AMD's opensource lies exposed

[deanjo]

Pointing to a lack of support doesn't mean a person is making fun of it. A fact is a fact and if we were to put our heads under the soil like ostrichs do then why bother with computers and technology? we would shut ourselves to monasteries with pentium MMXes and linux installed with open source supported radeon 9800s.

Yup, well said.

[glxextxexlg]

Some more FUD.

They dropped support for 5-year old chipsets.

RS690 introduced March 2007, driver support dropped March 2009. Last time I checked that was only two years.

R580 X1950XT 512 mb introduced Q1 2007, driver support RIP in March 2009. Any more hollow claims?

[deanjo]

opensource "nvidia" drivers deliver 0 results.
radeon is delivering maybe 30% of nvidia blob results.
I still do not understand amd though.

Ummm when it comes to opencl all open source drivers are presently delivering 0 results.

[deanjo]

That's the best argument I've heard yet.

If you use open source drivers, you kill children.

Hey just countering Q's claim on his own level.

[pingufunkybeat]

RS690 introduced March 2007, driver support dropped March 2009. Last time I checked that was only two years.

That's a glaring exception affecting a small number of people. The vast majority of products were older than 5 years at the time when the binary drivers stopped following bleeding edge X versions and kernels (they still work just fine with older software).

Why when there is a perfectly good blob pre-made and ready to use that offers all the features needed.

Becasue he cares about open source?

Why code KDE when there's a perfectly good Mac OSX? Why code for vim when there's a perfectly good Visual Studio?

There are many very good reasons for open drivers -- easier to maintain for distributions, it is LEGAL (!!!) to distribute them, they do not violate the GPL (!!!), they work with multi-screen setups properly, support kernel modesetting, work out of the box, etc.

These are good reasons. If you're happy with your semi-legal blob which the kernel hackers hate, the distros hate, which is basically illegal according to Linus and many others, which replaces half of your system with unknown code, then that's fine, but please to respect our wish to have a clean, open solution.

If our compiler is open, our kernel is open, our editors, desktops, browsers, servers and pretty much all other drivers are open, then why should we depend on the benevolence of one single company and their magical, closed, illegal blob?

The blob makes the developers' lives difficult. All because nvidia has secret sauce. Screw them. I can live with 50% or 75% performance. I don't want the Windows graphics subsystem injected into my kernel

[pingufunkybeat]

- it will never reach full possible speed due to absence of close-to-hardware tools and direct communication with designers.

there is no such thing as "full possible speed", that's a poor argument which keeps popping up.

The blobs are faster, and taken as a reference, but that doesn't mean that this is "full speed", just that they have been optimised better. They could possibly be even faster with some more optimisations, some of which are unreasonable.

Same goes for GCC and other compilers. Using the binary blob on Linux is like compiling your whole Linux using Visual Studio because it produces more optimised code.

80% of the blob is plenty fast. It's just that gamers get really touchy about their FPS. An Intel i7 running GCC-compiled software is much slower than when it runs ICC-compiled stuff, yet people don't bitch and whine about not getting full performance from their expensive CPU, because it doesn't impact the FPS in Vorkraft :P

[deanjo]

That's a glaring exception affecting a small number of people. The vast majority of products were older than 5 years at the time when the binary drivers stopped following bleeding edge X versions and kernels (they still work just fine with older software).

Small subset? It included every IGP solution AMD had until the 700 series.

Becasue he cares about open source?

Umm you were responding to me. You told me to start coding or stfu. I responded as such.

Why code KDE when there's a perfectly good Mac OSX? Why code for vim when there's a perfectly good Visual Studio?

I don't code KDE, I absolutely don't use vim let alone code for it either.

There are many very good reasons for open drivers -- easier to maintain for distributions, it is LEGAL (!!!) to distribute them, they do not violate the GPL (!!!), they work with multi-screen setups properly, support kernel modesetting, work out of the box, etc.

They don't do what I need them to do so all of that is useless to me. I'd rather take the 30 seconds to install a blob then take years to get full support for the hardware. If the open source drivers had that capability especially at time of purchase then I would have zero issues using open source. Chances are though they will never match the features and performance of the blobs. KMS is useless to me as it serves little practical use.

