How well does HD4670 perform with open source drivers?

[bridgman]

Taking 12 months (it's a bit more than that even) to complete the planned architectural changes *is* slow progress.

Ahh, OK... sounds like we agree on the facts and just have different views of what constitutes a reasonable rate of progress.

Carry on

[RealNC]

@RealNC

why don't you just go back to the ati blob (or buy an nvidia +blob) and stop bitching??

I'll do whatever I feel like without asking for your permission, thank you very much. Funny though, how you support "open" and "free", but then tell others to shut up. Hey, I'm free to express myself too, you know. Fascist.

[89c51]

I'll do whatever I feel like without asking for your permission, thank you very much. Funny though, how you support "open" and "free", but then tell others to shut up. Hey, I'm free to express myself too, you know. Fascist.

fine with me

i just proposed a solution to your problem

keep bitchin

[Not so pretty now]

How about this:

Will radeon HD4670+open source drivers give me super stable 125fps in quakelive with "r_finish 1" at resolution 1920x1020/2=960x510 (cpu will be AMD Athlon II X2 255)? And if I will be patient enough then I will get pretty much same benefits one day as nvidia blob's vdpau?

[elanthis]

its much better to have an full featured driver than an fast but only openGL2.1 driver without any wine extansions and s3tc and openCL and without shader based video acceleration..

I'm not sure I should get involved in this, but... that really depends on what you're doing with your computer.

If you're trying to run an OpenGL 2.1 application that doesn't need Wine, s3tc, or OpenCL, and isn't playing videos, but it does render an awful lot of triangles with complicated shaders, then I'm willing to bet speed becomes a lot more important to you than any of those other things.

In the general case, it's still a trade-on. I don't know of any interesting Free/Open apps that make use of OpenCL, but I have more than a few that drag ass on my desktop so badly that they are totally unusable as they run more like a slideshow than an interactive application. I'd much, much, MUCH rather have some speed updates before I get something like Wine compatibility (I have two other computers with Win7 on them, and XP in a VirtualBox VM on my main Linux desktop which can passthru DX/GL to the host OS, so Wine is completely and utterly useless to me). Likewise, video playback is optional for many users (I have two 42" TVs and hardware dedicated to playing both discs and streaming videos, so I don't need exceptional video playback on LInux). OpenGL 3/4 support would be nice, but not if the driver still can't run OpenGL 2.1 apps with a usable level of performance.

Ideally of course there'd be enough manpower to work on performance _and_ features.

Practically, there isn't, and I'm not putting any effort into the driver (nor are the vast majority of people in this thread), and I'm pretty sure no more than a precious few of the Mesa/kernel/driver/Gallium/X devs have ever received any money from me directly or via an employer I purchased products from, so I have zero rights to bitch about or demand any particular development priorities.

[BlackStar]

Taking 12 months (it's a bit more than that even) to complete the planned architectural changes *is* slow progress.

(a) Going from zero to OpenGL 2.1 in 12 months isn't slow progress, it's amazing progress.

(b) Fglrx/nvidia devs start working on driver support somewhere between 6-18 months before the hardware is released (Bridgman can provide a more accurate timeframe). Open-source development used to start *after* the hardware was released, hence the 12month gap. (This has changed now and future hardware will be brought up faster).

(c) Fglrx has been rewritten once, just like the OSS drivers. Gathering from bits and pieces of information, the process took around 24-36 months - just like the OSS drivers! (Brigdman can again give a more exact timeline).

(d) The OSS drivers have to develop and support many more devices than either fglrx or nvidia. Look at the big picture for a moment.

Now, I know your attitude is "my 4870 runs better on fglrx" but this is short-sighted. You are losing the forest for the tree.

[pingufunkybeat]

for opensource software you can just drop the openGL2.1 renderer and port the engine to OGL3 and this ogl3 renderer is just faster means faster per watt usage for the same result.

why is openGL2.1 so importand for opensource software ?

opensource means any OS-Game can be portet to an modern OGL3 engine.

Sounds great.

When are you going to do it?

[Panix]

Panix, are you sure you are looking at the right page ?

6xx/7xx/Evergreen 3D boxes are all marked "MOSTLY" (ie implemented with known bugs) except for the just-added GL3/GL4 feature rows (eg geometry shaders) and performance enhancements like texture tiling. The transition from "MOSTLY" to "DONE" is always slow because there is no clear rule for when the transition should happen.

The WIP entries are for Northern Islands not Evergreen. Only one of the NI models has even been released...

"MOSTLY" has been unchanged since? Forever? I don't know what it means.

Much of the feature set is still not available via an OSS option. That was my point. So, a new architecture even based to some degree on the previous... one can expect another long wait.

Also, posts like "HD 5xxx' still sucks in Linux" doesn't inspire...

I want to buy a card for Xmas but Radeon fans aren't convincing yet.

[Kano]

gtx 580 or

[bridgman]

"MOSTLY" has been unchanged since? Forever? I don't know what it means.

It's been at mostly for about 11 months. Last time you asked it was about 10 months. There is a legend at the front of the wiki page:

?"DONE" means that it is implemented and relatively bug-free.
?"MOSTLY" means that it is implemented but has some known bugs.

As you can see, the transition from MOSTLY to DONE is controlled by the interpretation of "relatively bug-free" vs "some known bugs", which is not particularly specific.

Right now I believe the community is focusing on moving to r600g and getting *that* relatively bug free rather than worrying about whether the classic r600 driver meets that criteria. It probably does, but I don't think anyone cares.

Much of the feature set is still not available via an OSS option. That was my point. So, a new architecture even based to some degree on the previous... one can expect another long wait.

... but most of the feature set used by today's applications *is* available via an OSS option. If you are buying a card to run Heaven, or to develop using GL 3 or 4 then you need to plan on using the proprietary drivers for a while longer.

Also, posts like "HD 5xxx' still sucks in Linux" doesn't inspire...

How recent is the post and was the poster actually running current code ?