airlied ([info]airlied) wrote,
@ 2005-11-23 21:04:00
Previous Entry  Add to memories!  Tell a Friend!  Next Entry
ATI YOU SELL CHIPS!!! NOT DRIVERS!!!
I noticed someone making a point on /. about the fact that no 3D card with DirectX9 capabilities as had any sort of specifications released for it to open source developers.

And thinking about it, it seems like a very valid point, ATI stopped even talking to 3D open source developers with the r300 chips.

Am I being paranoid or are MS involved, in a totally non-monopoly crushing all opposition fashion, the alignment of DirectX9 functionality and lack of specs is a bit iffy alright.

Also I've just discovered that I have information about the r300 cards under NDA, via a Windows register explorer utility supplied by ATI to OEMs, I don't have any documents just some text files the application loads which maps register names to register addresses. I'm not even sure they could be considered secret anymore as the r300 project has already RE most of them and a lot of them are available via the r200 header files.

It places me in a slightly precarious position in the future with respect to the r300 driver project, the r200 XvMC should be okay as I reverse engineered that before I ever even seen the register explorer utility.

I just find it a very unusual business decision for a company who makes their money selling chips to want to stop people from using those chips in situations that ATIs closed-source driver don't support, i.e. Linux on PPC, 3D graphics without 2D framework (mesa-solo). In case you missed it ATI *YOU SELL CHIPS NOT DRIVERS*.

I'm just reminded of the Simpsons when Kodos and Kang take over the two nominees for US president, "so who else you going to vote for? its a two party system! Perot?"
in this case ATI/Nvidia are the two parties with Intel the Perot, Intel do mostly the right thing with respect to their drivers (they are open), but they have no power :-)

(Post a new comment)


[info]fooishbar
2005-11-23 12:08 am UTC (link)
DON'T BLAME ME, I VOTED FOR KODOS!

(Reply to this)

intel no power??
(Anonymous)
2005-11-23 12:25 am UTC (link)
Intel actually sells more chips than ati or nvidia...I wouldn't say they have no power...

(Reply to this) (Thread)
Re: intel no power??
[info]ajaxxx
2005-11-23 12:30 am UTC (link)
when was the last time you bought an intel video card?

(Reply to this) (Parent)(Thread)
Re: intel no power??
(Anonymous)
2005-11-23 12:40 am UTC (link)
Intel integrates their graphic chips in their mainboards. And intel owns ~80% of the desktop market and ~90% of the server market (and lets not talk about the laptop market)...even if not all the intel cpus use a intel graphic chip, they easily outsell nvidia and ati. Of course you'll hear a lot more about nvidia and ati, those who buy extra graphics cards are the geeks who care about speaking on internet.

Check yourself the numbers for the 2005 Q3... http://www.theregister.co.uk/2005/11/01/graphics_chip_mkt_q3_05/

"The large number of integrated shipments naturally favoured Intel, which remains the dominant desktop graphics chip vendor - and, indeed, the leading graphics chip maker period. Across the desktop arena, Intel took 36.9 per cent of the market, ATI 23.6 per cent, while Nvidia was close behind with a 22 per cent market share."

(Reply to this) (Parent)(Thread)
Re: intel no power??
[info]fooishbar
2005-11-23 12:43 am UTC (link)
'And intel owns ~80% of the desktop market'
'Across the desktop arena, Intel took 36.9 per cent of the market'

Something I'm missing?

(Reply to this) (Parent)(Thread)
Re: intel no power??
(Anonymous)
2005-11-23 12:53 am UTC (link)
'And intel owns ~80% of the desktop market' -> CPUs
'Across the desktop arena, Intel took 36.9 per cent of the market' -> GPUs

Sorry about that :)

(They are top 1 chip graphic maker since 2003 - http://www.theregister.co.uk/2003/10/30/intel_outsells_nvidia_ati/ - personally, I _like_ the idea that the top 1 graphic chip maker somewhat supports open source drivers)

(Reply to this) (Parent)
Re: intel no power??
(Anonymous)
2005-11-23 12:33 am UTC (link)
(and I was meaning: graphic chips not cpus of course! :)



By they way, I remember a interview where Linus talked about this... and remembered from the days when the ide port was connected to the sound card and every chip was different until intel started integrating the ide controller in the mainboards, then everybody started doing the same and followed the intel "standard" and compared it with graphic cards and predicted that graphic chips like everything else would have to go through the same phase and it'd be "easy" to support graphic chips - it looked like a good point.

(Reply to this) (Parent)(Thread)
Re: intel no power??
[info]fooishbar
2005-11-23 12:36 am UTC (link)
Er, graphics cards are significantly more complex than a CD-ROM. Any idiot can build a CD-ROM controller, be the interface proprietary or standard IDE. But you'd need hojillions of dollars, and the brightest minds these dollars could buy, to get your cards anywhere near where ATI and NVIDIA are today.

Intel have good chips, don't get me wrong. They're solid, they do the job, and I have one in my laptop. But they don't play HL2.

(Reply to this) (Parent)(Thread)
Re: intel no power??
(Anonymous)
2005-11-23 12:47 am UTC (link)
"Er, graphics cards are significantly more complex than a CD-ROM. Any idiot can build a CD-ROM controller, be the interface proprietary or standard IDE. But you'd need hojillions of dollars, and the brightest minds these dollars could buy, to get your cards anywhere near where ATI and NVIDIA are today."


