Slashdot Mirror

The resulting WEC7 driver for Radeon GPUs is proprietary, but that's allowed per the MIT license that the ATI-AMD Linux driver code is provided under."

I wonder what the source of the article is, that the driver will be proprietary.

I wonder if taking the code written by the community, port it to windows embedded and add some proprietary sauce to it was also part of the original plan. So maybe AMD sees Linux community as a form of cheap labor?

Nope. This is a fairly recent project, conceived and started years after we got back into supporting open source graphics drivers, and the request was for a driver that could be released in source code form.

Source

Sorry, I started dismissing it when they sat on a fairly major bug/regression for a week, and then had a big article about how bugs/regressions do not get fixed in the kernel.

If you find a bug/regression report it to LKML, don't expect the linux devs to watch the output of your little test suite. Don't get me wrong, I really like the testing, just that when they see something like this, and have it narrowed down to a few commits, freaking report it so it can get fixed.

It seems likely that politics has a role to play here. Qt came into the MeeGo project from Nokia. Despite recent moves towards open governance, is still very much associated with Nokia. Intel were unhappy that Nokia switched to Windows Phone and the member of LiMo (including Samsung) may prefer to avoid mentioning or relying on what is perceived to be a competitor's asset.

Not just that. Samsung has been sponsoring Enlightenment, and they may see it as being better for low-powered devices than Qt.

http://www.phoronix.com/scan.php?page=news_item&px=OTQxNg
Right now, such a machine, as it is delivered to you today, can't run Linux. Well it can, really, using 1.8Gb of RAM.
This is happening, right now.

"except maybe AMD supporting more of them at the same time on most cards"
Oh man, I just checked and you're right ! Damn, guess I'll really have to consider retiring my current nVidia gpu + Matrox Dualhead2go setup...

Starts on Page 2.

And interestingly enough, KDE vs Unity on Nvidia shows the same numbers as BSD vs Ubuntu did: http://phoronix.com/forums/showthread.php?60243-FreeBSD-A-Faster-Platform-For-Linux-Gaming-Than-Linux&p=226522#post226522
So this is simply Nvidia having regressions on Unity/Gnome but not with KDE.

This claim keeps getting repeated, over and over by people who didn't RTFA. IT IS THE SAME HARDWARE. The operating systems report it differently.

As for Phoronix, nobody here seemed to have a problem with their previous benchmark showing Windows games running better on Linux Cedega. Now that Linux is shown to be losing in a benchmark, suddenly there are all these "problems" with the benchmark. This community is so biased.

What's funny is that people are finding any reason they can to dismiss the benchmarks (my favorite is claiming the hardware is different, when it's not).

Meanwhile, nobody seemed to have a problem with Phoronix's previous benchmark showing Wine/Cedega games running faster on Linux than on Windows. The difference now is that Linux is on the losing end of the benchmark, so it simply must be incorrect in some way.

OS bias is a funny and bizarre thing.

It should be "Linux 3D Games Run Faster On PC-BSD with KDE than on Ubuntu Linux with Unity, on NVidia hardware". It's worth to note that KDE outperforms every gtk based desktop environment (gnome, unity, lxde...) when running on NVidia hardware.

Didn't this happen when Linux started emulating Windows?

"Games run faster in Linux/Wine(Cedega) than in Windows"

http://www.phoronix.com/scan.php?page=article&item=681&num=2

Why is everyone so shocked that an emulation layer can be faster now, when before it was "look at us, we're great?"

Nope. NVidia themselves say that API churn in Linux is not so bad: http://www.phoronix.com/scan.php?page=article&item=nvidia_qa_linux&num=7 In fact, you can check NVidia's compatibility layer (it's distributed in source code). It's tiny, any adjustments are easy to do.

Anyway, closed-source drivers are not going anywhere. OpenSource drivers are slowly (very slowly) catching up with them, and native Linux drivers have huge advantage, they JustWork(tm).

