Valve Ships An AMD Preview Driver For SteamOS

[dffx]

They are marketing Iris Pro based Gigabyte Brix as Steamboxes. Inherent to that, you are already limited to the GL version of Mesa.

All the whiz-bang features of ogl 4+ aren't getting used anyway until Mesa catches up. Might as well have AMD work on that rather than keep their Goliath 50MB blob mess of Windows code that keeps breaking.

While this is true, there are a couple things to remember:

1. The Gigabyte machine is a low-to-medium powered machine that is not likely to do a lot of heavy lifting, ie, games that require complex or demanding graphics (think: primarily intended for a lot of the 2/2.5D indie games that have compromised the bulk of Linux releases over the past year). AMD cards offer much more powerful hardware, and are much pricier, so you're going to want to get as much performance out of them as possible, for the fanciest games on the market, which the Mesa drivers do not offer yet.

2. You're right that most games don't require OGL4, but many AAA games *do* require at least OGL3.3 -- which the Intel drivers offer, but the AMD radeon Gallium drivers do not.

[dffx]

Because no edit...

I also want to just make it clear that I agree with the sentiment that the Gallium driver should be a priority -- hell, that's what I use on a daily basis, I haven't touched Catalyst in about 3 months now.

But when OEMs are looking to build machines with up to date, modern AMD hardware you have to consider the fact that they want to market it to people who want big, powerful machines -- and if you're putting GCN-based AMD cards in a machine, you don't want to only be getting a smaller fraction of possible performance that can't play big-name titles, you want to use the drivers that give you the most performance possible. And that means using Catalyst for the time being.

I have no idea how many people AMD has working on the proprietary driver, but I feel as strongly as anyone that more attention should be focused on the open source drivers. Hell, that's the reason I haven't gone out and bought a new Nvidia card, because I'm excited about the prospect of having an entirely functional, performant open source driver stack in the very near future (we have one currently, depending on your measuring stick).

[V10lator]

Metro Last Light isn't.

Let me guess: Bug #71239?

If you're a lucky winner (able to use fglrx) then please have a look here http://phoronix.com/forums/showthrea...tro-Last-Light cause we really need that damn trace.

[mrugiero]

2. You're right that most games don't require OGL4, but many AAA games *do* require at least OGL3.3 -- which the Intel drivers offer, but the AMD radeon Gallium drivers do not.

Yes, but implementing 3.3 support is not such a great task, starting where they are now. Even more, if the idea is to reallocate the resources to it, it should get done pretty fast. It's not like someone suggested that somehow the open drivers should catch up in a short time to 4.4.

I also want to just make it clear that I agree with the sentiment that the Gallium driver should be a priority -- hell, that's what I use on a daily basis, I haven't touched Catalyst in about 3 months now.

But when OEMs are looking to build machines with up to date, modern AMD hardware you have to consider the fact that they want to market it to people who want big, powerful machines -- and if you're putting GCN-based AMD cards in a machine, you don't want to only be getting a smaller fraction of possible performance that can't play big-name titles, you want to use the drivers that give you the most performance possible. And that means using Catalyst for the time being.

I have no idea how many people AMD has working on the proprietary driver, but I feel as strongly as anyone that more attention should be focused on the open source drivers. Hell, that's the reason I haven't gone out and bought a new Nvidia card, because I'm excited about the prospect of having an entirely functional, performant open source driver stack in the very near future (we have one currently, depending on your measuring stick).

True, but that doesn't stop you from making the open driver the default in custom installs of SteamOS. The OEMs can change this on the images they use, it shouldn't be hard.

[blackout23]

They are marketing Iris Pro based Gigabyte Brix as Steamboxes. Inherent to that, you are already limited to the GL version of Mesa.

All the whiz-bang features of ogl 4+ aren't getting used anyway until Mesa catches up. Might as well have AMD work on that rather than keep their Goliath 50MB blob mess of Windows code that keeps breaking.

