OSNews

No, you haven't.

AMD is trying to tell a hurting man that that being poked by a branding iron is more painful than being poked by a cattle prod.

This may be an internal strategy paper and it does show that AMD is putting down a tight release schedule for drivers under Linux and that many common criticisms towards AMD seem to be unfounded. Beyond that the article has nothing to say but that AMD's effort is so much better than NVidia's. For free and open source advocates there is little in there. However, news of open sourcing the code are reserved for another article. Or so TFA says...

Exactly. What they say doesn't help that I had a X1600XT and I had to replace it with a GeForce because the ATI wouldn't work with most Linux distributions.

Read the summary a bit more carefully:

"while this issue has come up again more recently, we will cover this particular topic in a different article"

I thought the point of TFA was that they are trying to make the drivers less horrible, even if not open. That was what I got out of it at least.

trying, yes, but failing miserably.

as i always say to things like this, rome was not built in a day...

True enough. But if Rome had been constructed as ineptly as ATI's Linux drivers, then people would probably be praising Nero today.

Yes, because improving such terrible drivers is simple in...how long has it been? A month? A few?


If they say they're working on decent drivers and don't deliver, I'll be one of the first to yell about it...

But jeeze man, give them some time.

Rome wasn't built in a day?

Decent software is, in my experience, never written in a day. Especially something as complex as GPU drivers!

I really do appreciate that ATI/AMD are making an effort and that maybe they're seriously trying to turn around the deficiencies the driwers have suffered from in the past!

However, I do agree with you that it'd be really nice to see something come out of this, like AIGLX support etc. I've found the ATI drivers that came with my Ubuntu Feisty install to have some annoying bugs as well as the lack of up-to-date features. I'm hoping future releases will show the benefits of the improved attention the drivers are getting.

What good will opensourcing the drivers do. R200 has been around for a while and is opensource yet it doesn't support quite a few things that frankly it should have right now. Are you going to blame ATI for that too? If the opensource nvidia drivers are as good the developers say its going to be then we should be looking at the opensource radeon drivers developers and find out why they haven't gotten there crap together. Start from scratch do something.

Nvidia rarely releases drivers, but they work and they work well, and that's a testament to there dev team. Fglrx is way behind from lack of attention and now the devs have to work double time releasing monthly to get feedback from users as to what works and what doesn't. I think that if ATI can start to develop their drivers right on *nix they can probably whip there windows drivers into shape too. Those are almost just as bad as linux in some cases. I do like to see monthly releases though, its nice anticipating aiglx. When I see that in the release log I'll cry for joy.

Exactly, which is why people ask for OPEN SPECIFICATIONS.

There is a clear difference between having the actual source code for the operation of hardware, knowing everything about it and how to function with it and reverse engineering the code by looking what calls the hardware is making and making the code up from that. I believe that ATi hasn't given anything in the way of specification to the hardware because they have to "think of their shareholders". nVidia have worked with the development of open source drivers and have spent a lot more development time on the drivers, which has been said before.

I don't see how anyone can quibble with the fact that open source is not the best solution for the user since there is a greater chance that the code can be improved rather than in a small, propriety group. So open source code and specification would mean more features available to be built by the community. I believe that the nVidia drivers aren't perfect, but if they were open source then they would be going in the right direction to.

Actually, the r200 open source drivers do support pixel shaders. It was implemented on r200 after it was reverse engineered on r300.

And while XvMC and TV-Out may not work, saying the drivers suck is a little harsh. Pretty much all 3D features of the cards are supported.

Nouevau (for nVidia cards) is planned to support all that stuff, and even hardware H.264 decoding on supported cards when they figure out the best solution..

Planning and implementing are two entirely different things. Please let us know when those items are actually supported.

Adam

Actually, the r200 open source drivers do support pixel shaders. It was implemented on r200 after it was reverse engineered on r300.

They do? Atleast not on my card, even though the card does support pixel-shaders under Windows. Hmm :/

Like for example, TV-output is possible only when building from the sources and patching them

We can all thank Macrovision for that. How is Nouveau going to get around their patents?

We can all thank Macrovision for that. How is Nouveau going to get around their patents?

This is a bit off-topic, but I decided to answer anyway. The situation is that the R200 drivers are built around whatever information ATI provided the devs with, which obviously lacked the TV-Out support information, but the Nouveau drivers are completely built around reverse-engineered information. As such, they don't need to care about Macrovision. All they need is the information about which registers to write to, and what values, and they apparently already have that information.

Actually, both hyper-z and pixel shaders were implemented in the r200 driver through reverse engineering. It's not like the nouveau developers have cornered the market on that :-)

Adam

AMD shouldn't have bought ATI and instead should have invested in their CPU programme and increased production facilities.

Maybe in yesterday's hardware industry, but not tomorrow's. In under 5 years, the vast majority of client processors will have on-die graphics units. Midrange MPUs will feature something like 4-8 CPU cores and 32+ GPU cores. AMD didn't have the skills, resources, or patent portfolio to get from here to there. Intel does, and so they will develop their own graphics technologies in-house. NVIDIA will be reserved for high-end gaming and visualization markets.

