folks you will find that's not an isolated occurrence. > The WebPack doesn't support all parts. Xilinx already has Linux > tools. All we need is a Linux version of the WebPack. They seem to > have decided that they must differentiate the WebPack version from > the commercial version in some way, and the primary way seems to be > to limit the supported devices. I'm well aware of that, and I'm working on it. Regards, Daniel _______________________________________________ Open-graphics mailing list [email protected] http://lists.duskglow.com/mailman/listinfo/open-graphics List service provided by Duskglow Consulting, LLC (www.duskglow.com)
