Slashdot Mirror
SGI Sues ATI for Patent Infringement
Ynsats writes "The Register is reporting that SGI is filing suit against ATI for patent infringement. The suit alleges that ATI violated patent number 6,650,327, "Display system having floating point rasterization and floating point framebuffering", which was filed in 1998 and granted in 2003, in its Radeon graphics cards. This is coming fast on the heels of AMD's announcement of the intention to buy ATI for $4.2B and it doesn't seem to be swaying AMD's intentions. AMD hopes to finish the takeover by the end of this year. SGI has also issued an ominous statement stating that they have plenty of intellectual property left and there will be more litigation to come."
If you can't beat them, sue them...
There's 10 types of people in this world, those who understand binary and those who don't.
So SGI has been reborn as a patent troll? Welcome to the party.
Sorry SGI, you are done. Your products are done, all of your competition can bring better products to the table. This isn't 1998 where you brought these trendy cases and we all "ooed" over them.
Bye bye.
Looks like the folks at SGI really did have a plan for restructuring. It's a shame that is' predicated on litigation rather than innovation. Although with the way companies are greedily gobbling up "intellectual property" (hah!), I don't feel too much pity for ATI.
I wonder if the cost of going forward with this suit will hamper SGI's ability to produce and market new products?
My guess is that they will, ultimately, lose this case, and that they will end up, in 18 months or so, filing Chapter 22.
Transistors and Beer!!
The SCO-iffying of sgi. I used to love SGI. I still love their old hardware, from Indys to Reality Engines, from the 4D85 I started on (before they gave fancy names) to the Onyx Infinite Reality that we ran virtual sets on in real time long before PCs could even think about doing this stuff, and the sgi's ran a lot of our live TV well into the PC era, doing a better job than PCs could years after the sgis were released.
But now it's over and sgi has become an office with a few lawyers, and this is what the call emerging from bankrupcy.
Stupid sexy Flanders.
Can't do business by developing, Building, and selling products/services? Thats OK, we have this great business plan, it can't fail! Business by litigation.
Power Corrupts,Absolute Power Corrupts Absolutely, leaving one person(group)in charge is absolutely corrupt.
No, RTFA. They patented floating-point frame buffers, which didn't exist before SGI invented them.
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
Anyone else remember they gave NVidia the same treatment back in the heady day's of '98? This is nothing new for SGI. "Rattle the cage, and try to stave off the end with another lawsuit." How did that last one work for SGI? Not so well....
The problem is that this patent fails the obviousness test about 100%. The patent itself, if you follow the link, says that "People have used floating point before, just in emulation because hardware cost too much. Now that hardware is cheap, we just do floating point rasterization from the framebuffer instead of through emulation."
I don't understand how the USPTO granted a patent that says "This method has been known for some time, but now we just have the capability to do it."
I'm all for granting legitimate patents (they do actually exist) but this one does not pass the sanity check.
"There are a dozen opinions on a matter until you know the truth. Then there is only one." - CS Lewis (paraprhase)
No, they patented the use of floating point math in graphics hardware.
Ofcourse, OpenGL (and probably a lot of others) was using floating point math in graphics way before the patent was filed, but that obviously has nothing to do with this.
Slashdot social media options: AIM, ICQ, Yahoo, Jabber and Mobile Text. Why no MySpace?
not because of any of this, mind you....just cause of the lack of decent PPC ATI drivers.
no i have not shot my gun in the air and gone 'Ahh!'
When I read about stuff like this, it makes me annoyed. Not because any sense of fairness or ethics (companies don't have morals), but because of the wasted resources. Litigation is money spent without any production at the end. You pay a bunch of bloodsuckers to fight another bunch of bloodsuckers and either you take money from the other guy or the other guy takes money from you, but the only people guaranteed to get paid are the bloodsuckers.
Imagine if the money spent on spurious litigation went into actual R&D, capital investment for fabrication centers, engineer salaries, hell even advertising. Anything but litigation!
But as long as there's an avenue to make money this way, you can't really expect companies like SGI to behave any differently. You're providing a way for companies that are no long profitable (either because they have no product, e.g. SGI, or because they have an antiquate business model e.g. **AA) to leech off of the market instead of exiting it. Of course they're going to try to survive and not just go quietly into that good night. So, while I'm annoyed at this behavior, you have to realize that it's intellectual property laws that are the problem. We need fewer and simpler IP laws. Of course, trying to get lawmakers to pass fewer laws is like asking a competitive eater to "take it slow", and that's not even mentioning that the bloodsuckers aren't going to be happy to see yet another cash cow disappear anytime soon.
How long will it take for public outrage to really grow until real reform is made?
-stormin
The Southern Baptist Convention has creationism. On Slashdot, we have porn.
