NVIDIA Denies Opening Up Its Driver

[Kano]

@Redeeman

That's a value in your dreams. Because several things are completely missing like GLSL support and glxgears is really no benchmark - that it would be possible to archive a similar thruput for extra simple operations yes, but you must be really lucky to get maybe 25% speed for more demanding things.

[Redeeman]

You need to take a chill pill. Seriously. You are more of a hindrance than a help right at the moment with the attitude.

Pick your fights- something you're not doing right now. Honest.

i guess thats the difference between the two of us, i pick fights that are just and right, you seem to pick fights "you can win"

[Redeeman]

@Redeeman

That's a value in your dreams. Because several things are completely missing like GLSL support and glxgears is really no benchmark - that it would be possible to archive a similar thruput for extra simple operations yes, but you must be really lucky to get maybe 25% speed for more demanding things.

glsl should be coming some time, atleast with gallium.

which things are you saying have only 25% speed?

anyway, that stuff SHOULD become faster with the documentation released, right?

[ethana2]

Will we need GLSL when we have gallium?

[Redeeman]

yes, we will, glsl is what the applications use to do the shaders.. gallium will then read the glsl and transform it into something the hardware can do.

[Svartalf]

i guess thats the difference between the two of us, i pick fights that are just and right, you seem to pick fights "you can win"

You tilt at windmills- what you consider is "just and right" only distorts the message you try to convey and you get NOWHERE fast.

[edit]
You have your heart in the right place- but stridence does you NO favors. You seem to enjoy bringing strife wherever you go and you accomplish little with it.

[bridgman]

glsl should be coming some time, atleast with gallium.

yep; definitely needs a good memory manager first, and by the time we have the memory manager Gallium should be far enough along that it makes sense to do the rest of the work there. The open issue with Gallium is how effectively llvm will be able to pack work into the SP's, but that primarily affects performance not functionality.

EDIT -- it's probably worth mentioning that GLSL doesn't "come for free with Gallium" -- someone has to do a whole heap of work to make it happen. The attractive thing about Gallium is that it seems like a sufficiently useful foundation that there's a good chance that the work will get done.

which things are you saying have only 25% speed?

I realize you weren't asking me but I'll answer anyways

The whole speed thing is hard to put a single number on. I do expect that some things will run as fast on the open driver as on our proprietary driver -- but in other cases there may be a 5:1 difference. We do expect that open source drivers will be fast enough for many users, and some of the rest will simply get a slightly faster graphics card and not worry about performance.

anyway, that stuff SHOULD become faster with the documentation released, right?

We will be providing enough information and support to make open source drivers which run just as fast as the proprietary ones. The question is whether the open source community has the resources to write a driver which takes full advantage of the information even including development resources provided or funded by HW vendors.

[Svartalf]

@Redeeman

That's a value in your dreams. Because several things are completely missing like GLSL support and glxgears is really no benchmark - that it would be possible to archive a similar thruput for extra simple operations yes, but you must be really lucky to get maybe 25% speed for more demanding things.

Once the GLSL is in place with an optimizing compiler the performance will majorly improve. But, that's a bit off yet, I fear- at least another 6-12 months before we start really seeing the results of the work in progress with Gallium3D, etc.

[duby229]

We will be providing enough information and support to make open source drivers which run just as fast as the proprietary ones. The question is whether the open source community has the resources to write a driver which takes full advantage of the information even including development resources provided or funded by HW vendors.

My guess is it'll prolly never be quite as fast as the closed drivers simply due to the vast amount of resources that are devoted to that code base. But I'd be happy with 3/4.

But in the same aspect it also makes sense that the open drivers will improve with each new release. It may never catch up with the closed driver, but by the time a card reaches the end of it's life it should be pretty close.

[bridgman]

Yep. I do agree that most people will be happy with 3/4.

One of the little revelations that made this plan possible was "y'know, GPUs *are* pretty fast these days"