XBMC Developers Criticize AMD's Linux Driver
An anonymous reader writes "It's not only the NVIDIA Linux driver that has been publicly slammed over lacking support; the AMD Catalyst driver is now facing scrutiny from developers of the XBMC media and entertainment software. The developers aren't happy with AMD due to not properly supporting video acceleration under Linux. The AMD Linux driver is even lacking support for MPEG2 video acceleration and newer levels of H.264. AMD reportedly has the support coded, but they're refusing to turn it on in their public Linux driver."
But do they at least support this functionality?
ATI has always sucked balls. To this day, I cannot fathom why AMD paid real money for that company.
What leeches? The drivers don't cost the user anything extra (far as I know?). If I've already paid for the hardware, I expect drivers that work and support all the functionality, and there is no valid excuse for any hardware manufacturer to withhold them.
To me, the most interesting part of the summary was:
AMD reportedly has the support coded, but they're refusing to turn it on in their public Linux driver
The relevant point from the article seems to be
Our sources say that these features are implemented in fglrx since a long time, but simply not activated within the driver. Nobody seems to know why.
Forgive me for being skeptical with Phoronix, but does anyone with more direct knowledge of these "sources" want to comment? I'd like to have a better view of the situation than just the words "Our sources."
Just this past week I rebuilt my HTPC going from Boxee (which orphaned its support of Linux) and went to XBMC. I have personal knowledge of the dumb problems with the Catalyst driver.
XBMC is a project whose users take a lot of advantage of old hardware. The other part are dealing with small form factor hardware. A lot of it does happen to be proprietary garbage. In my case I purchased a Dell Zino several years ago for the task. There isn't much choice about for these items, and rolling you own at this size is often clunky (though a lot more feasible now than 3 or so years ago). You're going to find a lot of Nvidia (no fucking way) and AMD.
So you have one group of people that are re purposing and one group with specialty hardware. Not a lot of hardware choice in either, really.
So, yeah, this is a big deal. There is no real reason from my point of view not to provide a good driver for my platform of choice.
Windows users don't pay for their video drivers. Both Windows and Linux users have paid the same amount for the hardware, though. You must be kind of stupid.
Does this mean we can publicly criticize the XBMC UI?
YHBT
> Why should they support Linux leeches? Get a job and pay for your software losers and then maybe you'd see some support.
This is HARDWARE you moron. Everyone that has the hardware has PAID for it you moron. Get back under your bridge.
Linus may want to cuss out the guys at Nvidia but they're doing a better job in this case.
A Pirate and a Puritan look the same on a balance sheet.
Why should they support Linux leeches?
There's a reason why all my Linux boxes except the oldest one have either Nvidia or Intel graphics. In the case of the old one I had to manually patch the ATI driver kludge source because ATI dropped support and a kernel change broke it.
So, not a penny of my IT budget has gone to ATI since 2008 because their drivers aren't very good and they don't support them.
It's clear you don't know a thing about hardware, linux, amd, ati, nvidia and drivers. Seriously, get a fucking clue before you post your crap. Your getting on lots of peoples nerves. flaming bitch whore
I don't see the validity of your argument. Are you saying that because you bought the hardware, and because AMD offers free drivers for Windows users, they should be required to provide drivers for any operating system you choose to run, simply because you bought the hardware?
Nobody cares unless somebody drops an F-Bomb and gives the middle finger. Until then we should all just move along.
Well, drivers advertised as available at the time you purchased the hardware should be available and supported well enough. Drivers not advertised, on the other hand... OEMs can't support each and every OS, kernel version, ... especially when the market share is marginal, and revenue almost nil.
I understand that sucks and, frankly, it's the main thing that' keeping me away, again and again, from Linux. But I also understand that companies are not charities and have to make a business case for investing $$$ in dev and support. Especially when, as is probably the case here, there's 3rd party IP in the mix, which would cost a lot to buy out and "open", or replicate w/o getting embroiled in endless lawsuits.
The Cloud - because you don't care if your apps and data are up in the air.
He was replying to the AC that was apparently saying Linux shouldn't be supported because those users don't pay for software (generally). The AC which you replied to was simply pointing out the original AC's argument was faulty because Windows users don't pay for the software in question either. However, both Windows and Linux users paid the same price for the hardware, the price of which includes support for the drivers. If they offer Linux drivers, then it's only fair to expect the same level of support offered for other platforms.
Nice rant. Any relevance to the topic at hand ?
The Cloud - because you don't care if your apps and data are up in the air.
That's exactly what he is saying. I think you are too hung up on the word free. It's not as if a better paid option is available. All the drivers for each platform are free. As a user who purchased hardware at the full price I fully expect the hardware creator to release fully functional drivers for their device, and to not half ass it. This shock and awe that Linux should be treated differently or somehow less free than the windows version is the ridiculous argument here.
Yes, Intel provides excellent Linux support. However it comes at a slight price premium which most Linux users can't afford. They would rather mailorder "budget" AMD systems and fight with buggy drivers.
We liked you better when you were posting as an A.C. Obviously the same person.