SGI created OpenGL.
Perhaps SGI caught wind of AMD/ATI's new "Fusion" CPU/GPU combination?
Aikon-
"Those who can, do. Those who can't, sue."
There are some people that if they don't know, you can't tell 'em.
I seem to recall OpenGL being originally developed by SGI. I don't think prior art counts when the company in question developed the prior art.
Cool Jurassic CG wars in full HD !
makes me want to play with my SGI 540's Cobalt chip set in full glory !
bæ8Ã0sÃOE?5r©oÂÃ?âz:ÃÃAÃ?ÃOEÂ6fXÃ?]Â
I would definitely welcome another competitor in the video card market.
So they have been the first to implement an obvious idea (storing numeric values using floats instead of integers to get the kind of advantages one usually get by using floating point numbers instead of integers? OMG THAT'S BRILLIANT RUSH TO TEH PATENT OFFICE NOW!!1!).
Why should they be entitled them to reap benefit from other people managing to build a successful product using the same, painfully obvious idea?
I just hate the patent system for this.
SGI is late to the table to become a patent troll. If there's any lesson to be learned in the past 5 years in the tech world it's that a business plan built around litigation is no plan at all (unless you are a law firm, then you're basically printing your own money).
It's a shame too, SGI was a great company with some very good products too.
However, I would point out that it's not unexpected. One of the reasons that vendors of video cards don't provide hardware programming specs or open source drivers for their products has been for fear of litigation. It's been a prevalent rumor for years that many vendors feel that their products potentially run afoul of a bunch of patents and that's why they are so cagey with letting people understand how to program for their products and to get the best performance out of them. If SGI wins in this suit, expect a horrible blood-letting in the graphics adapter business and prices for premium technology to go up across the board.
IANAGCD but it still seems the obvious thing you shouldn't be able to patent. You got something encoded with integers, then you think: "Hey, now the cost of doing FP arithmetics is negligible, why not add resolution in the most transparent way by going to FP?". Happened to audio file and picture formats, maybe hundreds of times in the history of computing.
---- MISSING MISCELLANEOUS DATA SEGMENT --- [sigdash] trolololol
To any company involved in bringing to market saleable goods or services, the Patent Lawsuit strategy is like starting a Nuclear War. Only those who consider they have no future in the market will risk the inevitable mutually assured destruction from the counter lawsuits.
That is another reason why Lawsuit mushroom clouds rise over the remains of USA's Tech industries
What worries me is if this New Business model will have an impact on Open Source. For the past few years, SGI has been one of the major contributors to Open Source in general and Linux in particular, including thread handling, non-uniform memory access (NUMA), access control lists, file stream support, and lots of other things. Will this dry up? And, more importantly, will the SGI remainder turn SCO on us and sue for something they've already given away? I sure hope not!
SGI had some of the best engineers, best ideas and for a while the best products in the world. They deserve a comeback, but it hurts to see them doing it through litigation and not new innovation.
Regards,
--
*Art
SGI always poured the lions share of its income into research, and to the best of my knowledge they, even now, continue to do so.
SGI is the company that today has the very fastest Linux computer - the Altix shared memory multiprocessing family - available at any price, really a technological marvel because it runs a single OS kernel and has memory architecture which is truly phenomenal - it scales better than any other multiprocessing/clustering solution.
So any defense of their patents, however 'unpopular' with the video gaming set, should be welcomed because it could help a company that we really owe a lot to in many ways get back into the game. Honestly.
They would not be a 'patent troll'. Don't forget, SGI open sources a LOT of its technology. Much more than most other hardware vendors. Much more.
I used to work at NASA and our division was largely an SGI shop, and yes, they were expensive, but at the time, there was nothing else out there that was comparable in ANY way. You won't ever find me saying anything bad about SGI except maybe that it would be great if they were cheaper.
Why? Because they are the best.
that SGI would be able to come up with a working business model and pull themselves out of debt and stay around a while, if it means they're going to sit on their IP and sue the bejesus out of everyone, I wish they'd just go away. You hear me SGI? Innovate or die means innovate now, not 20 years ago.
SGI is the market leader in high performance graphics.
Someone makes cool 3d video game with a VGA.
SGI laughs, continues selling workstations for $10k.
Someone releases a commodity 3d graphics card.
SGI laughs, continues selling workstations for $10k.
Someone releases a fast commodity 3d graphics card.
SGI laughs, but to placate the market, throws half-hearted PC graphics effort over the wall (Fahrenheit, x86 workstations, etc.) Effort is severely overpriced due to SGI's existing value network/cost structures. No one buys it.
SGI thinks little of it, decides to let the commodity vendors have their razor thin margins, they're doing them a favor by leaving all of the fat deals to them, right?