That's not what they mean here. They are talking about making input scroll events (mouse wheel, presumably) be less jumpy and more smoothed out. See this article: http://www.phoronix.com/scan.php?page=news_item&px=OTUyNw

I went with 64-bit on the laptop because it was a fresh install and my laptop has 8GB of RAM. There's probably no need to "fix" your install if you have invested much extra time in setting it up for yourself. 64-bit can be a bit faster, but probably not so much that you'd notice it for everyday tasks. Though some tasks, like audio encoding and general compression, are a bit faster: http://www.phoronix.com/scan.php?page=article&item=ubuntu_natty_pae64

In other news, Fedora 15 just pushed out Linux kernel 2.6.40 (aka 3.0)! Thanks Fedora people! I hope it fixes my stability issues. If it does, I'll be happy to stay with Fedora and will probably move my desktop machine to it instead of Kubuntu.

You're wrong. It IS possible and it is very simple with xen 4.1 and Linux 3.0 as dom0.
An early example here, with a previous kernel:
Ubisoft Is Playing With Linux & Xen Virtualization

Which test of ZFS do you talk about? The most recent one http://www.phoronix.com/scan.php?page=article&item=zfs_ext4_btrfs&num=1 was with an AMD64 build of FreeBSD 8.1 on a machine with 4GiB RAM.

Um, no. The 3D driver in question is the R600 Gallium driver

The Gallium3D driver has been under development for nearly 4 years.

The Documentation for the R6XX has been released in December 2008 and the R6XX/R7XX programming guide in May 2009. It may be that the core Gallium 3D driver is in development for four years, but given the availability of the documentation, the R600 was developed for at most two and a half years. Actually, first support for the HD6000 series of graphics cards was added in January 2011.

Um, no. The 3D driver in question is the R600 Gallium driver

The Gallium3D driver has been under development for nearly 4 years.

The Documentation for the R6XX has been released in December 2008 and the R6XX/R7XX programming guide in May 2009. It may be that the core Gallium 3D driver is in development for four years, but given the availability of the documentation, the R600 was developed for at most two and a half years. Actually, first support for the HD6000 series of graphics cards was added in January 2011.

Um, no. The 3D driver in question is the R600 Gallium driver

The Gallium3D driver has been under development for nearly 4 years.

The Documentation for the R6XX has been released in December 2008 and the R6XX/R7XX programming guide in May 2009. It may be that the core Gallium 3D driver is in development for four years, but given the availability of the documentation, the R600 was developed for at most two and a half years. Actually, first support for the HD6000 series of graphics cards was added in January 2011.

...it's only advantage is being Open?

I can see how many people may not see a great cost/benefits ratio there...

Being open means that these drivers won't simply go away once the product line is deprecated in favour of the newest and coolest graphics card, and that it will be able to receive improvements and bug fixes essentially until the last working piece of hardware dies off.

Wewt! I can get speed improvements! Now, at their current rate or increase, it will only take 5 years for the driver to be able to perform at the same level as the proprietary driver.

Your prediction is likely incorrect. Look at what the open source nvidia guys were able to do through reverse engineering alone and no specs:
On Low-End GPUs, Nouveau Speeds Past The NVIDIA Driver

Hey I know some of the guys at Phoronix, and they began over on Linuxgames.Com. They're wholesome to report the abilities of their hardware as used on Linux comparisons.

Going back to the Benchmarks, did any of you actually read the article? The Gallium3D drivers outperformed the proprietary drivers in a couple tests, such as with V-Drift ( http://www.phoronix.com/scan.php?page=article&item=amd_hd6000_open&num=5 ). That's worth noting that something even in the Catalyst drivers is simply not up to par after all these years despite barely keeping above par with open source ( http://www.phoronix.com/scan.php?page=article&item=amd_hd6000_open&num=4 ).

I raise your grains of salt with my assalted beachfront property in Afghanistan, brought to you by The Weather Channel (the initial donators for ATI Radeon open source DRI driver development in XFree that POWERS the green-screen of their weather-tracker on Television).