If AMD stays committed to their open bus technologies and successfully integrates ATi's DNA, they will remain relevant in the client market. Otherwise, they'll only compete with Intel in the low-end of the server market.

provide a quote.

Intel had until recently 60% market share but dropping and Nvidia 20%. There are about 1 billion computers worldwide. This is ignoring smaller makes like sis, but these add up to about 800million computers worldwide.

Latest Generation console sales add up to about 20million, and are dwarfed by those of its own PC based GPU's sales about 200million. Thats a fortieth of that of GPU's by Intel and NVidia.

Seriously what are you on about.

Provide a quote.

Ok ...

Moulinneuf Quote :

1) Anyone who know graphic/video cards know that Nvidia make *zero* graphic cards. They make the chipset and the architecture , company like ASUS , MSI , BFG , etc ... build the video/graphic cards. That was partially why Nvidia had an advantage over ATI int the past ( more vendor and OEM ).

Intel is almost the same , except they build other hardware with graphic/video card included , but no graphic card.

So yes the quote "ATI sells more video cards" is 100% accurate. But a bit skewed.

"Intel had until recently 60% market share but dropping and Nvidia 20%."

Where do you get your numbers ? What Market ?

"There are about 1 billion computers worldwide. "

No , computers are in everything from Bus to elevator to refrigerator this days ...

"This is ignoring smaller makes like sis,"

SIS is not a small maker , they just don't have competitive product's in the video game market , but with on-board graphic they are in a lot of devices.

"Latest Generation console sales"

here is a nice estimate :

Worldwide Hardware Sales (End of March 2007)
PS2 - 117.89 million
Xbox - 24.5 million (unofficially estimated)
GameCube - 21.59 million
Xbox 360 - 11 million
Wii - 5.84 million
PS3 - 5.5 million
Game Boy Advance - 79.46 million
Nintendo DS - 40.29 million
PSP - 25.39 million

Note : There is a short supply of WII people want them but there none available on the market. Not all *Console* are listed either.

"Thats a fortieth of that of GPU's by Intel and NVidia."

No , thats More sales that Nvidia and Intel lost to ATI on top of the current PC related products. One of ATI main problem in the past was that they could not meet the demand ( something a lot of people seem to be forgetting or are unaware of ).

People need to realize that Drivers , Graphic chipset and Fabrication and manufacture are not thing that innovate or grow on there own or in tree or by *waiting*.

You know something else people don't really respond well when you talk about **market share** , first Microsoft lies about there's , and most people they only think of market share : "The product they have."

Interesting numbers, though I am not sure their relevance. First of all they are quite open to debate. Here is an alternate set:

http://www.vgchartz.com/

1) The PS2 is not ATI as far as I know.
2) PS3 is Nvidia.
3) The embedded market is a growing one to be sure. Are GBA, DS, or the PSP using ATI? I did some googling on the specs of the 3 and could find no reference to an ATI GPU being used. If you tell me they are I will believe you, but if you could provide a link I would be grateful. This is more because I am interested in the technology than proof.
4) Those numbers are for lifetime sales on the selected consoles. Lets compare them against computer world wide sales for just two years (2005 and 2006):

http://macdailynews.com/index.php/weblog/comments/12349/

While that is just a Mac blog, the source for the numbers is IDC. This puts sales at 400M for 2 years, so 1 billion in use is quite possible.

Now stripping off non ATI consoles, ATI's numbers from consoles are not terribly significant.

None of this changes the fact that ATI has not shown a lot of interest in the Linux market. Linux users can choose not to buy ATI due to lack of good support, and many do choose not to. ATI as of this moment seems to have looked at the numbers and decided it was not cost effective for now. You don't buy ATI, and ATI accepts these lost sales as part of their cost/profit comparisons.

But at the same time, you seem to be forgetting three important things; firstly, if all alternative operating system people got together and refused to purchase ATI and AMD hardware, it would be a sizable chunk.

Secondly, as 'geeks' we're prepared to pay for the expensive gear; we aren't the 'charlie cheapskates' of the PC world, we'll go out and spend $400 on a graphics card, the type of card which have the good margins - a decent market with people prepared to spend the money.

Thirdly, we're the goto guys when it comes to IT advice for friends and family; all we have to do is say, "na, I'd avoid that computer, this one with the Nvidia graphics card would be better suited" and voila, no sale for ATI or AMD.

Its time for AMD to be punished, they've been given 6+ months, they've done nothing to improve the situation - boycott all AMD and ATI products till such time they actually fix the situation.

Sony Playstation 3 is using a customized Nvidia G70 videocard aka Geforce 7800 GTX.

Edited 2007-06-01 20:18

"No they shouldn't" and "It's their decision" contradict each other. Linux might be a small community but it is a vocal one. Intel is getting a lot of good press regarding their own relatively weak driver. Linux users are a large proportion of the technical community...those people who people ask advise and make purchasing choices for large companies and friends alike.