Commodity 3d graphics vendor offers lucrative deal to SGI top talent.
SGI top talent, looking for new and exciting and more money jump ship.
SGI, instead of getting the message, continues to focus on moving up-market and ignoring commodity markets.
Commodity graphics grows into a dozens of billions of dollar market.
SGI participates in none of it. Dies instead.
Clap. Clap. Clap.
Passing enough electricity through a wire so it glows to provide light is obvious - AFTER someone else thought of it. Its so obvious someone else did and Edison bought the patent off them.
If AMD can buy ATI for $4.2B, can't they simply add a few bucks to buy SGI too?
They have much in common.
They offered fast hardware with a robust OS.
Now you can have an equally robust OS ranging from 0$ to less than 500$ (whether you choose Linux/*BSD or MacOS/X or Win); most of them have hardware accelerated OpenGL, etc etc. Linux, MacOS/X and BSDs are even unix-like and posix compliant (most of them at 0$). Intel and AMD offer multicore processors at a wonderfully low price (at least if you compare them to the system SGI was selling..); don't even mention graphic hardware from nVidia and ATI.
When you are put in this situation (which is common between previous *nix vendors) you either 1) embrace a new business method (IBM/Sun/Novell) or 2) sue who trampled your market (SGI/SCO).
Perhaps because, when they did it, the cost of doing FP arithmetic was phenomenal? Because at the time, the idea was madness, not obvious?
SGI hasn't exactly been doing so well lately. One of the guest speakers in business school told me straight out: have good legal representation because, when the chips are down, that's all you may have left. All this lawsuit is is a manifestation of that principle and the "cornered animal" principle.
Their patent is a little more specific than this. A company called Chromatics had floating point hardware in their CX 1536 graphics engine back in the 80's (so named for it's 1536 x 1152 resolution) and used floating-point to represent the initial coordinates. What they didn't do, but SGI does, is store computed pixel attributes in the framebuffer in floating point.
The problem is that this patent fails the obviousness test about 100%.
If it's so obvious, why didn't you do it first?
Implementing floating-point framebuffers is non-trivial problem, and SGIs solutions to doing so are why they deserved the patent.
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
I don't even think so-called legitimate patents exist. I've yet to see a patent that isn't anymore than an extension of an idea using a previous technology and applying it to a more contemporary one.
The idea of changing a software to hardware method isn't anymore impressive than a caveman changing a written communique into a verbal one. "He wrote 'ugh' but I was first to say 'ugh' so I patent it!"
Looks like SGI is the next SCO :)
Except that SGI actually did develop its own technology, so maybe it has a bit more ground to stand on here than SCO...
I've yet to see a patent that isn't anymore than an extension of an idea using a previous technology and applying it to a more contemporary one.
Try looking up the patent on Xerography, then. People come up with novel solutions all the time, and many of them apply for and get patents on them.
-jcr
The only title of honor that a tyrant can grant is "Enemy of the State."
I am sure that if NVIDIA and ATI were to open their drivers or specs, it would make it much easier for companies with patents to go after them.
"No, RTFA. They patented floating-point frame buffers, which didn't exist before SGI invented them."
This statement is completely incorrect. Floating point point frame buffers have existed forever. Back in the day it was all done in software ofcourse. Essentially every pretty rendered image from the 90's or earlier that was not ray traced used them. The problem is that the initial 3d accelerators used integer based frame and zbuffers because they were cheaper and easier. SGI did not invent anything. They patented a hardware version of existing software methodology.
They are in fact patent trolls if they think this patent is valid.
As of today, they've announced that they've completed the merger. Now begins the integration of the companies in question...
I am not merely a "consumer" or a "taxpayer". I am a Citizen of the State of Texas
Edison's patent on the incandescent light bulb was ruled invalid.
Actually, no... All the technical data does is tell you how to drive the chips. It doesn't reveal all too much of how the silicon actually does what's being asked of it. If they've got any mojo that can't be revealed in that manner, they've got piece-parts of tech where they shouldn't be. Intel opened up everything except how to drive the Macrovision pieces of the chip (per their agreements with Macrovision and DVD CCA/MPAA...)- if it were Patent concerns, I assure you, it'd not been open sourced.
I am not merely a "consumer" or a "taxpayer". I am a Citizen of the State of Texas
I don't think SCO's case can be considered typical -- for one, when pressured to present specific claims, they don't seem to even have any at all! It's really quite unusual to sue someone without even presenting your OWN account of what you think is evidence, especially your biggest rival, one of the biggest companies in the world, and second only to tobacco companies for playing legal hardball. Deliberate suicide is something people do, watching SCO slit its wrists over and over again is something fairly unprecedented. That it's all a stock manipulation scheme is something Novell, and apparently now IBM are specifically alleging ... but that's a pretty atypical approach, even for patent trolls.
