I've got only one word for this...

Story: The Open-Source Graphics Card Is DeadTotal Replies: 5
Author Content
BernardSwiss

May 23, 2012
10:07 PM EST
RATS!!!

(Actually, a rather stronger expression comes to mind, but I can't say it here...)
mbaehrlxer

May 24, 2012
11:28 AM EST
patience, my friend, patience. it will come back, i am certain.

in the long term, since the first computers more and more sophisticated devices are being created by hobbyists. it isall a matter of the cost of components and the availability of affordable tools. just recently oshpark.com was launched. a site that will print custom circuit boards. this site may not be the first of its kind (i have no clue), but it certainly is not the last. and just like the 3d printers, circuitboard printing will be accessible to more people, opening up room for more development and the emergence of open hardware.

incidently, just today i attended a meeting here in beijing of openbrd. a group of young enthusiastic developers intent in making a completely open version a miniature form factor arm based computer. (you know like the raspberry pi or the beagleboard, but all specs, drivers, etc completely open)

so the failure of existing projects to produce such a board is a setback, but it's not the end. just wait!

greetings, eMBee.
Khamul

May 24, 2012
11:50 AM EST
No big loss, the whole idea was harebrained anyway. There's no way you're going to beat even a low-end Intel GPU with an FPGA, and it'll end up being far more expensive and power-consuming too. If these guys had the ability to have a foundry like TSMC build custom GPU chips for them, it would have been an ambitious but possibly worthwhile project, but Intel has the bar set pretty high with their ultra-cheap (frequently built into chipsets) and decent-performing GPUs which have fully open-sourced drivers. If you can't do better than that, there's no point in even trying; spend your time writing open-source software that'll actually be used instead if you want to make a difference in the FLOSS world.

This "open hardware" stuff really doesn't make a lot of sense until we have the ability to fabricate advanced ICs at home or in very inexpensive facilities rather than multi-billion-dollar fabs. The only exception here is highly custom hardware where it really does make sense to just go ahead and use an FPGA because the economics will never support building an ASIC.
mbaehrlxer

May 24, 2012
12:26 PM EST
Quoting:This "open hardware" stuff really doesn't make a lot of sense until we have the ability to fabricate advanced ICs at home or in very inexpensive facilities rather than multi-billion-dollar fabs.


right. but i am convinced that this will happen eventually. circuit boards today, ICs tomorrow...

greetings, eMBee.
BernardSwiss

May 24, 2012
1:37 PM EST
@Khamul

My understanding was that this wasn't, itself, intended as a consumer product, but as hardware for developers and enthusiasts to experiment with and work on design.
Khamul

May 24, 2012
2:29 PM EST
@Bernard: Doesn't matter; what's the end purpose? Just to be something for hobbyists to play around with? If so, then the FOSS community hasn't lost anything of substance, because none of them would ever use such a thing for anything important when you can get off-the-shelf stuff that's much cheaper and better and fully supported.

@mbaehrlxer: Not likely, not until replicators are invented. Chips have gotten more expensive to produce over time, not less. The only reason chips cost less is because of better technology and greater volumes; the initial start-up costs however are much, much higher than ever before. If you don't have $5Billion lying around, you cannot expect to get into semiconductor production, as that's what it costs now to build a single fab. You might be able to do things cheaper somehow (buy some 25-year-old fab somewhere), and that's fine if you want to produce some very low-tech ICs, but if you're trying to compete with GPUs, that won't work as those are all pretty close to the state-of-the-art in their manufacturing. Even hobbyist projects have to be self-supporting somehow; someone has to pay for the hardware to be made. Today, that means either using an FPGA (which are rather expensive) and a custom board, or just using an off-the-shelf Intel chip which is already built into every cheap laptop out there. If you're a hobbyist, what are you trying to do? Play around with 3D drivers or GPGPU programming? You can do that cheaply and easily with existing Nvidia or Intel chips. The Nouveau project can use all the help it can get, or if you want to see how some complete drivers look, go look at the open-sourced Intel drivers. Maybe this project makes sense if your burning desire is to learn how to design RTL logic for a GPU, but if you're an RTL designer, why don't you just go to work for one of the GPU companies? Obviously, there weren't enough people wanting to do GPU design in their spare time to support this project.

Additionally, even things like PC boards aren't really getting easier to make at home; if you're doing anything moderately fine-pitch, it makes far more sense to send your design to a dedicated manufacturer to be made for you. There are machines which can mill PCBs from blank copper boards, but that only works for 2-layer boards; any reasonably high-speed or complex circuit requires a 4-layer (or 6- or 8-) board, mainly for noise issues. I've never seen hobbyists figure out how to make those themselves, but for a couple hundred dollars I can easily get my 4-layer design made for me. It's true, hobbyists are able to do things now they couldn't do 10+ years ago (they've even figured out how to solder BGAs at home, though not terribly reliably), but they're always far behind the state of the art. I fail to see the point in expending a bunch of effort to try to build something that matches the performance of a circa-1999 Nvidia GPU and probably costs more than the highest-end Nvidia card now. As I said before, if you want to get involved in open-source GPUs, volunteer to work with the Nouveau team. They can use all the help they can get, and that work will benefit a large portion of the FOSS community. And your startup costs will be cheap: all you need is an Nvidia card, and you can get one of those for $20 now (GT100 or GT220).

Posting in this forum is limited to members of the group: [ForumMods, SITEADMINS, MEMBERS.]

Becoming a member of LXer is easy and free. Join Us!