AMD Catalyst 9.11 For Linux Released

[yotambien]

First lines from the Wikipedia entry for ATI FireGL:

The ATI FireGL range of video cards is a series fabricated by ATI for use with CAD (Computer Aided Design) and DCC (Digital Content Creation) programs, usually found in workstations. They are very similar in hardware to the Radeon range.

And Nvidia Quadro:

The Nvidia Quadro series of AGP and PCI Express graphics-cards comes from the NVIDIA Corporation. Their designers aimed to accelerate CAD (Computer-Aided Design) and DCC (digital content creation), and the cards are usually featured in workstations. (Compared to the NVIDIA GeForce product-line, which specifically targets computer-gaming).

Nvidia explains the differences between the gaming and the professional cards here (pdf document).

So while you can certainly play a game with these cards, it is not their principal purpose. And even though the hardware appears to be very similar, you also pay for a firmware, driver support and feature set targeted at their intended use, which is different to that of gaming chips. The effort in terms of man power of keeping a more rigurous QA testing, development of features and optimisations in a market where these things actually matter and quite possibly a more direct relation between customer and manufacturer can not be dismissed only because they are not visible in the hardware itself.

[barkas]

The effort in terms of man power of keeping a more rigurous QA testing, development of features and optimisations in a market where these things actually matter and quite possibly a more direct relation between customer and manufacturer can not be dismissed only because they are not visible in the hardware itself.

You may believe that if you choose to. But don't be too sure that something is better just because it's more expensive and has a fancy name.

The reality is: same drivers, same hardware, at least in that case. That mobility firegl is not supported by either the firegl drivers nor the radeon mobility drivers, the only windows driver is an over 2 year old lenovo driver, which does not work all that well.
Regarding linux drivers, it uses the very same fglrx (with all the same bugs).

So, to come back to my original point, the OSS drivers are still better than fglrx, even for low texture/high poly workloads that CAD workstations have been tuned for in the past.
I also have not been able to spot any better quality or optimization for anything. I have had much the same experience with ati drivers everyone else here has probably had. The only reason I even bought the hardware was because AMD had started to open source the hardware documentation, so I could hope for a decent driver in the distant future.

By now I'm quite happy with 2.6.31/KMS/DRI2/radeon.

[yotambien]

I agree that market fragmentation must play a role here, but I doubt that companies are so incompetent as to keep buying more expensive hardware due to marketing hype.

Also, I bet that many of the perceived bugs of fglrx typically mentioned in this forum have their roots in running the driver in environments where it wasn't thorougly tested or not even intended to run in the first place. Look at the usual complains like 2D slowness with a composited desktop, lack of video hardware acceleration or incompatibility with weeks old kernels or X servers...Well, guess what, I compile a kernel once a year or so, I upgrade important packages only if there's a reason for it, and I don't use fancy, useless stuff. With this in mind I find that fglrx runs extremelly well here (although it hasn't been always the case), wiping the floor with radeon in terms of 3D performance and power management. Now I'm kind of locked to a certain kernel release because my hardware is not supported anymore (which sucks), but I'm yet to see what good reason would I have to throw this out of the window and use an experimental graphics stack.

If you are happy with DRI2/KMS/radeon and what not, good for you. We all choose what works best out of the available options, unless you have 'issues' with some of those options.

[barkas]

I agree that market fragmentation must play a role here, but I doubt that companies are so incompetent as to keep buying more expensive hardware due to marketing hype.

Also, I bet that many of the perceived bugs of fglrx typically mentioned in this forum have their roots in running the driver in environments where it wasn't thorougly tested or not even intended to run in the first place. Look at the usual complains like 2D slowness with a composited desktop, lack of video hardware acceleration or incompatibility with weeks old kernels or X servers...Well, guess what, I compile a kernel once a year or so, I upgrade important packages only if there's a reason for it, and I don't use fancy, useless stuff. With this in mind I find that fglrx runs extremelly well here (although it hasn't been always the case), wiping the floor with radeon in terms of 3D performance and power management. Now I'm kind of locked to a certain kernel release because my hardware is not supported anymore (which sucks), but I'm yet to see what good reason would I have to throw this out of the window and use an experimental graphics stack.

If you are happy with DRI2/KMS/radeon and what not, good for you. We all choose what works best out of the available options, unless you have 'issues' with some of those options.

They may have unlocked some features like line antialiasing and higher depth visuals, but I'm not sure if that's been the case since fglrx no longer is the workstation class opengl driver but the common driver for all linux stuff.

Point being, it's still not a good driver. As you also point out, it locks you to software components that are by now a bit old and is no longer updated for r500 and down. It will not be long until those software components hold you back from so much newer and nicer stuff that you will have to switch, too.