Done with slashdot, done with nerds, getting a life.
Aww Taco, you had a chance to use the old SGI icon again and missed it ;(
Poor thing rarely sees the light of day anymore.
...
... the honest hard working way.
Take em to court and sue for millions to keep your company alive.
I am sure now we do not want SGI to make a come-back. People here have mentioned that they have made interesting products in the past, but now that they are unable to compete, they will become a purely litigious company. They could not and cannot keep up, so they are effectively determined to hold everyone else back. This is yet another example of a failed patent system defeating what is demonstrably productive market economics. SGI, when it was king of graphics, failed to deliver the substantial improvements in technology at better prices demanded by the market, so the market turned elsewhere. That is precisely what should happen and it does not matter who came up with those developments in the first place.
Why bother.
Tags: amd, ati, sgi.
Media that can be recorded and distributed can be recorded and distributed.
-kfg
Except SGI could actually have legit patents.
SGI should have thought of spinning off it's graphics IP a long time ago. Take a look at ARM. They make nothing but IP and money.
See my blog http://ilovecookes.blogspot.com/ for light hearted technical information.
You are highly confused. The incandescent light bulb was one of the few patents of Edison's that was not overturned. In fact, Edison strengthed his case by buying off previous patents for similar work, making sure that there was an unbroken chain of Intellectual Property. So when Sawyer and Man attempted to challenge Edison's patent, they had to do so on addendums they added to their own patent application. The Supreme Court found that Sawyer and Man's claims were too broad, and that their addendum was an afterthought rather than core to their invention. Thus Edison's patent was upheld as valid.
Javascript + Nintendo DSi = DSiCade
Enough said.
So SGI is making cloudy suggestions that there might be more IP to blackmail over.
Someone seems to be really desperate to get bought by one of the remaining big players (most likely they are hoping for intel). So the whole company is really only worth the legal matches to light the other guy's factory.
[i have an opinion and i am not afraid to use it]
SGI is back from the dead and is now trying to feast on living companies. If that doesn't fit the halloween season I don't know what does.
to welcome our new rasterization and floating point framebuffering overlords!
You guys are really falling out of practice here.
If it's so obvious, why didn't you do it first?
Ok, so maybe SGI did it first. Then let's give them a nice gold-plated medal or something. Regardless of how much time and money you spent training for the competition, winning a race shouldn't give you the right to charge an arbitrary fee to anyone else who wants to run that course in the future (or prevent others from doing it at all).
Try this out.
The method to harvest wheat had been known for a long time but was handled in software (human + sickle). The mechanical wheat thrasher was invented and patented using the same method and concept but now in hardware.
The Intel chips are much less sophisticated than offerings from ATI or Nvidia (to the extent that Intel graphics chips are seen as something of a pariah by gamers ;-), so there is probably a lot less risk in that they are not doing nearly as much either in the chip or in the drivers. In the case of ATI and Nvidia I think a sufficent amount of code remains in the software due to the constantly-evolving and highly competative nature of consumer graphics card market (where it's common to see driver updates that allow existing cards to support the latest effects). If they crammed as much into hardware as was possible, that would reduce the flexibility of the platform quite a bit.
I forget which but either ATI and Nvidia (possibly both) have said the reason why they haven't open sourced their drivers is because they contain valuable intellectual property that could be to the advantage of a competitor (i.e. each other - I assume), and I'm inclined to belive them to a large extent. If that's the case, I think it's certainly conceivable that if it's possible to determine a bit more about how the system operates, that it would make it easier for those looking for both trivial and non-trivial patent violations to make a case, and that might be part of the reason why neither party has been inclined to released open source drivers despite both being sufficently aware of demand that they've provided binary drivers for Linux on x86.
Actually this is probably just their odd way of saying 'buy me too' to AMD.
Given that their licences to SGI's IP, is one of ATI's excuses for no OSS support, this may not be a bad idea. If AMD buys all of SGI's old IP and release good OSS drivers it is going to take a big swing at nVidia and regain some ground ageist Intel's open chipsets.
Do you have any facts to back up this claim? I know that some legislators are lawyers, but in the past whenever I've contacted my own representatives (which I've done many times) I've been struck by how utterly clueless their staff seemed to be. (And, by the way, Senator Allen R-VA is a lawyer who, judging by his staff and subsequent form letter, has no idea what net neutrality even means.) Furthermore, law is a specialized profession. You don't just have to be a lawyer to really understand this, you have to be a lawyer with at least a rudimentary understanding of technology who also specializes in IP law (as opposed to tax law, criminal law, family law, constitutional law, etc.) How many legislators fit that bill? I think you're drastically over-simplifying this case. I find that annoying because every time some wide-eyed fanatic claims "all the legistlators are lawyers and teh lawyers are all out to get us!" he or she takes away from serious analysis of a serious problem. No one is going to take you seriously if you sling around unfounded assumptions, and that gets in the way of people like me who actually want to change things instead of just be melodramatic about them.