But 100,000 Windows users have paid $1 of their final purchase price to software development. 10,000 Linux users have paid $1 of their final purchase price to software development. (It's hyperbole for an example.) Each user has paid the same amount for hardware but the "Develop Windows Drivers" pool has $90k more in it.
So much for all those nerds praising amd in the nvidia stories..the facts are these companies
don't want to give away their sauce and I don't blame them
I usually don't pay too close of attention to ATI vs Nvidia war, but I had built out a slick HTPC machine to run xbmc on Linux, and videos had all sorts of problems on the ATI card.. especially with decent quality videos. Hitching, crashing, general instability despite trying different drivers and config combinations.
Threw in a fanless nvidia, VDPAU works fine, totally different experience.
So, I'll stick with Nvidia on Linux for anything more serious than web browsing; their closed source binary driver is a little obnoxious, but at least it works.
What kind of piss-poor cpu can't decode mpeg2 in several times realtime?
The article implies h.264 acceleration for levels less-than-or-equal-to 4.1 works fine as well. Scene rules for x264 releases say respect 4.1, and most hardware players top out at that as well... so who's clamoring for it, and why?
Yeah, only losers have choices.
Yes; seems like I'm a bloody "lucky" winner. I bought a reasonably top end AMD card specifically because they promised open source support. Of course it turns out that only the proprietary driver works properly. Fine "support is coming; they do the right thing and give over the documentation; install it for now and to free later; I don't mind". Except that because it's stupid proprietary code it doesn't get automatically distributed by my distro vendor (today that's Ubuntu; who knows tomorrow). Every time I get an X-org update it breaks.
I really don't care about the high speed graphics most of the time. The free driver will be fine. Just make sure they have the specs so that the colours can be made to come out right on decent monitors and I will buy your stuff. AMD; you almost have our goodwill; You've already made the investment; Just go that last few inches; get it finished and make sure you fully cooperate with the developers. We will pay extra for your stuff. We will be glad to never see NVIDIA again. You will get better integration to Android. This will be worth it.
=~ s,(.*),<sarcasm>$1</sarcasm>,g if any_point_you_wish();
Yeah sure, best OS would be microsoft. No possible virus could infect that system. No problems, no patch needed. It's the perfect system. Perfect company... stfu
Not to be too much of a shill, but this is one area Intel seems to always be better. :-) *
Their Gfx performance may not be up to the other two, but their support is better.
Maybe Intel should takeover nVidia
-nB
* when pigs fly I assume
whois gawk date unzip strip find touch finger mount join nice man top fsck grep eject more yes exit umount sleep dump
Think different, just like every other consumer with ample disposable income.
Isn't OS X built on Linux? So your criticism is really an issue of window manager, right? MacOS v. Gnome / KDE / Unity?
Fine. Don't support Linux. If you say you support Linux, then REALLY support it. There shouldn't be a middle ground in this issue. It's pretty simple.
isn't going to happen when you keep calling them assholes because they won't support all 6 of you.
If you can't understand why Linux support is going to continue to suck, you really don't need to be posting on the Internet. You like a basic understanding of the way the world works.
I'll give you one basic hint though, as I said, when you keep rallying behind people in your community that throw temper tantrums when the rest of the world doesn't see it the same way and they can't understand why ... you pretty much lose any chance of any intelligent company supporting you. The only people who are going to support you are those that can't make it main stream for various reasons and they need a niche market to help them survive ... this companies also tend to produce shit products, hence why they are relegated to niche markets for standard products.
Keep calling people assholes and then wondering why you have no friends, go ahead, we'll continue to shake our heads in disappointment that you still haven't grown up.
Persistent Volume manager for Kubernetes - https://github.com/dwimsey/openshift-pvmanager
I'm currently using a mac mini with the Intel i910 driver and a broadcom crystalhd mini-pci-E card. 1920x1080, both CPUs run at about 30% decoding 1080p. Works very well for me.
-- Its survival of the fittest...and we got the fucking guns!!!
Intel cooperates with the community. That doesn't mean that their kit is better or that the associated drivers are better. It also doesn't make them a premium option of any kind.
That cooperation also hasn't led to feature or support parity with the Nvidia blob.
Intel is the same sort of force bundled cheap stuff that AMD is.
A Pirate and a Puritan look the same on a balance sheet.
If you've ever lurked Doom9, you'll realize that AMD/ATi cards are very narrow in the types of AVC files they will accelerate, when compared to nVidia. And this is on the well supported and well funded Windows side.
They might have some support in the drivers, but Linux video acceleration is a clusterfuck, some really convoluted setup that makes PulseAudio/ALSA look like a sane design.
In Windows, video players will simply drop to software decoding if it can hardware decode, but in Linux, the video players try to hardware decode anyways, resulting in garbled video.
My nicest card is AMD, but my ancient Pentium 4 with a NV 8400 can DxVA more files... and it needs to.
Their video acceleration hardware has DRM built into it. The reason they can't release the specs is most likely because their lawyers said not to, for fear of breaking some DRM-related legal contract(s).
we are talking about video drivers here and the drivers only work with the hardware from that company mentioned. Exactly what software should the OSS people purchase from AMD to get this support you think you know about?
figures, an AC...