Hey I know some of the guys at Phoronix, and they began over on Linuxgames.Com. They're wholesome to report the abilities of their hardware as used on Linux comparisons.

Going back to the Benchmarks, did any of you actually read the article? The Gallium3D drivers outperformed the proprietary drivers in a couple tests, such as with V-Drift ( http://www.phoronix.com/scan.php?page=article&item=amd_hd6000_open&num=5 ). That's worth noting that something even in the Catalyst drivers is simply not up to par after all these years despite barely keeping above par with open source ( http://www.phoronix.com/scan.php?page=article&item=amd_hd6000_open&num=4 ).

I raise your grains of salt with my assalted beachfront property in Afghanistan, brought to you by The Weather Channel (the initial donators for ATI Radeon open source DRI driver development in XFree that POWERS the green-screen of their weather-tracker on Television).

Yet good hardware accelerated 3D graphics is still not available in an open source driver for Linux. Phoronix's benchmarks show this quite clearly.

Yet, it's slowing coming in.

This just in. Apparently, on the HD6000 side of things, the performance gap is shrinking. We're still talking about a 2:1 gap, but it's better than the 10:1 reported earlier.
It takes time. Time for the OSS developper getting used to write good drivers for it. Time for good consolidated architectures to emerge and gain momentum (like the Gallium3D with a good separation between API front-ends and hw back-ends). Time for hw manufacturer to integrate OSS in their development pipeline : it took enormous time before the first documentation could be green-lighted for publishing by their legal department, currently they only lag a few weeks to a few months behind. By the time their reach the HD8000 generation (or whatever it will be called at that point in time). AMD promised that the OSS will be integrated into the development process from the beginning.

Contrast it with Intel : they've been much longer in the OSS game. Currently, as (non-PowerVR-based) hardware is rolled to stores, there is driver support available (okay: there are still hicups - initial sandy bridge was a buggy, wasn't available in mainstream distros and required pulling the latest development version). But that's still support released almost simultaneously. And benchmark show almost similar performance between the Linux (opensource only) and the Windows (blob) drivers.

I have 2 computers with Radeon cards, an X1500, and an HD5450. [...] Probably the open source drivers are emulating the 3D with software.

You're doing something wrong... You should check on your distro's forum if you didn't miss something somewhere.
Specially with the HD5450 :
- it's a mid-range card (the biggest performance gap happens on highest range of cards)
- it's a previous-to-latest generation of cards (by now bugs must have been ironed out).
It should perform decently.

Are you sure that you're getting latest up-to-date drivers from your distro's repositories? (Some distro use additional repositories to get the latest versions, otherwise you only get bug- and security- fixes for whatever version comes with the stock distro)
Are you sure you're running the *Gallium3d* variant of drivers? (the "r600g" driver ?) (Some distro still used the older variant "r600" by default. Gallium3d has been making gigantic leaps forward in the latest months).

(Also there's a bug affecting some AMD hardware users: you might need to add "irqpoll" on the boot parameters. read your /var/log/message log. If it complains about "irq nn: nobody cared (try booting with the "irqpoll" option)" and then "Disabling IRQ #nn", you might be affected by a bug which slowly brings the 3D acceleration to a crawl)

One of my machine has an AGP variant of HD4500. I run latest openSUSE + official repository from SUSE for latest X11 version. (I still have the stock kernel, so I don't benefit from some advantages of the latest kernel modules). I got an up-to-date Xorg and Mesa. The performance isn't stellar, but its decent enough.

I think overall it a sign of the whole graphics situation. Sufficient and decent OSS solution have started to appear. But saddly, it's not always clearly documented and made easy. For performance people need often the lastet version, a version newer than the one which came with stock distribution.
- But this latest version doesn't always exist (in the sandy bridge case, initial enthousiast needed to pull the source and compile it themselves ). It would be better if more collaboration between distribution+developpers+manufacturer happens. So we see more "official additional repositories" (the repositroy on openSUSE's buildservice to get latest X11 and Mesa is a nice starting point).