In any case, I visited this site: Congress Merge to do a search of how many members of congress were lawyers. I first did a search for "lawyer" in the profession field. I got 6 hits (3R, 3D, if you're curious). Then I did a search for "attorney" and got 193 hits. Assuming no overlap, we've got 199 lawyers in congress (102D to 90R, if you're curious, and I'm assuming 1I). Congress has 635 members (435 in the House, 200 in the Senate).
So out of 635 members, 199 were lawyers or attorneys at any point in the career. That's 31%. Hardly enough to say, in my opinion, that "Legislators are by and large lawyers". 31% of those are lawyers, how many do you think are IP lawyers? Sorry to rain on your simplistic world, but it's more complex than just "teh lawyers are everywhere!"
-stormin
The Southern Baptist Convention has creationism. On Slashdot, we have porn.
The difference I see is that the complete idea for using a floating point framebuffer existed - just nobody could afford it. I would equally have a problem if someone tried to patent the idea for a mechanical wheat thrasher. Now, patenting a specific mechanical mechanism, or a specific method of implementing a floating-point frame buffer, is fine.
That's ultimate difference in my book: implementations should be patentable, but ideas should not. (Similarly: Information (such as a chemical formula or genetic sequence) should not be patentable, but methods to get it should - and not the idea of coming up with a way to measure the information, but the actual method itself.) Granted, that may actually take some effort to separate those two things, but that's the crux of my disagreement with the patent system.
"There are a dozen opinions on a matter until you know the truth. Then there is only one." - CS Lewis (paraprhase)
See my comment below for more details, but I wouldn't care if they patented the specific implementation. It appears that they patented something far more general than that (the idea of floating point framebuffers - not "this one way to make a floating point framebuffer").
Incidentally, part of "obvious" in my book is "It's so obvious, why bother trying to patent it?"
"There are a dozen opinions on a matter until you know the truth. Then there is only one." - CS Lewis (paraprhase)
I think we can now see the first salvo of the Patent Wars we have all feared were coming. It seems every dying company decides that they need to 'monitize their patent portfolio.' as soon as the customers disappear. And SGI will be horrible in their death throes. Thankfully most of Xerox was bought instead of us all having to suffer through their death spasms because they had even more patents to abuse when they were dying, although by now many of the most dangerous ones are probably expired.
But this is still unfocused thrashing. Wait until they, like SCO, sucumb to the temptations of the monopolist in Redmond to focus their attack.
The patent system doesn't need reform, it needs to be scraped and rethought. I'd say cap em at 1000 per year. With a number that low only real inventions would make it through and the number in any particular industry would be small enough anyone in that industry could be expected to be aware of them.
Democrat delenda est
"Display system having floating point rasterization and floating point framebuffering", which was filed in 1998 and granted in 2003, in its Radeon graphics cards.
This is insane. Look at this snippet from the OpenGL extension ARB_texture_float:
SGI owns US Patent #6,650,327, issued November 18, 2003. SGI believes this patent contains necessary IP for graphics systems implementing floating point (FP) rasterization and FP framebuffer capabilities. SGI will not grant the ARB royalty-free use of this IP for use in OpenGL, but will discuss licensing on RAND terms, on an individual basis with companies wishing to use this IP in the context of conformant OpenGL implementations. SGI does not plan to make any special exemption for open source implementations. Contact Doug Crisman at SGI Legal for the complete IP disclosure.
Did the ARB cough up money to SGI to be able to use floating point textures? What about DirectX's use of floating point textures?
I know for a fact that Nvidia supports "full 128-bit (4 component) floating point precision throughout the entire pipeline". Why isn't nvidia getting slapped with a lawsuit as well?
Also, a lot of graphics techniques were invented almost 30 years ago (such as bump mapping). Surely there has to be SOME prior art out there..... no one needed floating point textures/framebuffers until SGI invented/patented it?
Why doesn't SGI have a patent on integer-based textures/framebuffers? Probably because its rediculous to patent the use of a specific numerical type. Its like saying SGI invented floating point numbers. Rediculous.
How can we get around this stupid patent? Someone needs to invent a double-precision version of the floating point texture, and tell SGI to get lost.
That's a non-argument. In fact, in Europe novelty and non-obviousness are even two completely distinct patentability requirements. This argument also misses the point that the fact that nobody already published something about doing that (since this is the actual criterium for novelty/non-obviousness) does not mean it was not obvious.