LoB
"Anyone who stands out in the middle of a road looks like roadkill to me." --Linus
I bet you came when you finished typing that.
The video codecs are the least of my problems with linux support from both NVidia and AMD. Neither of them off any kind of support for switchable graphics under linux. I have laptops with modern graphics cards from each of these guys, and in both cases it has been a long up hill battle getting the graphics cards to work correctly.
If you don't need gaming graphics Intel is the place to be in terms of linux support. I know what my next purchase is going to be. I just wish Intel would expand their market and try to compete on the high end. I would love to see chipzilla enter this fight with thier opensource record.
120 char limit :(
whois gawk date unzip strip find touch finger mount join nice man top fsck grep eject more yes exit umount sleep dump
I use XBMC and find it pretty solid. There are things I'd like, mostly not GUI related
-ATSC embedded closed caption support (this feature would make me drop mythfrontend in a *second*)
-No 'headless' xbmc. For central library, an xbmc instance must be running. It does pretty much everything needed to do api calls to do database maintenance, but has to have a display
-A more simplified cookie cutter setup for centralized database. I know a lot of people will say 'PMS', but I've found that less useful (and more burdensome) than XBMC database.
I found the UI pretty decent actually. Now 'x' versus 'Tab' versus 'Escape is a little non-obvious at first in terms of 'how the hell do I get back to the video that didn't stop when I hit escape' or 'why didn't my video stop when I hit escape', but on a remote, things actually can be intuitively mapped.
XML is like violence. If it doesn't solve the problem, use more.
They tried (larabee IIRC) and failed. Intel == low-end graphics, that is just the way it is.
I wish it were different, but such is the state of affairs. (They are getting better, but really only maintaining the gab, not closing it).
-nB
whois gawk date unzip strip find touch finger mount join nice man top fsck grep eject more yes exit umount sleep dump
None of this would have happened if we had stuck to the original XBOX hardware!! :D
This statement gets my complete support. Been using Intel i915 for XBMC for at least 3 years and it's been a rock solid experience. I had some initial trouble when the kernel mode setting was introduced (before you needed an app to write to the video bios so you could get the 1920x1080 resolution) but extremely minor and long past, I have to agree, I'll be looking for Intel net time I put together a HTPC.
-- Its survival of the fittest...and we got the fucking guns!!!
in the early 90's, I bought and loved my ATI VGAWonder card and Mach32 card. They were the best, and I like to support home-grown tech (I'm Canadian). Things worked fine when I moved to Linux (pre-1.0 kernel, no less!).
But, it's all downhill from there... the lowlight was buying a top-line Radeon AGP-Pro X, which quite simply ~never~ worked, under either Linux or Windows.
They started great, made some great products, won the hearts of many. Then, they sputtered, blew off customer-complaints, and spiralled down... ...almost exactly like RIM, just a decade earlier.
Would you care to explain how AMD/ATI's revenue is different because I choose to use Linux instead of windows?
I still paid the same amount of money for the card.
Furthermore, we all know for a fact (because it's happened for every other piece of hardware) that if they released the details needed for the Linux community to write it's own drivers, they'd never have to write another one for Linux, ever, AND they would benefit from being able to take the concepts and optimizations created by the Linux community and fold them into their windows drivers.
I'm an Apple-hating Linux fanboy, but you're really really wrong.
If OS X used Linux code, Apple would have to release the source (GPL, remember?).
There is a lot of free software in OS X, but not copyleft. Check out Darwin.
If it says it supports that OS on the outside of the box then YES, I DO expect it to be included.
For me, out of nVidia and ATI proprietary drivers under linux, ATI's were always more problematic, especially under 64 bit Ubuntu.
Every couple years I try to get an ATI card and make linux work with it and every time I get very frustrated (last one was 5830 card in 2010). Yes, I don't have much expertise in kernel hacking, but I don't expect that I have to do that much tweaking to install a freaking driver. So far nVidia never presented a problem under Ubuntu. It might re-compile itself, but that's as bad as it gets. The amount of trial and error for ATI is waaay worse (how on earth can "official" binary driver complain about unsupported hardware? open driver being so slow you can see windows painting? yuk)
So, for me, out of two evils, the lesser one is nVidia. Maybe it's almost time to try a new ATI card yet again :) Though if even mpeg2 acceleration is not working, probably I should wait another couple years.
Hyperom.com
I stopped reading at the second reference to big brother.
You do realize that this is actually going to make it /easier/ for the government to install boot-loader level software on your rig, right? All any entity will need to do is pull the appropriate strings at M$, and they get a free digitally-signed pass past all of your scanners and other protection.
What leeches? The drivers don't cost the user anything extra (far as I know?). If I've already paid for the hardware, I expect drivers that work and support all the functionality, and there is no valid excuse for any hardware manufacturer to withhold them.
They do work on the hardware listed on the box when you purchased it.
People only have finite funds available, if you have a budget of $1000 for a computer and you decide to spend $400 of it on software then you only have a $600 computer.