Yes games will continue to target OpenGL 3.2 thanks to Mac OSX, which is probably still the prime platform for OpenGL Mac/Linux ports. Still catalyst is faster no matter how much that hurts some FLOSS fanboys feelings.

[dffx]

Yes, but implementing 3.3 support is not such a great task, starting where they are now. Even more, if the idea is to reallocate the resources to it, it should get done pretty fast. It's not like someone suggested that somehow the open drivers should catch up in a short time to 4.4.

I agree. I want RadeonSI to get up to 3.3 more than anything, but having watched the mesa-dev list for the past few months it seems like they hit 3.1 and have kind of taken a break from working on OpenGL compliance for a bit while they clean up code, implement bug fixes, etc. Maybe you're right and putting more people on the project would help. I'm not honestly sure.

True, but that doesn't stop you from making the open driver the default in custom installs of SteamOS. The OEMs can change this on the images they use, it shouldn't be hard.

No, but I think the idea here is that SteamOS is intended to be directed towards people who don't have much (or any) experience with Linux. I agree that the OS could provide some handy guides for installation (ie, "Do you plan on using this for small indie games or more hardcore 3D games?" and install accordingly), but then, I think there's a lot of things Valve should be doing with SteamOS to push adoption that I don't think they're doing (but that's for another topic). Point is that having to make people choose

[GreatEmerald]

Still catalyst is faster no matter how much that hurts some FLOSS fanboys feelings.

Faster, but less reliable. For instance, the last demanding game I tried running on fglrx, Expeditions:Conquistador, fglrx had major performance problems while scrolling text. Seriously. When not doing that, the performance was OK, but if at any point you decide to scroll text, bam, you get huge lag spikes. That is really silly.

Personally I'm using an NVIDIA card right now, but that's mostly because I didn't have any and it's nice to have cards from all three vendors to monitor the situation. I agree that AMD should be pushing the FOSS driver instead of Catalyst going forward, but yes, that might take some time. Their progress on the FOSS driver as of recently has been quite amazing, and thus they deserve some credit for that. It's now pretty much their strong point. For instance, now I'm contemplating getting a simple video card that would be cool and quiet (as opposed to the large monstrosity I'm sing right now) for my HTPC, and AMD has pretty much no competition there. Intel is not an option since they don't have dedicated cards, and NVIDIA is not worth it because it doesn't make sense to burden such a device with a giant graphics blob, and Nouveau doesn't have DPM support that is rather essential. So perhaps an R5 210 will be in order, to see how radeonsi evolves, once it's released (if it ends up using radeonsi).

Overall it's interesting to see trends changing. Just last year AMD cards were mostly good for netbooks, where its competition was using PowerVR chips. Now Intel has pushed in front in that regard, with dropping PowerVR and using their own integrated graphics (although they still have their UEFI situation etc. to sort out). A minor point for AMD was that it allowed greater flexibility with two drivers that were both OK at some tasks. Now the FOSS driver certainly fares better, which means that they are a good choice for lower-end dedicated cards and light gaming, as well as being not far behind Intel with regards to laptops. And even NVIDIA has come forward with support for Nouveau, so we might see even that changing.

[agd5f]

I agree. I want RadeonSI to get up to 3.3 more than anything, but having watched the mesa-dev list for the past few months it seems like they hit 3.1 and have kind of taken a break from working on OpenGL compliance for a bit while they clean up code, implement bug fixes, etc.

We've been working on 3.3 support internally for a while. When the code is ready we'll push it out.

[dffx]

We've been working on 3.3 support internally for a while. When the code is ready we'll push it out.

I'm excited for the release! Thanks again for all the hard work you open source ATI/AMD guys have been giving lately. Like I mentioned earlier, radeon has been my daily driver not just for desktop stuff but for 3D stuff for a few months now, and it's been a great ride.

[Gps4l]

That driver is no improvement over 13.12 or other drivers in L4D2. I even patched the steamos package for kernel 3.13 support just to verify how bad that driver is. Then i could have stayed with 13.12 too...

Thank you but

I was hoping to finally have LFD2 run good, without the stutter / freezing.