In fact, it may have been so obvious that nobody actually saw the point of publishing something about it (like what's the case with many software patents), which paradoxically makes it quite hard to invalidate the patent based on absence of novelty/non-obviousness.
Donate free food here
No, the summary description doesn't define the scope of the patent, it describes what the thing patented is. They patented an implementation of a hardware graphic adapter, and are claiming ATI infringed by using that particular implementation, or at least elements of it that are substantial enough to infringe on what is protected by the patent.
I can't see what the correspondence during the patent prosecution was, nor what prior art was actually discussed, since the application from which this patent issued was too old have an online public Image File Wrapper, but under current patent law (but under Supreme Court review) to show Obviousness one needs to show positive evidence that "suggests" each aspect of the patent claims. Simple hand waving won't cut it; i. e. saying "It's generally obvious that doing anything in sw can also be done in hw (and of course vice versa).
So, can anyone find a good prior reference that explicitly mentions the desirability of using a floating point based frame buffer in a graphics environment? I know nothing, specifically about this art, but this would be needed to establish a good case. This would have a good shot at invalidating claim 22, the broadest claim to this feature.
Also I noticed that a number of claims, notably the first few independent claims seem to define mention floating point only in the context of individual sub-units of the rasterizing hardware, which, to my perhaps naive mind, do not require that the frame buffer actually store floating point color data, and at least one of the prior art patents cited (Intel's 5,995,122) uses fp in parts of the rasterizing hw. I know I would have used such a reference against to make a rejection, if, for no other reason, to get the applicant on record as saying these claims implicitly requires fp color data in the frame buffer.
Yeah, but if SGI patented using floating-point numbers in the framebuffer, and you have something like
// loads the frame buffer with floating point data
int loadFrameBuffer(float* data);
in a header file lying around somewhere, then it doesn't matter what your underlying implementation is. You're violating the patent.
:(){
I forget which but either ATI and Nvidia (possibly both) have said the reason why they haven't open sourced their drivers is because they contain valuable intellectual property that could be to the advantage of a competitor (i.e. each other - I assume)
Should ATI and nVidia figure out each other's tricks, wouldn't everyone benefit from the technology developed? (or at least the Linux crowd...) On top of this, the manufacturers would be forced to continue innovating to maintain their presence in the market.
:(){
I know for a fact that Nvidia supports "full 128-bit (4 component) floating point precision throughout the entire pipeline". Why isn't nvidia getting slapped with a lawsuit as well?
nVidia, last I knew, had a fairly broad cross-licensing agreement with SGI, mainly as a result of settling the suit between them about 6 years ago.
Why doesn't SGI have a patent on integer-based textures/framebuffers? Probably because its rediculous to patent the use of a specific numerical type. Its like saying SGI invented floating point numbers. Rediculous.
Because they didn't invent "integer-based textures/framebufers". They did, however, come up with the first working implementation of a floating point framebuffer and patented it.
So your point is?
In my view, its not 'Property' if you unknowingly infringe on patents that companies have "invested a lot of research and development into"
Its far more like 'research things that are as likely as possible to be thought of eventually by someone else then sue them into the ground for "damages" and royalties'
Which in effect means, as the vast majority of us know, the patent system is ripe for abuse and has been abused since its creation. [/rant]
renames new comany Stigmadia
OSGGFG - Open Source Gamers Guide to Free Games
much of their business is coming from government, defense and education. SGI is still selling MIPS based systems until the end of 2006, and will continue to manufacture MIPS/IRIX systems for several more years(most likely goverment amd defense contract obligations). most of the IRIX enginners have already been shifted to migrating the core technology from IRIX to linux. i don't believe SGI is turning into a SCO-like litigation machine, even if some PR monkey makes a wide sweeping statement about the vast amount of IP SGI is holding.
and while i'm on the topic of SGI and SCO, wasn't SGI initially accused by SCO of having violated licensing agreements in the whole SCO V. World debacle?
SGI has admitted defeat in the CPU market by porting EVERYTHING from IRIX to linux. it's a very risky process, and they've made a pretty good go of it. apple is the only other company whose actually managed a platform change THREE times and survived(four if you include OS change), perhaps SGI is hoping for similar results. if any company in the tech arena has a strong following, SGI certainly does.
three can keep a secret, if two are dead - benjamin franklin
Go read the background of this patent. The patent itself admits that there was plenty of graphics prior art that used floating point values to do the calculations. All they are claiming a patent on is implementing this system in hardware! And they include a line in their patent about how it has now become possible/desierable to implement the floating point stuff in hardware.
To be fair it seems their 'advancement' is that they kept the data in floating point format in the framebuffer rather than converting it to fixed point. Now their implementation of this approach probably contained some genuine advancement but just the notion of doing everything in floating point is pretty fucking obvious.