If you decide to use free software, then you can buy a $1000 computer which will be more powerful.
AMD sell hardware...
http://spamdecoy.net - free throwaway anonymous email - avoid spam!
H.264 patent: The last expiration is US 7826532 on 29 nov 2027
MPEG-2 patent: The last expiration is US 7334248 in 2026 (but if 6181712 is held to be prior art, move that up to 2018)
Otherwise there are per unit royalties without a Microsoft to pre-pay them for you so the OS itself pays no royalties for the driver (or you pay for the driver, or you drive the chip cost up relative to Intel, for whom Microsoft also pays the royalty).
-- Terry
Would you care to explain how AMD/ATI's revenue is different because I choose to use Linux instead of windows? I still paid the same amount of money for the card.
Furthermore, we all know for a fact (because it's happened for every other piece of hardware) that if they released the details needed for the Linux community to write it's own drivers, they'd never have to write another one for Linux, ever, AND they would benefit from being able to take the concepts and optimizations created by the Linux community and fold them into their windows drivers.
I'll bite, although I generally agree with you and will make a different case in a minute.
Their given revenue may currently be based off of a model where they spend X number of person hours on developing drivers for given hardware and devoting those hours to the less common user base of Linux in the degree to which it would theoretically be necessary would take more than the given hours.
On the second point, if they were to release all of their hardware specs, then, it's theoretically possible for someone with a bigger budget (Intel comes to mind) to come in, build these exact same chipsets in higher quantities and undercut AMD at sales.
Now, while neither of these are extremely strong arguments, they're the type I'd expect to hear from corporate honchos who don't really know as much as they should. Now my big argument is that, we, the Linux crowd, is some of the most outspoken individuals in computer. I have a half dozen friends who, every time they need a new computer, they ask me to come shopping with them. You find me a company that gives true, quality support for Linux, and I'll tell you which company I'm going to recommend over the other. Many Linux geeks are in charge of (or at least have a say in) specing systems for everything from corporate servers to company desktops as well. Sure, the boost in sales won't be apparent or even clearly directly correlated, but there'll likely be some improvements...
"Don't meddle in the affairs of a patent dragon, for thou art tasty and good with ketchup." ~ohcrapitssteve
Would you care to explain how AMD/ATI's revenue is different because I choose to use Linux instead of windows? I still paid the same amount of money for the card.
Furthermore, we all know for a fact (because it's happened for every other piece of hardware) that if they released the details needed for the Linux community to write it's own drivers, they'd never have to write another one for Linux, ever, AND they would benefit from being able to take the concepts and optimizations created by the Linux community and fold them into their windows drivers.
Statistics show that people aren't going to go out and spend $500 on a nice video card to display X on Linux. The high powered video cards are for games and gaming is non-existent on Linux. So their revenue is directly affected by drivers working *great* when gaming on Windows. Successful gaming equals new cards purchased in the future. If the Linux community could optimize OpenGL to compete with Directx you might have a solid argument. But OpenGL is slow, buggy and doesn't compete so why would AMD/Nvidia have any interest in the input of the open source community on their drivers?
You cannot always go for intel - unfortunately. Try buying a high-end laptop with 17" full-HD screen and 8 i7 cores and the latest chipset - whatever it may be right now. Then ask them for intel integrated graphichs! Minds will boggle!
Impossible. You can have intel graphics on medium level laptops with "hd-ready" screens and i3/i5 processors. But not on the high end, even though intel perform better (when restricted to available, open-source drivers.)
Fortunately, a high-end amd card is not that bad either, but it eats much more power. 80 min. on battery max - oh well.
Why should they support Linux leeches?
There's a reason why all my Linux boxes except the oldest one have either Nvidia or Intel graphics. In the case of the old one I had to manually patch the ATI driver kludge source because ATI dropped support and a kernel change broke it.
So, not a penny of my IT budget has gone to ATI since 2008 because their drivers aren't very good and they don't support them.
A kernel change broke it for a card 4 years old? So the driver worked fine until Linus broke it? The biggest problem here is there is no standard for changes make by the kernel team. They have ultimate power and break anything they want, forcing companies to change. Microsoft does this also but only on major revisions of their OS. Windows XP, Vista, 7, etc. If Microsoft broke video drivers once a month when Windows updates came out there would be an uproar as well.
Nvidia drivers have always been better than ATI drivers on Linux. The only reason someone uses ATI is because it seemed like the best option for their windows partition.
Replying to my own ftw. It's funny how many people put up with this crap with Linux. When it's Windows it could be a 10,000 page thread bitching about the evil Microsoft. But when Linux does it, it's assumed as something natural and innovative because its open source. You can't have it both ways...
Has it coded, but won't turn it on? Dude, that's just retarded!
Not to be too much of a shill, but this is one area Intel seems to always be better. Their Gfx performance may not be up to the other two, but their support is better. Maybe Intel should takeover nVidia :-) *
-nB
I actually have what was supposed to be an okay Intel chip in the machine as well (built in on the motherboard). It was fine from the driver point of view but failed when it came to driving my big monitor properly and stably. Next time I probably will go with an Intel discrete solution, however.
