Jawed
Legend
http://hardware.newsforge.com/article.pl?sid=06/06/16/2110211&from=rss
Jawed
Watch out for the prices
Jawed
Open Graphics Hardware
- Thread starter Jawed
- Start date
the maddman
Regular
Jawed said:http://hardware.newsforge.com/article.pl?sid=06/06/16/2110211&from=rss
Watch out for the prices
Jawed
OUCH. I would have gotten one if it was under $400, but.... Well, the optimistic guesses are now coming in way off, is anyone taking odds they'll make it to a fixed function card at all?
And what's the chance the card will be any faster than a Radeon 9000? Since that's about the last card with enough documentation to have an open source driver that does OpenGL.
arjan de lumens
Veteran
Dunno - given that they apparently haven't even begun on the actual 3d hardware core yet, I'm quite skeptical. Then again, they are showing a fair bit of persistence (which is usually the main part lacking in failed opensource projects), so I expect at least something to result from this project, even though I still expect them to fall short of their performance goals.the maddman said:
An FPGA-based card in the Radeon9000 performance range is not even remotely realistic. They may have a slight chance to reach that kind of performance if they can get to the ASIC production stage AND take enough care while desigining their core to allow for easy scalability. As for open-source drivers, there exists a GL driver for ATI R3xx/R4xx cards, however this driver is a product of ongoing reverse-engineering and as such not particularly stable yet.
JF_Aidan_Pryde
Regular
The feature set looks roughly like DX7. But what kind of work can only be done on this card but not on a cheap DX9 ATI/NV card?
Work by zealots who believe that if something is not 100% open it's completely worthless? Yes, I sound like a jerk, but am I wrong? The feature set is lacking, performance is lacking, it's ~15x as expensive as cards whose only flaw (compared to this) is an incomplete open driver (and a fully functional closed driver)... How many of these will they actually sell?JF_Aidan_Pryde said:
arjan de lumens
Veteran
Buggy unfixable closed-source drivers have been a source of endless headaches for a very long time, indeed for as long as closed-source drivers have existed at all - as you cannot check or modify what they're doing, they are a permanent threat against both security and system stability. AFAIK, most crashes in both Windows and Linux are directly attributable to buggy drivers rather than weaknesses in the core OSes themselves - this is the direct reason why Microsoft has decided to introduce a rather harsh driver quality rating scheme with Vista.The Baron said:
Isn`t that a bit of an overstatement?Permanent threat to security?Do I really want my drivers to be written by Joe Schmoe?Do I really want to lose time fixing my drivers by myself?Would I or Joe Schmoe be better then paid professionals who have considerably more resources dedicated to this?
JF_Aidan_Pryde
Regular
arjan de lumens said:
What kind of work will not tolerate an obscure driver bug but will tolerate rudimentary 3D features? For DX7 class effects, you might as well as do it on today's modern CPUs!
But there are open reverse-engineered R200 drivers for Linux, for one. Open R3xx/R4xx drivers are coming along, as you pointed out. Where is the market for this? $1000 for much lower performance than a $50 card where the only advantage is an open driver doesn't seem like much of a challenge.arjan de lumens said:
hoom
Veteran
Pros get paid to do Open Source too & who says that the entirely anonymous people who write closed source drivers are more competent than Joe? You can tell the competence of Joe by looking at his code but the closed source drivers well...
I'm not saying I'd buy this but I do understand & support the idea.
arjan de lumens said:
consider also replacing the phrase "closed-source drivers" with "hardware". add to that overclocking, unreliable hardware designs (no correction on busses, memory, crappy core logic implementations, ...). A significant number of bluescreens can be attributed to hardware issues like bit flips in memory.
Killer-Kris
Regular
Something that I've been wondering, but I don't believe I've heard an answer to yet, is why not simply implement the 3D part on the CPU kind of like Nick's swShader/SwiftShader, but only for OpenGL instead?
Obviously there will still need to be some sort of board because you need to get the signal to display, but it would seem like using the CPU as a GPU would be an all around better way of doing it. With the CPU unlike an FPGA there's no real limit to how much logic is used. Not to mention everyone who would be using this already has a CPU.
So other than the purely academic reason of implenting all of this in verilog and on an FPGA, why would they go that route?
Obviously there will still need to be some sort of board because you need to get the signal to display, but it would seem like using the CPU as a GPU would be an all around better way of doing it. With the CPU unlike an FPGA there's no real limit to how much logic is used. Not to mention everyone who would be using this already has a CPU.
So other than the purely academic reason of implenting all of this in verilog and on an FPGA, why would they go that route?
The founder of the project is an electronics engineer, so he doesn't really want to hear about software based solutions.Killer-Kris said:
Their goal is to sell enough FGPA based cards to manufacture an ASIC, then sell enough of those to create a second generation on a faster process, etc. So eventually they could outperform optimized software renderers (but still lack in features for some generations). In their eyes, software rendering wouldn't last long. Personally I believe that CPUs are rapidly becoming powerful enough to satisfy the needs of many 3D tasks except cutting-edge gaming. If Open Graphics is always going to lag behind NVIDIA/ATI's products then it's in the same market and any performance difference isn't going to convince people to spend hundreds of dollars.
Anyway, I don't think software rendering is the definitive answer to their problem either (although it certainly has long-term potential for a certain market segment). A better cooperation with NVIDIA/ATI would already make a world of difference...
arjan de lumens
Veteran
Killer-Kris
Regular
arrrse said:
That thought had crossed my mind as well.
Similar threads