What I don't understand about this is that surely SGI has some much more substantial patents in their pockets. Is the problem that they gave them away with OpenGL/don't want to scare people away from opengl?
My personal guess is that SGI isn't serious about pursuing this particular patent infringement. Rather they are using a very broad and simple patent to go on a fishing expedition about ATI's hardware. During discovery on this patent SGI will be able to get details about how ATI's hardware/software works that they would otherwise not be able to get. I suspect SGI thinks ATI is infringing on a more substantial patent somewhere and is going to use the discovery during this case to find out.
If you liked this thought maybe you would find my blog nice too:
OpenGL has become a standard in graphics. You cannot reasonably do OpenGL without floating point operations in your rasterizer (Trust me, I've tried to find a way around it). OpenGL is an "open standard" that was pushed by SGI. So SGI wants others to use OpenGL. But, given the patents, they DON'T want others to use OpenGL, at least not without everyone paying them lots of money.
It just me, or does it seem to be that SGI are pulling a Rambus on us here?
Also, given the OpenGL spec, everyone doing 3D graphics (ATI, nVidia, Matrox, 3DLabs, S3, etc., etc., etc.) is violating this patent. No?
Years after SGI's failure to commercialize their invention they try to coin in using their patent portfolio. The main reason why I think this is wrong is that SGI isn't competing in this space (anymore) and thus is not loosing any money.
Personally, I'm not convinced that SGI invented the floating point display buffer. But they did patent it before anyone else did. Even if they did invent (as in "first to be there") this technology, they never managed to make a commercial product out of it. They tried and failed due to their own incompetence, slipping schedules and eventually dropping that product all together. This was years ago.
Now they twist patent law to make a dishonest living. Yet another example of why the US Patent system is flawed.
File for chapter 11..
...
re-organize...
sue the pants off of everyone...
PROFIT!!
The research paper InfiniteReality: a real-time graphics system provides some clues as to why.
It's more than just saying:
float framebuffer[1024][1280][4];
The engineers had to resolve the problems of increased video memory bandwidth requirements (512 bits or 64 bytes per pixel) at a 1280x1024 resolution at a consistent 60Hz, and use parallel processing techniques to eliminate major bottlenecks (4 geometry processors built from custom ASIC's, 80 image engines, and texture mapping hardware to support virtual 2048x2048 textures using clipmaps).
Vintage computer adverts: http://www.vintageadbrowser.com/computers-and-software-ads
In fact, it may have been so obvious that nobody actually saw the point of publishing something about it
I vaguely recall a talk at a conference several years back where an NVidia guy was talking about the need for floating point framebuffers in hardware. IIRC, he was specifically referring to their use in improving the accuracy of Z-buffers, and it sounded like they were in development at NVidia. I think it was the GDC road tour conference back around 1998. There must be some published material on the subject from around that time.
There are 0x40000000 types of people: those who understand 32-bit IEEE 754 floating point, and those who don't.
Not sure if anyone has noted this before, but in retrieving any entries from the PTO online database...
Every page has the SUN logo on it! Innocent accident you say? I think not!!!
Perhaps until the PTO can check for a single corporation's logo branding every single patent's online posting, we shouldn't hold them to such high standards for checking the actual patents they are issuing.
I just have the Dylan Thomas line ringing around in my head upon hearing this. "Rage rage against the dying of the light" Would a patent infringement suit have been adversely impacted had it occurred prior to or during bankruptcy proceedings?
ATI and nVidia are chock full of ex-SGI employees. For example, a good chunk of my friends from the MIPS division are at ATI. There's also the story of how they got rid of the desktop graphics division. The story goes that the entire team was pulled into the cafe. As they walked in their badges were taken. They were then told that some would be going to nVidia and some would be going home. So there is probably a whole bunch of SGI guys at nVidia as well. I wouldn't be surprised if some SGI-patented ideas leaked in....
"Where quality is like a dead stinking rat - you just can't miss it."
SGI is done as a technology company they won't be contributing anything to FOSS anymore. At least it's not like they can take the things they did give us back.
Jaysyn
There is a war going on for your mind.
Heh... They both sit on the ARB board, come up with new OpenGL extensions together. They can figure out the other's tricks easily enough. It's not REALLY that they're not being nice. It's because of some bad advice from IP attorneys that think EVERYTHING should be kept secret, lest they lose the right of patentability, copyright, etc. Believe me when I tell you this; it's a story that keeps playing itself out at each and every place I've been contracting at for the last 3-4 years now.
