Slashdot Mirror
Open-Source GPU Drivers Show Less Than Ideal Experience For SteamOS/Linux Gaming (phoronix.com)
An anonymous reader writes: Phoronix's recent 22-Way SteamOS Graphics Card Comparison showed that NVIDIA wins across the board when it comes to closed-source OpenGL driver performance. However, when it comes to the open-source driver performance for Steam Linux gaming, no one is really the winner. A new article, "Are The Open-Source Graphics Drivers Good Enough For Steam Linux Gaming?" answers that question with "heck no" by its author. While AMD is generally regarded as having better open-source support, their newer graphics cards still can't run at their rated clock frequencies due to lack of power management support, the lack of enough OpenGL 4.x support means many AAA Linux games simply cannot run yet, not enough QA means regressions are common, and other issues were noted when it comes to testing a number of modern graphics cards on the open-source drivers.
Playing the eternal game of catch-up.
So what?
The closed-source drivers run on Linux and they run fine. I don't see what's the problem with switching to them while gaming - the games aren't going to be Open Source either, for the most part.
Things that were true for the past 20 years are still true.
What does licensing have anything to do with performance? If NVIDIA released an open source driver for their hardware, then open source drivers would win the performance shoot out. Their drivers will always be better than reverse engineered drivers (open source or not). The question really is, if you care about licenses and gaming performance, what hardware will fit the "good enough/open enough" requirement?
There are ways to help fix these types of problems, but users need to put there money where there mouth is and buy hardware from the companies working on projects to resolve these sorts of issues- and preferably not the frauds (there are *many* fraudsters taking advantage of the community claiming 'open source' and 'freedom' who are doing absolutely nothing to actually fix the problems they're claiming to be 'working on').
The Free Software Foundation maintains a list of companies that prioritize freedom and make sure the community is in a position to fix bugs and improve the overall quality of the code (unfortunately while it makes it easier even of the companies on this list only two out of four are really making a difference.. unfortunately the criteria doesn't require you to make a difference- only to sell free software friendly hardware which has enabled people to swipe in and take advantage of *other* companies efforts- which might be fine if they were actually contributing back- but they aren't- it's one thing to build off the works of others when you contribute back- but when you don't your undermining the part of the movement that is trying to fix the problems).
Sadly a lot of people just don't understand that by buying random 'cheap' hardware it makes these problems unfixable too. Working in the industry the segment of the population having the biggest impact in contributing (financially generally) to solving these types of problems are 1. free software advocates 2. less technical users adopting freedom friendly hardware from companies who are contributing back, but they're doing it because it 'just works' (mostly two is resulting in the funding that is needed to work on these projects).
I'm disappointed, although not really surprised.
I read stuff like this, and, you know, sometimes it seems like we haven't moved an inch since 1991. Sometimes it seems like we are absolutely NOWHERE.
Just sayin.
LOL@vword: "fooled". You can say that again, brother.
I already have 95% of my apps and formats cross platform and open.
All I am concerned about now is hardware support.
Then once that is fixed, I switch.
I am hoping SteamOS and Steam is the catalyst for the hardware driver momentum increase. I am running it in VirtualBox and it is blisteringly fast even on that, I can't wait to see it running native on my 5 year old netbook :)
Counter-Strike Source and Left4Dead still rock the house.
Useless review without mentioning Intel GPUs.
not enough QA means regressions are common
Regression testing is borrrrrrrrring! I wanna do the fun happy path stuff!
(-1: Post disagrees with my already-settled worldview) is not a valid mod option.
So the system that requires buying, reverse engineering and basically guessing what they can do, all under threat of hideously huge lawsuit if they even LOOK like they're doing something the "owners" of the IP don't like, lags behind the output of the people who get to start their work BEFORE THE DAMN THING IS ON THE MARKET.
Who WOULD have guessed.
What I don't understand is why AMD bothers to keep Catalyst around, when a) they've already shown to be supportive towards open source, b) Catalyst drivers have always been considered crap compared vs. their Nvidia counterpart (by most gamers anyway), and c) the open drivers have made leaps & bounds of progress in the last few years.
Supporting both the open and closed source drivers will surely take more resources than focussing the effort on one of them. And I kind of doubt that AMD has much resources to spare for this kind of thing. The open drivers have caught up to Catalyst quite a bit lately, why not work towards replacing Catalyst with it? That would make everybody on AMD side of the fence happy I think (not to mention future buyers of AMD videocards / APU's).
With nVidia alone maybe the argument is that they won't release open source drivers just because they are dicks. Ok, yet AMD has thrown in with open source, however they still have a faster proprietary driver. Why? Perhaps the licensed code has something to do with that. They license various things they can't open source, and those things are some of what help make it faster. AMD didn't just release the Catalyst source because they can't. It contains code they paid licenses for and can't open up.
The situation isn't quite as simple as some people would like to make it. If it was, well then the OSS AMD drivers would have full support, be faster than Windows, perfectly stable, etc, etc. However it turns out that writing a graphics driver is REALLY HARD and very complex.
Anyone that would use Steam would have no issues using closed-source drivers.
It's good for servers, and even a lot of office desktop applications, it just lags far behind in gaming because market-share is small, and some extremists go nuts if the video driver isn't open source.
Care to explain how? If you mean by hacking a driver such that it produces more fps, then (by that logic) simply plugging in a faster GPU would qualify as cheating too. Note that the 'faster-GPU-cheat' is considered perfectly acceptable for online gaming, only exception being pro gaming tournaments where I'd expect all participants to have same-configured machines.