Thanks for the comment anyway.
=~ s,(.*),<sarcasm>$1</sarcasm>,g if any_point_you_wish();
There are a lot of people buying high-end graphics cards for Linux workstations or even clusters with tens of thousands of nodes so if you don't release drivers for them you are losing a substantial chunk of the technical computing market.
distros don't matter if you put the the driver in f'ing kernal where it belongs
---Saying gnome 3 is better than windows 8 not so much a compliment as it is damning with light praise.
Window managers have nothing to do with drivers.
Video drivers on Linux (for hardware accelerated devices, anyway) have a kernel component and a driver for the X Window System (which isn't a window manager, it's the graphical system). Things like GNOME, KDE, etc. run on top of X and don't care what the underlying driver is, although of course certain features won't work well if you don't have accelerated hardware coupled with a driver that supports it.
OSX doesn't use the X Window System for its main GUI. It also isn't based on Linux (it's a highly customized BSD derivative). OSX drivers would be as useless as Windows drivers on a Linux system.
Those who can't do, teach. Those who can't teach either, do tech support.
A person can use nVidia hardware and assuming the proprietary driver exists and works for their kernel, VDPAU works well.
A person can use Intel hardware and the free driver has VA-API, which most stuff is able to use now.
A person can use AMD hardware and they're screwed. The free driver lacks the most basic of features (unlike Intel's relatively featureful one) and the proprietary driver isn't as reliable as Nvidia's. AMD has their niche and I can see why someone would buy their hardware, but for HTPCs it's pretty much the worst choice you can make, and whichever competitor you go to instead, you'll end up happier.
Statistics show
[citation needed]
There used to be a time cards had vesa support. I'd love to see vesa accelerated and with 3d functions added. It doesn't have to support everything, it doesn't have to be fast but it should enable cross platform support in the most easy way possible. I don't even have xv support on bsd for ati. I can't run 1920x1080 desktop in plan9. What's wrong with supporting a standard instead of all this mess. I really don't get it. Same for all the talk about rewriting old software with less functionality. Please someone explain to me why.
Writing drivers is not free. It costs money to hire and pay a development team.
AMD/nVidia would love it if everyone used one version of Windows (even Windows has driver compatibility problems between versions - Vista/7 drivers don't always work on XP/2K, and never on 95/98/ME, and so on). Not to mention 32-bit/64-bit drivers. They would cut their driver development costs dramatically.
On the other hand, users would love it if they supported every single revision of even the most niche OS, from Mac OS 9 to Minix to freaking CP/M.
Obviously, neither extreme is actually viable. At the very least, they'll need to support four Windows drivers (32/64-bit XP and Vista/7, probably going to become Vista/7 and 8 soon), and OS X, if only because between those two you get > 95% of the desktop market (and Apple's willing to subsidize your dev costs because they need it to sell their hardware).
Linux support is iffy because it's just on the border of "is it worth our money to make drivers for this OS?". It's big enough that you'll get *some* sales, but not enough to justify the kind of development work that goes into the Windows drivers. So it seems most companies are content to half-ass it to hedge their bets - if the Year of Linux on the Desktop finally arrives, they can quickly ramp up development and ship out awesome, Grade-A drivers, but if Linux on the desktop totally dies (at least to the level of *BSD), they didn't waste much money.
That also explains, to an extent, why Intel's Linux drivers rock. Intel graphics are found much more often on servers and ultra-portable laptops than AMD/nVidia's graphics. And, coincidentally, Linux dominates the server market, and is making inroads on netbooks and such. So Intel gets much more out of having good Linux drivers (it also helps that to Intel, GPUs are a product bundled with their actual money-maker, while to nVidia GPUs *are* their money-maker).
And yet, I bought a quite nice graphics card because I wanted 6 outputs. For a multiple monitor rig, 3840x2400 and 2048x1536, each requiring two DVI connections.
The problem is that in many cases, the hardware might be designed to support a
particular feature either directly or indirectly through register program
a particular combination of features but thatcombination is already covered by a patent currently owned by patent trol.
So long as the programming capabilities aren't known
publicly, the patent trolls don't have anything to chew on. Just
a single keyword is enough to start a lawsuit.
We don't need them to spend money developing Linux drivers; we only need them to stop throwing roadblocks in the way so that others can write and maintain them. That's the way Linux has always worked. The only cost to the manufacturers is proper documentation, which must already exist internally (or so one would hope).
I assume that's because Intel uses their video chipsets to augment their processor market (basically creating a low end market). AMD and nVidia both have a large portion of their revenue tied up in video cards, so they can't risk it even though they probably should.
OEMs can't support each and every OS, kernel version, ... especially when the market share is marginal, and revenue almost nil.
Well, it's been argued that the whole Linux ecosystem can't possibly exist for just that reason, much like flight for bees being aerodynamically impossible. And yet, there they are. Drives you nuts every time you update your OS and software, doesn't it?
A kernel change broke it for a card 4 years old? So the driver worked fine until Linus broke it?