How do you define "far superior"? According to most benchmarks, LLVM still has some miles to go before it produces binaries that are faster than gcc (it does produce a few special cases where LLVM is faster though so it does show promise for the future). For example check out: http://www.phoronix.com/scan.php?page=article&item=gcc_46_llvm29&num=1

I did a little digging for those wondering: it does run Linux, but only with the proprietary Catalyst driver at the moment. Might be interesting once the open source driver catches up (assuming AMD shares the required info).

I think you are a little lost on this one.

Once upon a time BAPCo didnt even bother pretending that they weren't actually Intel.

These days, BAPCo pretends that they arent Intel. Its still Intel tho.

BAPCo is Intel.

The last time Intel so blatantly rigged the benchmark game was when the Athlon XP's were beating the shit out of the Pentium 4's. AMD has recently made a mockery of Intels Atom solutions, and the one leaked benchmark for the Bulldozer design must have Intel more than a little worried about its future bragging rights.. so here we are, with Intel blatantly rigging the benchmark game again.

Expect a new version of ICC shortly.

Will it run Linux?

I'm not being facetious, I got stung by the lack of support by Nvidia for their Optimus graphics cards on my ASUS U30JC.

Thankfully Martin Juhl has been working on a solution using VirtualGL, which gives us the use of our Nvidia cards under linux

wtf is this, ubuntu?
http://www.tomshardware.com/reviews/ubuntu-11.04-natty-narwhal,2943-13.html
2 hours lost?!!?
how can anyone write code that causes such a huge battery life reduction?

Ask the Linux Kernel developers as it's their fault... Check on Phoronix for details...
Another Major Linux Power Regression Spotted

First, Ubuntu 10.10 is based on Linux but it is not Linux any more than any other distribution built on top of a linux kernel is Linux.

So you're saying Ubuntu 10.10 isn't the true Linux because Linux works better than Windows. That sounds a bit like the No True Scotsman fallacy to me.

Consider the following Gedankenexperiment.

I upgrade my notebook to Windows 8. Suppose if my network card/graphics card/power management on my laptop starts behaving like like something from the 1990s performance-wise in Windows 8. I.e. the network card is slow (according to you), the graphics poorly accelerated (actually this applies to Ubuntu compared to Windows - see here http://www.phoronix.com/scan.php?page=article&item=linux_windows_part1&num=11) and the battery life much poorer (actually this applies to Ubuntu 10.10 compared to earlier versions - see here http://www.phoronix.com/scan.php?page=article&item=ubuntu_maverick_netbook&num=1).

Would that mean

1) Windows 8 sucks and I should switch back to the OS I got for free with the machine which actually works.

2) I'll still be secure in the knowledge that I made the right choice of OS. After all, just because all my hardware performs poorly in the current version of Windows doesn't mean the platonic ideal of Windows is in any way besmirched.

Now furthermore consider if the laptop came with a pre-installed, totally free copy of XP where all my hardware performed well and instead of Windows 8 I'd upgraded to Ubuntu where all my hardware performed badly. Would that mean Ubuntu sucks or not?

First, Ubuntu 10.10 is based on Linux but it is not Linux any more than any other distribution built on top of a linux kernel is Linux.

So you're saying Ubuntu 10.10 isn't the true Linux because Linux works better than Windows. That sounds a bit like the No True Scotsman fallacy to me.

Consider the following Gedankenexperiment.

I upgrade my notebook to Windows 8. Suppose if my network card/graphics card/power management on my laptop starts behaving like like something from the 1990s performance-wise in Windows 8. I.e. the network card is slow (according to you), the graphics poorly accelerated (actually this applies to Ubuntu compared to Windows - see here http://www.phoronix.com/scan.php?page=article&item=linux_windows_part1&num=11) and the battery life much poorer (actually this applies to Ubuntu 10.10 compared to earlier versions - see here http://www.phoronix.com/scan.php?page=article&item=ubuntu_maverick_netbook&num=1).

Would that mean

1) Windows 8 sucks and I should switch back to the OS I got for free with the machine which actually works.

