Open ATI R600/700 3D Graphics For Christmas?

[chelobaka]

It must be noted that there are also hidden threats for AMD opening specs. If enthusiast programmers will find these specs useful for brining better open source drivers to the community, then rival companies may find it as a good chance to compromise DRM path and bury AMD's business by leaking some decryption program into the Internet. Even software companies producing source-closed operating systems may be interested in this. No good hardware with open drivers - no good truly open Linux distributions with solid 3D support - no users switching to Linux, especially then AMD's closed drivers for Linux really suck for WINE gaming imo.

I have two RV670 cards I bought only for AMD's open source politics and like most of you I don't want this company dead. We just have to wait.

[duby229]

I am certainly --NOT-- endorsing downloading illegal media. And I am not suggesting that AMD make doing such possible. That just makes the case for DRM stronger.

What I am saying is that the content industry is actually limiting there potential customer base significantly by implementing DRM the way they have.. The goal for content should be exposure right? You would think that should be the purpose right? DRM makes that difficult....

And they wonder why there market is shrinking....

If I could watch a movie on linux in some uber high definition format unrestricted on linux, I'd gladly pay for it. Or maybe on my smart phone. Or maybe my PS3. Or maybe my Netbook. etc, etc, etc..... The list goes on and on and on. It is the content industries duty to make distribution more effective. Effective in this case means efficiency. You want to be able to make a profit over those copies that are being watched.... DRM makes that dream impossible....

Theyt limit there contents exposure to only a tiny range of formats, players, and hardware... And then wonder why there market is shrinking....

[curaga]

What "tricked" us with the 6xx was that the internal docs contained information on a lot of functionality we are not using in any drivers today, so the amount of information we had to cut out was much larger than in previous generations so it took a few attempts to find the right balance.

Bridgman, back on page two you said this interesting bit. I don't however understand, why would you not let open drivers use functionality the closed ones don't use (yet). What is the worst case of revealing that your card X has some extra functionality and it can be used this way? At worst it could be that the competing companies also add that extra functionality, and everyone benefits. Especially if this extra was something small, just something that hasn't conventionally been in a graphics card (hardware accelerated jpeg? hardware MD5?)
Of course if that unused functionality is something huge or secret, then I understand, but it would appear not all of it is like that. Is there any risk at all in revealing some small functionality?
That someone will prefer the open driver because of it?

[cjr2k3]

Ye that also puzzled me too.

What is the problem of releasing specs about hardware that is not used? Could it be used later?

[conspiracy mode = on]
I bet that is the part that makes an 4830 = 4850 = 4870... So if that
is opened everyone would buy a 4830 and end up with a 4870...
[conspiracy mode = off]

BTW, in a graphic family all the chips are the same right? I know CPU's are all equal and they get their speeds/cache ajusted after production in
a quality testing (If can perform full speed is full speed, otherwise
they sell it with a lower frequency). Is the same thing with GPUs?

[agd5f]

Bridgman, back on page two you said this interesting bit. I don't however understand, why would you not let open drivers use functionality the closed ones don't use (yet). What is the worst case of revealing that your card X has some extra functionality and it can be used this way? At worst it could be that the competing companies also add that extra functionality, and everyone benefits.

We aren't trying to deliberately hold back useful information. It's more an issue of whether certain features work or are supportable. For example when planning a new asic, new feature X may have been designed in and shows up in the specs. Somewhere along the way, someone decides that feature wasn't worthwhile or should be moved to a newer asic or maybe ended up having a hw bug. As such, feature X may not exist in hw, or it may have not been tested during hw verification (in which case it probably doesn't work), etc. Moreoever, if we aren't using that feature in any of our drivers, there's a good chance it will stop being verified or get dropped altogether on future asics. Also, if we don't use a certain feature in our drivers, it's often hard to find information on how that feature should work if it indeed does. So, the best way to figure out a common stable supportable feature set is usually to look at what our other drivers use.

[bridgman]

Exactly. The point I was trying to make is that the original design documentation we use as input included a lot of things which did not make it into the final chips, and also included some things which we decided not to use and therefore do not test in production ASICs. Normally we can use the original design docs as a good starting point for writing programming docs and drivers, but in the case of the 6xx/7xx family that didn't work so well and we ended up having to spend much more time working with the designers and driver devs.

curaga; almost all of these features fall into the "big and secret" category

[duby229]

Ye that also puzzled me too.