If I remember correctly, the ATI driver was calling a kernel function which they're not supposed to access, and a security fix changed the parameters or visibility for that function so the driver no longer worked.
It wouldn't have mattered, except ATI decided to drop support for the chip so there was no official route to update the driver; I believe that was about two years after I bought the machine.
Actually, this is why you can't have source code to the binary blobs.
-- Terry
Stetson-Harrison, 2012
You can't have it both ways...
As a matter of fact you can. If the source is open, then anyone can propose the fix (to either the driver or the kernel) and even do it themselves. If it's proprietary, then you're shit-out-of-luck unless the vendor sees it in their business interest to make the fix and distribute it. That's what this entire thread illustrates: the video drivers and hardware specs are closed, and so all we can do is plead and whine.
So yes, complaining about a problem in Linux is just the first step along the way to getting it addressed (if there's sharp disagreement about how best to do it, then you'll get your 10,000 page thread!). Believe me, if the capabilities of the hardware that you and I purchase weren't treated like state secrets, we wouldn't need to be having this discussion.
A kernel change broke it for a card 4 years old? So the driver worked fine until Linus broke it?
You know, while you have a point, you should take a look at the hardware support list for the nVidia driver, and then compare it to the list for the AMD driver, and it will blow your fucking mind. nVidia manages to maintain support for very old cards, and all in one driver, why can't AMD?
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
I need HD video processing. and Intel sucks at at. Nvidia owns the market at making a video chipset that will render any file format HD without any processor load.
I really wished that intel would get off their arses and make their GFX chipsets not suck.
Do not look at laser with remaining good eye.
If you ever read the LLVM/Clang Dev Lists you'd know they are releasing the stack for their Linux Community Drivers with OpenCL 1.x full support. They are cleaning up the code and the dump will soon begin.
So why don't you write a business plan based on that data, and send it to AMD so that they know about all the profit they're missing?
Simple, the number of people wanting Linux support is less. So AMD does not make revenue (in the future) by supporting and serving Linux customers well.
Should try arch, they have even written up a wiki article on how to automatically reinstall fglrx when needed.
-- no sig today
IMO Intele never wanted to actually enter the comercial high end 3D world. They do quite a lot of resears but as I see it their interests always reside in creating the know how to scale CPUs horizontally in massive dimensions (10k cores etc.) because they correctly see that this is going to happen. IMHO the graphics series are just a way to monetize their RnD a bit earlier instead of just dumping billions on eggheads for a payday somewhere in the next decade.
Still I'm in full support of what they actually do. The only HW i had hassle free installs on on the graphics side were Intel chips.
-- no sig today
On the second point, if they were to release all of their hardware specs, then, it's theoretically possible for someone with a bigger budget (Intel comes to mind) to come in, build these exact same chipsets in higher quantities and undercut AMD at sales.
Argumento ad merda bovis.
There is no magical interface2blueprint application that I know of. Details of the interface(software or hardware) could give an expert some faint hints about the underlying hardware design breakthroughs. Maybe. Why would anyone bother to do that when he could just browse patent applications?
AMD already releases specs for the graphics stack, so this video encoder secrecy is probably 100% media cartel enforced.
It's not shilling; I point this out from time to time myself. If you want the best-supported GPU on Linux, it's simple: buy Intel. There's absolutely no room to argue this either; it's a plain fact. With the other two, you get either a proprietary driver that may perform well, but doesn't integrate well with the rest of the distro as you've found out the hard way (breaks every time you apply security updates to X or kernel, doesn't support KMS, etc.), or you get an open-source driver that does integrate nicely and support KMS, but the performance is total crap. With Intel, you get excellent performance (relative to the GPU's potential), and excellent integration, because it's all open-sourced by Intel.
Of course, the downside is that Intel GPUs have terrible performance compared to the other two (because of the hardware itself).
However, there's some possible caveats here: the Intel GPUs are dirt-cheap (integrated into motherboard chipset), they don't use much power (the other two are usually power hogs), and the performance of a mid-range Nvidia GPU running the open-source Nouveau driver is probably equivalent to a recent Intel GPU running its open-source driver (because the Nouveau driver gets such terrible performance), while the Intel will get much better power efficiency getting that lackluster performance. Finally, the Intel GPU is probably perfectly adequate for the needs of most desktop/laptop users who are only doing things like watching videos, and using 3D effects on their desktop environment; obviously, you're not going to have a good experience playing a recent FPS game with an Intel, but if that's not part of your requirements list, then you probably have little reason to bother with Nvidia or AMD.
It's not shilling to point out that a cheap, low-performance option that's extremely well-supported is probably a better choice than a more expensive option that's so poorly supported that you can't realize anything near its performance potential.