2) I'll still be secure in the knowledge that I made the right choice of OS. After all, just because all my hardware performs poorly in the current version of Windows doesn't mean the platonic ideal of Windows is in any way besmirched.

Now furthermore consider if the laptop came with a pre-installed, totally free copy of XP where all my hardware performed well and instead of Windows 8 I'd upgraded to Ubuntu where all my hardware performed badly. Would that mean Ubuntu sucks or not?

And how's that working out for them?

http://www.phoronix.com/scan.php?page=article&item=amd_gardenshed_drm&num=6

Published on May 27, 2011
Written by Michael Larabel
Page 6 of 6

There's really no change compared to our previous round of open-source Radeon DRM + Gallium3D vs. Catalyst driver testing... Most users interested in playing any sort of semi-intense OpenGL game will certainly be best off with the Catalyst driver for the near future.

If taking a geometric mean of all the frame-rates for each of the tests, from the OpenBenchmarking.org result file, it clearly shows the Catalyst driver is still many times faster overall.

If then condensing the results from the four Radeon HD 4000/5000 series graphics cards, the AMD Catalyst 11.4 Linux driver is about 7.76x faster than the latest open-source Radeon Linux driver code as of yesterday..

Not so good it seems.

5900:
http://www.phoronix.com/scan.php?page=article&item=amd_firepro_v5900&num=1
7900:
http://www.phoronix.com/scan.php?page=article&item=amd_firepro_v7900&num=1

5900:
http://www.phoronix.com/scan.php?page=article&item=amd_firepro_v5900&num=1
7900:
http://www.phoronix.com/scan.php?page=article&item=amd_firepro_v7900&num=1

And for those of you who would like to actually see the Phoronix article mentioned in the summary, it's here

(Yes, there are obnoxious ads, but only if you turn off your ad blocker and Flash blocker and mouse over the double-underlined blue words.)

It's worth looking again: http://www.phoronix.com/scan.php?page=article&item=nouveau_2639_flip&num=1

Yeah! Right on!

Yeah! Right on!

Yeah! Right on!

On another note, look here for a performance comparison. We can see that NTFS loses by a pretty big margin.

NTFS in Windows XP. Please don't tell me you actually think thats still relevant?

Generally: I never said that windows completely sucks in all these regards (it surely improved a lot with 7 and they do quite some things in a good way!), just that they should concentrate on improving their core instead of building fluffy stuff.

Performance: of course Windows has advantages, but nevertheless Linux wins regarding boot time and I/O performance. I didn't want to say that Windows performance is bad overall, but that they should rather focus on areas where it isn't instead of building PDF viewers or image manipulation software.
Documentation: true, but overall I'm a lot more satisfied with man/apropos than with Windows documentation.
File Systems: Yes, what I really meant is that ext3/ext4 is practically a lot better regarding fragmentation than NTFS. In practice defragmentation may not be needed at all. On another note, look here for a performance comparison. We can see that NTFS loses by a pretty big margin.
UAC/SUDO: Basically UAC simply grants you full rights for one particular action right now, whereas sudo can grant you a wide range of rights for X amount of time for Y numbers of actions. You can argue about them, but they are not the same.
Hardware: By default sure, but you can customize linux to run on *very* low hardware in comparison to windows. And I'm not forced to use a bloated window manager.
Registry: Here's a pretty good summary of what's wrong with the windows registry.

I would not put too much faith on the theoretical 2x performance increase offered by AVX this generation. Although the processing hardware is there, the processor itself is not designed with enough fetch width to keep the units fed.

See Here

Summary: a few tests topped %20 improvement.

And Here

Summary: a couple of tests topped %20 improvement. Many tests produced slower results with AVX enabled.

The overall performance boost looks to be around %20-30, at least for now.

The only thing I would like to see is a free software/free driver challenge between the two.

Well that would be a walkover because nVidia doesn't play in that category. Whatever you may have heard of the Nouveau driver, it is not done by nVidia, they don't want the project, they don't help them with documentation, specifications or answer questions. It is not in any shape or form nVidia's project.