I am not merely a "consumer" or a "taxpayer". I am a Citizen of the State of Texas
They don't have to have source code, all they need do is provide the register level info- what each register in the PCI/PCIE space does and what you're supposed to set them with. That's all anyone had when they did the RagePRO, Rage128, i810, or G400 chipsets in the days of the Utah-GLX or the early days of the DRI project. I know, I was there DOING it. Doesn't matter much, though... Saying Intel's part is less complex isn't quite the case, and me bitching about how silly it all is to not release register info doesn't change the situation any. If AMD sees fit, they'll make ATI open up the info a little more. If they don't, well, just hope that their OpenGL driver group catches up with NVidia soon... :-D
I am not merely a "consumer" or a "taxpayer". I am a Citizen of the State of Texas
Yes, but at a faster rate. Currently the requirement for R&D budgets is driven by hard core gamers demanding faster frame rates, and new games coming out demanding faster hardware. The system has reached an equilibrium and everyone is happy. CxOs don't mind spending money, what they have a problem with is change. If ATI and nVidia together open sourced their drivers (and while they wouldn't be on the stage sucking each others cocks while they did it, if one did so within a month the other would almost have to) that would mean the CxOs would have to deal with change. Initially, to ship the drivers: more tech writers, more (real) technical support; in response to this from the other guy: more people in research figuring out what the other guy is doing; and with the research in hand: more hardware developers now producing new systems more frequently; with shorter product life cycles: screw up your whole production/marketing systems.
This new-video-card-world-order may be obtainable and viable. But would it, at the end of the day, mean more profits? There is only so much money around looking for video cards, shipping newer things faster wont necessarily suck any more money from that stone. So, from a business perspective; its a good ride now; changing the system opens up a lot of work, risk, with questionable outcome.
Quit making up nonsense excuses to forgive corporations who shit on their customers. The specs required to write video drivers do not tell anyone anything that makes it easier to sue for patent infringment. The list of features on the nvidia and ati websites gives as much info for patent trolls as programming docs would. It just tells you what functionality is supported, and how to tell the card to do it.
I loved the chips and the techies, but the legalists at SGI/Mips were startingly evil.
An unnamed former employer found their contract contained a clause that said the purchaser indemnified the vendor in the event of a patent suit brought in Canadian court...
--dave
davecb@spamcop.net
suing for patents may not work if you have no other products that sell, but SGI DOES sell products (not many graphics related ones) in the supercomputing realm.
Everybody likes to to try and point out you can't make a biz out of licensing patents, go look up how much money IBM made off of licensing their rather large patent portfolio.
The abstract, which you quoted, is not the patent. Read each of the "claims" in the patent for what is actually claimed as being covered by the patent.
That said, I don't know what the state of the art was in 1998, but how obvious was a floating-point frame buffer back then? To me, that seems to be the real question here, at least in terms of obviousness of the patent. When did ATI start using similar technology? How far behind does ATI have to be before the idea is no longer obvious?
http://www.forbes.com/technology/feeds/ap/2006/10/ 25/ap3119965.html/
That's not a good analogy. You're not trying to charge for use of the course, you're trying to charge for use of the techniques you used to reach the end. Others are free to hop, skip and jump where you sprinted, or carry the egg in their hand where you used a spoon.
It's official. Most of you are morons.
I would check into the original REYES hardware, which ultimately became Pixar's RenderMan. RenderMan has always assumed colors were meant to be floats, and the original implimentation was supposed to be in hardware...
I don't know for sure that it would have used an actual FP frame buffer, but it would probably be a good start for pointing out that FP for the frame buffer is bloody obvious. I still need to read through the patent to see if there is anything specific and novel that might be more impressive.
I personally welcome our AMD/ATI/SGI (AAS?) overlords.
Having to work for a living is the root of all evil.
The Bell Rings and SGI comes out of Bankruptcy swinging legal meat hooks. I guess the lean mean SGI machine is looking at all avenues to grow and claim any money that they can swing at.
They don't deserve anything except the opportunity of having a first-mover advantage. Patents are granted "To promote the progress of science and useful arts". Consider all the advances that were made in software before software became patentable. It's been clearly shown that patents are not needed, and all they do is gum up the works. Do you think the progress of "useful arts" in computer graphics would have been better off if all the early research that took place was under a patent umbrella?
And while a video card may be hardware, it's really just a realization of a software algorithm.
I'm curious what the actual hardware was....anyone actually know?
Max.
it does look like that. in 1998 it was only barely possible to implement that on Irix machines.. by 2003 when the patent was actually approved Nvidia and Ati were doing "floating point" but definately not the same type that SGI was doing.. their's was drastically different... much like comparing a chevy 350 to a mazda rx7... they're way different, I'd guess SGI ammended it's claims to include the "cheap" stuff they missed out on and fired all the engineers for.
They did, to nVidia, about 3 years ago.
Well, I did
6 85077
http://slashdot.org/comments.pl?sid=176905&cid=14
Good post! That's what I was thinking, but I couldn't think of a concise way to put it :-)