Still, if it works well for you, I won't argue with that, but it didn't work very well for me.

[yotambien]

Off topic (slightly)

[...]the only windows driver is an over 2 year old lenovo driver, which does not work all that well.

I think the laptop manufacturer is to blame there. The most recent ATI driver I can fetch from the manufacturer's site is from...2004 (and I bought the laptop less than 4 years ago). It doesn't even run Quake 3 properly. I had to use one of those modding things to install a decent version of Catalyst. It may or may not be what you are experiencing, I just thought I'd mention it.

[disi]

For me it's a really good improvement. Switching between workspaces is a little delayed if I use xcompmgr, tuxracer goes very well now on r740 HD4770

http://omploader.org/vMnVkcg

enemy-territory still crashes...

...loading libGL.so.1: Initializing OpenGL display
...setting mode 6: 1024 768
Using XFree86-VidModeExtension Version 2.2
XF86DGA Mouse (Version 2.0) initialized
XFree86-VidModeExtension Activated at 1024x768
Received signal 11, exiting...
Shutdown tty console

[cutterjohn]

Yup. I've gone back and re-enable compiz/beryl again. Notifications in Ubuntu now seem to work properly or at least show up when they're actually supposed to. (GNOME must be becoming highly dependent on compositing as is the Ubuntu notification system apparently.)

Video playback is MUCH slower with compositing enabled though, which I've noticed since I watched a lot (for me anyways) of videos the other day.

Compositing effects are, generally, horribly slow however once a window is open and in place it seems to have little affect on update speed for most 2d/3d apps excepting video playback(mplayer).

Looks like that thing with black rectangles poppping up is entirely gone as I still haven't seen it again in a while. Still no idea if it was an X.org, kernel, or fglrx bug as ALL of those changed for me right around the same time and then some while afterward noticed the lackof black rectangles.

Firegl: my understanding is that the hardware is identical to consumer GPU hw excepting that firegl tends to have MUCH more VRAM, has better warranty/support(apparently you get to talk to driver devs and request specific fixes), but also costs MUCH more because of that.

wine: nVidia/ATI is representational already. ATI has what? Something like 20% of the GPU market and nVidia 70 some%?

drivers: bottom line nVidia BLOB just works better in all applications be it workstation OR consumer as compared to fglrx/catalyst although AMD is making some progress, yet it's incredibly slow progress. (Still waiting for the OSS driver to get enough 3D support and then hope that's enough or I'll get serious about finding another MXM card plus get a heatsink made somewhere...or see if I can part one that would work from MSI...)

[bridgman]

Firegl: my understanding is that the hardware is identical to consumer GPU hw excepting that firegl tends to have MUCH more VRAM, has better warranty/support(apparently you get to talk to driver devs and request specific fixes), but also costs MUCH more because of that.

The driver optimizations are also quite different, at least on Windows. Not sure for Linux but will try to find out.

wine: nVidia/ATI is representational already. ATI has what? Something like 20% of the GPU market and nVidia 70 some%?

Last numbers I saw were 52.7% for Intel, 24.9% for NVidia and 19.8% for ATI/AMD.

http://jonpeddie.com/press-releases/...ddie-research/

[cutterjohn]

The driver optimizations are also quite different, at least on Windows. Not sure for Linux but will try to find out.



Last numbers I saw were 52.7% for Intel, 24.9% for NVidia and 19.8% for ATI/AMD.

http://jonpeddie.com/press-releases/...ddie-research/

Those are "this quarter" numbers which implies to me current sales which are meaningless since the vast majority of users don't upgrade GPUs every 3m, as much as nVidia, AMD, etc. would like them too. (Apple I'm looking into your cynical eyes w/o user-replaceable notebook batteries and aging GPUs...) Every 3y or so would be a more likely number, and the numbers that I referred to were based upon Steam's periodic surveys which have remained fairly constant over the years.

[bridgman]

Not sure I understand the connection between Steam numbers and market share. Steam surveys reach a specific subset of users and pretty much exclude the vendor with the largest market share (Intel). If you look at "Steam vs GPU sales" over the years you'll find they tend to be out of phase - some years one vendor is winning at gaming and the other is winning at everything else, then a couple of years later the positions reverse.

If you're really looking for installed base market share you need to do a weighted accumulation of market share numbers (using something like JP reports) for at least 5 years across desktop and mobile -- and that's going to come out more or less even between the three top vendors.

The marketing folks know these numbers a lot better than I do -- but I can say that Steam and total GPU sales numbers don't track very well, nor are they supposed to.