I don't think this is quite correct. I used to work at Intel, so maybe things have changed a little since I was there, but as I saw it, the main reasons they entered the 3D market was two-fold: 1) to secure their position in chipsets, and 2) to make money. When I was there (over 6 years ago), they were the world's largest GPU manufacturer. I imagine that hasn't changed. Yes, their GPUs were low-performance compared to the competition, but that wasn't all that important; their goal was to dominate chipsets, and they did then and I believe they still do now (honestly, it seems like very little has changed in the PC world in 6 years; lots has changed in mobile devices (phones, tablets), but not in PCs or laptops). Most PCs don't need high-end GPUs; most PCs are bought from places like Dell, in large quantities, and used in offices for corporate drones to read their Outlook email, write MS Word documents, etc. They only need 3D so they can run the graphical effects in Windows. Many more PCs (probably more laptops these days) are sold to individuals and corporate users, who again use them to read their email, use MS Office, and use a web browser. They only need 3D for graphical effects and to watch videos with GPU rendering. Some might play a low-end game here or there, but most don't. The people who do want to play games probably quickly find out that integrated graphics aren't very good for that, and upgrade to an Nvidia/AMD card, if they didn't do so from the outset.
By having a GPU built-in to their chipsets, they were able to get a lock on much of the chipset market. Instead of a PC buyer need to buy a motherboard w/ chipset, and then a separate graphics card, they could spend a couple bucks more, and get a motherboard with integrated graphics, and forgo the graphics card altogether, saving a bunch of money. Remember, before Intel got into 3D graphics, there were a bunch of chipset makers; these days, many of them seem to have withered away. They couldn't satisfy the low-end users by building an acceptable GPU into their chipsets, so everyone just switched to Intel.
They don't really need to; they've already succeeded at getting a giant majority of the chipset market by having an integrated GPU that sucks, but is good enough for average users who do little besides surf the web and maybe use MS Word.
Saying they need to make a GFX chipset that competes with Nvidia's and AMD's mid-to-high-end offerings is like saying KIA needs to make a car that competes with Ferrari. Not that it wouldn't be nice (since Intel's open-source support is so superior to the other guys'), but it's probably not exactly high on their priority list when they're already making buckets of money by covering the low-end market.
> There's absolutely no room to argue this either;
Sure there is. Intel peformance sucks and there are features missing from the driver. It doesn't matter how "open" it is.
It is not the "best-supported" on Linux.
A Pirate and a Puritan look the same on a balance sheet.
> Statistics show that people aren't going to go out and spend $500 on a nice video card to display X on Linux.
Most people don't in fact.
So your attempt to demonstrate that you have more money than you really do is rather pointless.
A Pirate and a Puritan look the same on a balance sheet.
I buy what's good kit for HTPC use or what's a reasonable but not absurd choice for light gaming. This isn't just a Linux thing. In truth, very few people are inclined to confuse their video card with their manhood. That's why Nvidia is hurting. They have no CPU to force bundle their GPU with.
Most people are cheap bastards. That's why Apple didn't destroy Microsoft in the 80s.
A Pirate and a Puritan look the same on a balance sheet.
Sure there is. Intel peformance sucks and there are features missing from the driver. It doesn't matter how "open" it is.
It is not the "best-supported" on Linux.
BS. I already said the performance sucks compared to the others, but that's orthogonal to "best supported". Intel is the best supported, hands-down. There is absolutely no room to argue on this; only a fool or a shill would disagree. With Intel, there's open source drivers available that get the best performance out of the hardware, and are well-integrated with the rest of the Linux distro. With the others, this simply does not exist. You can either get a proprietary driver that performs well but the integration is total crap, or you can get an open-source driver that integrates well, but the performance is horrid.
Intel's HW absolutely IS the best supported on Linux. This cannot be argued.
You want to run ancient hardware on a new system then perhaps you should run a new driver for it. If that ancient hardware is discontinued in the current driver then perhaps you should take the vendor's action in the matter to heart.
There's always the libre driver. If your card is ancient, the related performance issues probably aren't a problem.
A Pirate and a Puritan look the same on a balance sheet.
Things like hardware H264 decoding rely on tech that they do not own all the patents to. they are probably contractually barred from releasing the specs of this hardware to the OSS community. They HAVE released quite a lot of specs on the physical registers. The fact that their OSS drivers lag nVidia Nouveau OSS drivers has more to do with market share and developer mind share than not releasing the relevant specs. OSS developers will gladly reverse engineer nVidia hardware to develop the drivers, yet balk at working on an AMD driver with the specs known beforehand.
I can't add anything regarding ATI/AMD Catalyst drivers for Linux issues but I thought I'd mention results with nVidia so that someone can compare that to their experience with HTPC and ATI/AMD drivers.
Over a year ago I built an HTPC box based on Asus AT5ION-I motherboard with Intel Atom D525 (1.8 GHz 2-core + HyperThreading) and nVidia ION Next-Gen (~GT240 equivalent) video on board, 2GB of RAM, Intel X25-V 40GB SSD, and WD 1.5GB HDD. For software I used the old XBMCLive Ubuntu 9.x Linux distro that is now upgraded to Ubuntu 12.04 running XBMC. This system only consumes 50 Watts of power when running and decodes everything you can throw at it, even 8MB/s H.264 1080p with DTS or AC-3 Audio while only using 15-20% of CPU cycles on each logical core and 450MB of ram. This is all due to the VDPAU video acceleration available from the nVidia Linux drivers, currently 295.59 and always updating when new ones come up or new kernel is released since a re-compile is always required with this binary driver.