The only thing nVidia has open sourced - and even that is arguable since it was obfuscated - is an extremely simple 2D only driver, which is in maintenance mode and will never support Fermi+ graphics cards, DisplayPort or any form of acceleration. Now they ask you to limp on the VESA driver until you can install their proprietary driver.

As a user, it doesn't matter who wrote the driver. All that matters is how well it works, with or without the support of nVidia.

To be honest I don't get the Nouveau project at all, fair enough maybe these people have nVidia cards they want to make work but it looks like mission impossible. At least with AMD you do have the company's specs and backing and still it seems a very hard problem to solve. And you could work on all the common issues in the open source stack, of which there are plenty...

What makes you think it's "mission impossible"? Just take a look at recent 3D benchmarks; Nouveau's 3D performance is already competitive with the nVidia blob on virtually all recent cards except the very high-end ones.

Sure, it's far from having feature parity with the blob in most areas. I'm aware that it has stability issues on some hardware. But its mission of usable drivers for nVidia hardware is far from impossible - the progress that has already been made is impressive, and it's getting better very fast.

The only thing I would like to see is a free software/free driver challenge between the two.

Well that would be a walkover because nVidia doesn't play in that category. Whatever you may have heard of the Nouveau driver, it is not done by nVidia, they don't want the project, they don't help them with documentation, specifications or answer questions. It is not in any shape or form nVidia's project.

The only thing nVidia has open sourced - and even that is arguable since it was obfuscated - is an extremely simple 2D only driver, which is in maintenance mode and will never support Fermi+ graphics cards, DisplayPort or any form of acceleration. Now they ask you to limp on the VESA driver until you can install their proprietary driver.

As a user, it doesn't matter who wrote the driver. All that matters is how well it works, with or without the support of nVidia.

To be honest I don't get the Nouveau project at all, fair enough maybe these people have nVidia cards they want to make work but it looks like mission impossible. At least with AMD you do have the company's specs and backing and still it seems a very hard problem to solve. And you could work on all the common issues in the open source stack, of which there are plenty...

What makes you think it's "mission impossible"? Just take a look at recent 3D benchmarks; Nouveau's 3D performance is already competitive with the nVidia blob on virtually all recent cards except the very high-end ones.

Sure, it's far from having feature parity with the blob in most areas. I'm aware that it has stability issues on some hardware. But its mission of usable drivers for nVidia hardware is far from impossible - the progress that has already been made is impressive, and it's getting better very fast.

They only reported it because another website reported it falsely. Don't blame them for another website's mistake.

Give it a week. It will be on http://www.phoronix.com/ And it will be more likely to be accurate. Of course it will have real god and useful data soon at http://openbenchmarking.org/ but that is actually helpful and will not be reported by anyone.

There is a real God? With useful data.. AND His data will be accurate?
Wow...

Phoronix does these kinds of things all the time. http://www.phoronix.com/scan.php?page=category&item=Graphics%20Cards I wish more people know about it. And while they are a Linux Geek place, the benchmarks include Windows and OSX often.

Give it a week. It will be on http://www.phoronix.com/ And it will be more likely to be accurate. Of course it will have real god and useful data soon at http://openbenchmarking.org/ but that is actually helpful and will not be reported by anyone.

http://www.phoronix.com/scan.php?page=article&item=gcc_llvm_clang&num=2

LLVM/Clang generates binaries that are faster than GCC as often as not..

Similar results are obtained with any compiler vs. gcc, and even from one version of GCC to the next.

This release adds support for the AMD Fusion GPU+CPUs

Only if you want your box to reboot if you try a serious game. Maybe next time.

But Linux shall. Yes, native Direct3D.

For those thinking that Direct3D 10/11 on Linux will be sub-par, "Finally, a mature Direct3D 10/11 implementation is intrinsically going to be faster and more reliable than an OpenGL implementation, thanks to the dramatically smaller API and the segregation of all nontrivial work to object creation that the application must perform ahead of time."