These are good reasons. If you're happy with your semi-legal blob which the kernel hackers hate, the distros hate, which is basically illegal according to Linus and many others, which replaces half of your system with unknown code, then that's fine, but please to respect our wish to have a clean, open solution.

I have no problem with other people using them but on the same hand don't condemn others who prefer functionality, performance, and getting the best value of their hardware because they don't care if it is open or not, just the end result.

If our compiler is open, our kernel is open, our editors, desktops, browsers, servers and pretty much all other drivers are open, then why should we depend on the benevolence of one single company and their magical, closed, illegal blob?

Well first of all, it isn't illegal. I may be the opinion of some that it should be illegal reality is that it is not. Again best bang for the buck trumps here and FOSS drivers are far from the "best bang".

The blob makes the developers' lives difficult. All because nvidia has secret sauce. Screw them. I can live with 50% or 75% performance. I don't want the Windows graphics subsystem injected into my kernel

How does it make their life difficult? If anything it is easier for them, just look at recent testimonies from Mozilla, KDE, Carmack for example. If anything it is the FOSS drivers that are giving them grief.

[pingufunkybeat]

Here, for example

http://people.redhat.com/bkoz/benchmarks/

ICC produces exexutables up to 2x faster than GCC (though the difference is usually not that big). Yet this is what all distributions are compiled with. And much of the scientific simulation software.

Getting to 75% of the performance of the blob (like r300g) is not a problem, it's an excellent result. The gamerz can play Windows, which is what runs most of their games.

The main issue with the open drivers is functionality (for recent hardware). Getting to OpenGL 3 and higher is a very important step. Crying about FPS is unimportant. Stable, open source drivers are important. They provide legal, out-of-the-box functionality which works well with the rest of the system. This is important.

[deanjo]

Here, for example

http://people.redhat.com/bkoz/benchmarks/

ICC produces exexutables up to 2x faster than GCC (though the difference is usually not that big). Yet this is what all distributions are compiled with. And much of the scientific simulation software.

Getting to 75% of the performance of the blob (like r300g) is not a problem, it's an excellent result. The gamerz can play Windows, which is what runs most of their games.

The main issue with the open drivers is functionality (for recent hardware). Getting to OpenGL 3 and higher is a very important step. Crying about FPS is unimportant. Stable, open source drivers are important. They provide legal, out-of-the-box functionality which works well with the rest of the system. This is important.

Yes and where performance counts for people that want it people do recompile. If this wasn't the case there would be no need for alternative compilers for linux now would there?

[pingufunkybeat]

Chances are though they will never match the features and performance of the blobs. KMS is useless to me as it serves little practical use.

They did with r300g, for the most part. And KMS is useful for stability. Nvidia used to lock up on me when switching VTs. Never happened with free drivers and KMS, because it's properly handled.

Well first of all, it isn't illegal. I may be the opinion of some that it should be illegal reality is that it is not.

It is the opinion of Linux Torvalds and a number of high-profile kernel developers. The fact that they aren't pursuing it, but ignoring the existence of Nvidia and their blobs doesn't change this.

How does it make their life difficult? If anything it is easier for them, just look at recent testimonies from Mozilla, KDE, Carmack for example. If anything it is the FOSS drivers that are giving them grief.

Kernel hackers, for example, refuse to deal with bugs involving blobs. X people and kernel people refuse to introduce blob-only functionality into their software. Many distros refuse to ship them because of unclear legal situation. Package maintainers have problems because upgrading is difficult (for example, upgrading the Nvidia drivers automatically can b0rk your system when they drop support for your chipset, happened to me).

When there is a bug, it's impossible to trace. Lol at KDE. KDE was UNUSABLE on Nvidia for YEARS. Unusable! Because they didn't accelerate things which were important for desktop effects, but were not used in games. The only solution was waiting for Nvidia to fix it.

RedHat loves the blobs so much that they are funding nouveau developers to reverse engineer the blob. Novell loves them so much that they were working on RadeonHD.

What are you smoking? The blobs are foreign tissue which cause problems for developers, package maintainers, distribution, and users. Only gamers and some GPGPU guys prefer them because they absolutely need the performance.