Even if it is their choice how they use their code. It does not distract that particularly for drivers, that there advantages from doing so. Stability; Compatibility; Security; Free Advertising etc etc.

Linux is not an OS that is updated every 5+ years like Microsoft. X is updated every 6 months and Linux every 3 months.

I actually agree with you cyclops!

Drivers in general should be opensource, ESPECIALLY on linux. It drastically cuts down on support costs as there are often people who are more then willing to do the supporting for you. Drivers (typically) are the textbook situation of where you are giving away free software to sell something else. It is nothing but an expense to hardware companies, and it is an expense that can be virtually eliminated (or at least mitigated), by going the open source route.

They never asked you to use Linux.

WTF? No, but they did ask me to buy their hardware...

WTF? No, but they did ask me to buy their hardware...

Sure they did, but do you always do as people tell you to do? It is your decision if you buy their hardware, and you decide what OS you use. If your OS isn't compatible, it's not their fault.

You don't seem to get it. You seem to assume I'm buying their hardware and complaining about it, but I'm not. What I'm doing is not buying their hardware and informing them what they can do to change my mind. If my OS isn't compatible, it is their fault, because they haven't supported it. It certainly isn't my fault they haven't done so.

@Zedicus:

I know of more people who have had issues with their ATI driver than nvidia (Both Windows and Linux) but especially Linux. I myself have had a few ATI cards in the past, and they all sucked except for the 9800 Pro under Windows.

I have also had machines with nvidia cards and except for _one_ card they have all performed like champs under both Linux AND Windows.

As a programmer, all you have to do is take the input and manipulate it in order to get the desired output. How hard is that to do??

You're missing a WHOLE LOT of stuff in between. Like for example, handling AGP, PCI or DMA isn't that easy..There's always memory management issues to be solved, the driver has to utilize several other drivers in order to even use AGP, and there's always the fact that those cards are so full of features that it takes complex sequences of code to get them to do a certain thing. Even just a single error when writing to a register can crash the card.

But lets ignore the graphics; the simple fact is, Intel is focused on delivering a complete solution to vendors, wireless, chipset, processor, graphics, network etc. A turnkey solution to their customer (the OEM); the graphics chipset itself isn't being sold, but the whole solution.

AMD is doing the same thing; the difference is that they don't have a 'complete solution', they're merely got a piecemeal solution from different vendors. For all of AMD's so-called 'success' they're still far behind Intel when it comes to laptop related share.

Why does Intel 'opensource' (they didn't actualy opensource their whole driver btw IIRC) driver? because there are no 'strategic intellectual property' which Intel needs to bullcrap about as with the case of Nvidia and Ati; for Intel it is all about volume. Push through, and ensure that those fab machines are constantly pushing out more chips. What they lose from 'competitiveness' by not 'hiding behind closed source' they make up by volume, low pricing and lower cost. Intel don't need to ride on margins of just graphics cards - they make it one the whole system.

Community-driven in-fighting RARELY ( if ever ) really negatively affects the development process. It is the fuel by which it is driven.

ATI Could save untold sums of cash by using a controlled open-source paradigm for their driver development, 'merely' by releasing their specs and hiring a fews devs ( or using existing ones ) to oversee the development process.

The real stoppage in the sewer line isn't that there exist no business reasons to comply with the communities wishes, it is, rather, the desire to prevent their competition from using it ( even secretly ). Among the likely licensing issues.

I just want to know if they can or can't be, and if they can be... will they. I don't think that makes me 'ungrateful,' I'm certain they are not ungrateful for my business ( having purchased, for others and myself, nearly a hundred ATI cards, and probably nearly as many nVidia cards ).

--The loon

Ungrateful? I should be grateful exactly about what?

I don't have to be grateful for receiving drivers that work correctly for hardware I *paid* for, and much less for drivers that don't.

As far as BSD/Solaris, it comes across as a "boned either way" type of deal. With closed source drivers (ie, the current situation), the drivers only support the OS selections the vendors want you to run (ie Linux for ATI, Linux/FreeBSD/Solaris for nVidia as far as *nix goes). With an "open source" driver like Intel's, there's no specs, no docs without an NDA, and ultimately you're just left with undocumented, incomprehensible vendor-supplied source code that does you little better than an undocumented binary (Unless you're running Linux, in which case it'll be in the kernel tree and "just work.")

IIRC, Intel's driver code for the x86 architecture are completely open source. They do include some closed binary libraries that are not run by the x86, but shoved to the network processing chip, which has a different (and undocumented) architecture.

Why this is deemed good enough is because the binary bits are OS independent, and could have been hidden in a rom inside the network chip for all you care. When new versions of the Linux kernel come up, or when the open source driver is ported to BSD or OSX or whatever other OS, the same binary blob can be sent to the chip without caring about its contents, just as you don't have to care about what microcontroller is inside your USB flash bar.