What is the problem of releasing specs about hardware that is not used? Could it be used later?

[conspiracy mode = on]
I bet that is the part that makes an 4830 = 4850 = 4870... So if that
is opened everyone would buy a 4830 and end up with a 4870...
[conspiracy mode = off]

BTW, in a graphic family all the chips are the same right? I know CPU's are all equal and they get their speeds/cache ajusted after production in
a quality testing (If can perform full speed is full speed, otherwise
they sell it with a lower frequency). Is the same thing with GPUs?

What your describing is called binning. The reason why companies bin there products is becouse not every chip that comes off of the same wafer is equal. It has something to do with a concept called defect density. Some dies on a wafer can clock higher then others. Some dies might have a section of cache that doesnt work. Some might of a rendering pipeline that doesnt work. Some might have some other flaw that prevents it from being fully functional.

So what they do is bin them by defects. Flawless chips are the rarest, and so they get sold into the high priced high end where the volume of products will be the lowest. down the line where the chips are binned into progressively cheaper products.

Keep in mind also that ATi generally has 2 chips in production, a high end chip, and a low end chip. And both of these chips have there own set of bins. Traditionally they were differentiated by Rxxx and RVxxx, however in recent generations that line has been blurred.

[rbmorse]

Don't overlook the possibility there are features on the silicon that work perfectly well, but are only enabled for the cards sold into the proprietary workstation market.

That's a huge revenue generator for ATI (somebody has to pay the bills) and they are not going to compromise it by giving away "features" they don't have to.

[drag]

I am certainly --NOT-- endorsing downloading illegal media. And I am not suggesting that AMD make doing such possible. That just makes the case for DRM stronger.

Well it doesn't really matter how strong the case is for DRM. The sort of DRM that uses technological means to control what people do with their own possessions is just impractical from the get-go. Remember that all this sort of thing is based on math, which is a way to express the fundamental rules of the universe.. all computers are is on and off switches and they are limited by physical reality and by economic realities.

There is certainly a strong case for having anti-gravity devices, but spending millions of dollars and passing laws saying that people can't expose how anti-gravity works isn't going to make it any more practical.


-------------------

AMD is limited by contractional agreements and thats all there really is to it. They have patent licenses and contracts, agreements, and understandings with their business partners.

At this point it's pretty likely that AMD could not release documentation even if they dramitically wanted to. They've probably agreed to pay fines if they did not fufill their obligations and I would hate to see AMD dragged to court by Microsoft or S3 or anybody else for trying to help Linux folks out with open source drivers.

And in addition don't forget the DMCA. If they release information that will teach people how to circumvent digital copyright protections they will be in violation of federal law.

-------------------------

Just remember how convoluted and difficult this documentation out is. That's the sort of thing that happens when you deal significantly in proprietary software technology.

It ends up limiting the people working on the proprietary stuff as much as it limits end users, if not more. This sort of stuff is why open source is gaining traction.. it's not so much that it's superior technically, but that because of it's openness it's just so much less hassle to deal with and improve.

Breaking free is the hard part, but I hope that AMD gets enough real benefit (read: profits) that they can justify the costs of openning up the hardware and making their newer hardware more and more Linux friendly.

[duby229]

AMD is limited by contractional agreements and thats all there really is to it. They have patent licenses and contracts, agreements, and understandings with their business partners.

At this point it's pretty likely that AMD could not release documentation even if they dramitically wanted to. They've probably agreed to pay fines if they did not fufill their obligations and I would hate to see AMD dragged to court by Microsoft or S3 or anybody else for trying to help Linux folks out with open source drivers.

And in addition don't forget the DMCA. If they release information that will teach people how to circumvent digital copyright protections they will be in violation of federal law.

Well, I guess I'll just have to disagree with you. Like I said, I think AMD could win that lawsuit easily. If MS took AMD to court they would have to claim that becouse of the restrictions placed they have a superior product. I dont think that claim will stand up in court. If this issue ever went to court AMD would have an aweful lot of case law on there side.

I dont think MS could win.

And for the DMCA, it's already been proven to be unconstitutional. All it needs now is for a large group of people to bring a class action against it. There is already several huge petitions on the web, and I'd wager that the majority of those petitioners would gladly participate in a class action suit against the DMCA.

Like I siad it all comes down to whether or not we are willing to stand up and fight against our oppressors.