The biggest pain the arse is the recompiling of the nVidia driver every time that there is a Kernel update but the nVidia drivers work well enough with the command below that forces the re-download of a new driver, recompile, and reconfiguration. Then just killing Xorg process and starting uxlaunch service gets me back into XBMC frontend. This driver might be binary only and it might "taint" the kernel with its license but frankly it works well, every time, and never crashes. It provides video acceleration functions and decoding without any artifacts or issues.
sudo ./nvidia-installer --accept-license --force-update --update --no-questions --run-nvidia-xconfig
After a year of working using this I have been super satisfied since it plays everything I throw at it, old XVid, AVI, MPEG-2, H.264, etc content without any issues and with full audio with MP3, AAC, AC3, DTS, etc.
I would be weary of going with an AMD/ATI Fusion based motherboard after having such a great experience with Intel Atom and nVidia ION. I don't wish to harp on AMD since I loved the company years ago during the Athlon days but the sour taste that I got from the Opteron Dual-Core issues with core timings going out of sync and other game strangeness I am weary of that company. Although I did buy an ATI 6950 unlocked to 6970 for $225 USD that I have been happy with in Windows 7 playing last year's games but I've been religiously downloading their monthly ATI Catalyst drivers, updated to 12.4 just today.
And they would lose control.
You OSS guys are all hardon for keeping control... Yet you want AMD to give up control just so you can have a driver.
That sounds like a shitty deal. And you know it's a shitty deal. Why do the linux junkies always want someone ELSE to take the shitty end of the stick?
Every time.. Every damm time...
I chose an HTPC with AMD processor and graphics over the one with Intel/nVIDIA, thinking it would have better Linux support, and an nVIDIA based Android tablet thinking it will get good OEM and driver support. Turns out that now I am stuck with both.
Chat with other atheists http://secularchat.org
However, if I have both windows and linux, and would like to be able to use one card on both, guess which company's 500$ card I WON'T be buying?
Great Intellect...
On the second point, if they were to release all of their hardware specs, then, it's theoretically possible for someone with a bigger budget (Intel comes to mind) to come in, build these exact same chipsets in higher quantities and undercut AMD at sales.
That is ridiculous. The specifications needed to interface with a hardware device are roughly comparable to a complete definition of a software API. The implementation details are typically thousands of times more complicated.
For example, if you had the complete interface specification for a modern desktop microprocessor (freely available, easily a couple of thousand pages), just add several hundred million dollar research and development budget, and you too can produce a compatible device with comparable performance, if you are lucky.
Remember, before Intel got into 3D graphics, there were a bunch of chipset makers; these days, many of them seem to have withered away. They couldn't satisfy the low-end users by building an acceptable GPU into their chipsets, so everyone just switched to Intel.
I seem to remember Intel making it difficult for Nvidia to create chipsets for newer CPUs.
That could be too; if the goal is to dominate chipsets, then both actions are logical (making it hard for the competition, and offering features that render the competition's product unnecessary for as many users as possible).
Actually, I'd say that's only partially true. I mean, the Intel Sandybridge graphics driver works great for 2D hardware support, desktop compositing and 2D OpenGL-based games. But go 3D and the driver suddenly starts coming short (f.e. a fair amount of recent Humble Indie Bundle games just won't work properly with it).
I am not devoid of humor.
You're right; if you don't need gaming graphics, Intel is pretty much perfect in their support.
However, once you start trying to play games with an Intel GPU you'll find that there's plenty of games that don't even want to start up, or have issues that make it unplayable (Amnesia, Psychonauts, Bastion to name a few). Note that these are all issues I'm having personally with a Sandybridge, though.
It's a pity, really. Intel's graphic driver performance is surprisingly good (esp. when compared to their Windows-based drivers). They could have gone the extra mile and make sure more games launch at all.
I am not devoid of humor.
Statistics are also colored by the current state of affairs. What they actually show is that people won't shell out $500 for a video card to use in X when, due to driver brokenness, they will only get the features of a $100 card.
If they actually got the full $500 worth of features, the buying decision might change, but statistics can't tell us anything about that.
AMD and Linux: It is much more likely that MS will go with Intel for their new systems. MS has stated that they intend to make their own machines (a la Apple).
Where will that leave AMD? I guess that they may not even have a chance to have their software or hardware as a replacement for the Intel products.
If the above is true, then AMD should be at least considering to document the specs for Linux. The fear of patent lawsuits is a fear that all software vendors have today, AMD or Nvidia or Intel or another software vendor has to worry about the trolls. That fear is most probably the reason there is not better help from video card vendors. The algorithms to get performance from the GPUs is probably well known. I am thinking that the logic differences within the AMD and Nvidia cards is trivial.
Leslie Satenstein Montreal Quebec Canada
I'm putting you on my list of Trolls that support the Right Wing movement. I blame your parents for your lack of morals.
and there is no valid excuse for any hardware manufacturer to withhold them.
They aren't withholding them, AFAIK the features are in the proprietary drivers AMD produced for Linux.