My point is, that at some poing 3d cards will be no longer something only "brilliant minds" can do. We're seeing a lot of noise around graphics right now - 3d desktops, programmable GPUs, no real 2d engine on modern cards etc - but at some point (after some years) it has to calm down and become a "commodity". It always happens.

(Reply to this) (Parent)(Thread)
Re: intel no power??
[info]fooishbar
2005-11-23 12:56 am UTC (link)
I think a better analogy to IDE controllers would be, say, CPUs. Which are certainly not commodity, and still a matter of hot competition between -- you guessed it -- two vendors, with a marginal third. Heck, one of them's even Intel!

(Reply to this) (Parent)(Thread)
Re: intel no power??
[info]fooishbar
2005-11-23 12:56 am UTC (link)
And the marginal third's Via! Whoa.

(Reply to this) (Parent)(Thread)
Re: intel no power??
[info]libv
2005-11-24 09:38 pm UTC (link)
What? Is VIA third?

Time for another blog entry!

(Reply to this) (Parent)
Re: intel no power??
(Anonymous)
2005-11-23 03:49 am UTC (link)
Err..

You're taking the example of IDE chipsets, where.. you need a hardware specific driver to get remotely fast (DMA) transfers?

-Robert

(Reply to this) (Parent)

[info]hub_
2005-11-23 02:00 am UTC (link)
that is really lame that graphic cards are that hard to deal with in open source... stupid vendors.

(Reply to this)

IP
(Anonymous)
2005-11-23 04:04 am UTC (link)
I think the classic answer to this is that the vendors are afraid that published specs will point out where their hardware steps on the toes of other company's patents. And that by publishing specs, the will be showing that they 'knowingly' intruded on the patents. When everything's obfuscated, there's a much greater leeway for legal denial.

I'm not sure how valid this is.

But it's not as though it's just the big biopoly being painful. Even the smaller companies (who you'd think need any extra edge in the market they can get) can play that game just as well. VIA talk the talk but their 'open source' drivers end up being shoddy mixed binary & source releases. If you read Luc Verhaegen's blog you'll know that what little documentation does get released is often inconsistent and sketchy. XGI seem to have taken classes from VIA in this field too.

So really, you reverse engineering wizards are our only hope. And - let me add - I for one hugely appreciate what you do.

(Reply to this) (Thread)
Re: IP
[info]smitty1e
2005-11-23 11:26 am UTC (link)
The other avenue of attack in this situation is the market. If there is any kind of F/OSS card, and people make it popular on that basis, then the vendors will have to listen to the thing that talks, money.
Also, I think that IDE drivers are in the same lousy boat.
Fooey on the capitalists!

(Reply to this) (Parent)(Thread)
Re: IP
(Anonymous)
2005-11-23 02:40 pm UTC (link)
I may just be a pessimist, but I don't see a (usefully fast) Free graphics card design happening.

(Reply to this) (Parent)(Thread)
Re: IP
[info]smitty1e
2005-12-11 02:05 pm UTC (link)
That's realistic. I hope that there will eventually be a free card that is at least useful up through business graphics.
I agree that the 1337 g4ming rig will probabaly never be free.

(Reply to this) (Parent)
Paranoia hat on
(Anonymous)
2005-11-23 03:28 pm UTC (link)
ATI got lucrative contract with MS on designing Xbox360 GPU+northbridge chip. What were Microsoft's requests? We will never know.

(Reply to this)

XvMC support
(Anonymous)
2005-11-24 11:43 pm UTC (link)
So how's the XvMC code coming along?

(Reply to this)


[info]kosai
2005-11-29 09:16 am UTC (link)
Thanks for being vocal about this on linux-kernel; I think many of us hadn't realised we're heading for a crisis with the new Avivo chips, and with ATI's behaviour on the whole.

Is there anything we should do to help? Is DRI short of developers to help reverse-engineer X1000s, or would we get in the way? Would it help to give you an account on a PCIe system, or even to send you a X1300? Would being able to track down registers/memory locations get us anywhere, or is the card so mindless that only a full clone of whatever ATI's Win32 accelerated driver is doing will achieve anything?

Thanks again, sorry for so many questions.

- C.

(Reply to this)

I don't know even about FPGA vendors...
[info]vlnv
2005-12-25 03:52 pm UTC (link)
They are in the business to keep YOU (the individual/small business v.s. big corporation) OUT. That is their main goal - they are the superhuman, and you're - underhuman. That's the goal of legal departments, it has not much to do with making money. They make money to feel superiour - not the other way round.

For example, how would one explain that every FPGA(CPLD) vendor (Xilinx, Altera, QuickLogic, Cypress etc.) wants to charge an order of $5000-10000 USD for the software (VHDL,RTL) that enables using said FPGA chips, whereas none of them would disclose specification on FPGA programming that would make free software possible for these chips.

I would never consider myself bound by any kind of such NDA in moral terms, and would try to move to a jurisdiction where these pains are minimized or make any kind of trick to defeat the letter and spirit of said NDAs.

If you think - RMStallman was always right. It's a moral, freedom issue, not legal, business or engineering.

(Reply to this)

Create an Account
Forgot your login or password?
Login w/ OpenID
English • Español • Deutsch • Русский…