In case you were thinking about see-through-wall hacks, mods that help with aiming etc: those things are in game engines not graphics drivers.
They elected a new speaker and passed a budget this week. Or didn't CNBC mention that?
No, CNBC was too busy trying to sabotage the latest presidential debate...
Is there anything stopping me from just using the drivers that actually work? When the OSS drivers catch up we can revisit them - until then I'll use what works.
BeauHD. Worst editor since kdawson.
Preamble:
Just started using linux Mint - with no internet connection only one CUBE runs well on http://itsfoss.com/cube-lets-i....
Long story short Charter.com hacked me, keylogger (firewall caught that one) and much more
all because they thought I owned them $100 (I had auto payments setup). Cell phone number was reused and flagged.
Turned off the system and now in the process of using linux chntpw capturing what they did, and will send it to the appropriate people.
My windows era is over.
---
Haven't played any games with Linux yet, but have every intention of going to the Steam or SteamOS and playing what they have available.
I have an EVGA NVIDIA GeForce GTX 750, reading the article I'm going to be doing rather well. While it may not of come out on top other than Performance Per Dollar, it can handle all that was tested; each at least above 60 FPS lowest 90, and all at very low temps.
Sadly a lot of people just don't understand that by buying random 'cheap' hardware it makes these problems unfixable too.
You are under the assumption that "a lot of people" give a damm.
They don't, they just want something that works and is cheap, the rest is just background noise.
The Free Code purists in the Linux community are so dedicated to their cause that they intentionally make it hard for the average user to use the proprietary drivers.
They place hooks in the OS that may only be used by open-source code - yeah, they check the code to see if it's open source and if not this extra junk code added to the kernel refuses to cooperate..... so much for "freedom" for the end-user.
They do not place easy-to-use buttons on the install screens to let the user select to install a manufacturer's "binary blob" or run the manufacturer's install shell script
They auto-install the not-ready-for-prime-time open-source drivers, tying them in with the grub stuff that most users find confusing and scary (most users are too terrified to do what it takes to root-out drivers like Nouveau, and when they try and those drivers keep arising like walkers that haven't been stuck in the head, they give up and run away)
There is simply no valid reason other than ideological fanaticism that keeps the situation like this; it seems a political effort to make the experience so bad that users will give up and settle for the poor performing open source drivers. This is both sad and counter-productive. When Linux started to take-off, there were open source drivers for standard 2D VGA chipsets because those chips were fully documented in the freely-available data sheets. Then, as 3D chipsets arose and actually replaced the 2D chips, there was a lot of consternation about whether the video chipset makers would cooperate or Linux would by strangled by having no video card support. Nvidia came along and provided drivers for Linux, an act that cost them money and resources and which they never HAD to do and they've been great in providing constantly updated and excellent drivers for Linux at a release rate similar to their Windows drivers. It's an act of self-destructive insanity for Linux people to keep fighting this level of cooperation and support because it's not "pure enough" for some. If it's pure enough for the users, who all this "freedom" was supposedly about, then it ought to be pure enough for the purists to at least fully cooperate with.
...because I'm going to play proprietary games on my system.
It's the same BS line NVidia gave. They made the mistake early on by being specific (ish) and saying it was something from SGI, who, when approached, said that NOTHING they gave NVidia was something SGI were worried about not being disclosed, and they were free to open source it.
NVidia shut up about what this "IP" was after that.
Didn't stop them *claiming* it was there, though.
And now you're doing it for AMD.
You know what else is dead quiet and can handle almost any kind of video? Plex and a $30 android 'stick'.
It cannot handle a webserver, SSH, torrents, print server and NFS file server as the ION board does without breaking a sweat though.
I fail to see how any of those need a GPU. You could try using "Plex and a $30 android 'stick'" for playing video and the ION board for everything else. It's called division of labor.
If you want to run open drivers, dumping the Nvidia card seems to be the best option. Nouveau support for anything but the newest cards is basically zero, and even the newest cards can only run at half speed. I unfortunately have an older card (GT630), and there is zero support beyond VGA, and the only certain thing is that the nouveau folk have promised poor performance as they have no intentions of working on improving speed for it and similar (3-5ish year old cards). I don't know who is working on nouveau, but I don't think there have been any changes for about 18 months (at least). There was a guy who said he was working on mode setting support and higher speed/better performance, but that was about 18 months (at least) ago, and there isn't a card/driver you can point to and say "see, that's it". To be fair, Nvidia has done as little as possible to help the nouveau developers, but I think they gave up trying to improve performance a long time ago.
Wait for Vulkan. GL is a mess on Linux (it's not much better on Windows either).
I offered up my 300 baud modem.
Obsolete because the new tech (v.90) is literally 160 times better, pushing practical 48000 bits per second over the same line and providing substantial practical benefits. Old HTPC gear isn't quite as necessarily obsolete, as hardware that used to be able to push 1080p can still push 1080p.
You missed the point of why.
Then for the benefit of other readers, could you explain in more detail the point of why?
And if you're running Gaming rig, you're going to want high end current CPU, GPU lots of ram, lots of fans and ventilation, which isn't the same rig as a quiet HTPC.
Then how do Sony and Microsoft succeed at promoting PlayStation and Xbox family devices for both gaming and home theater applications?
The Linus Torvald's you know and love is but a figurehead puppet used by the open source conglomerates as clean shaven media representative. This man was born without birth certificates as part of a Middle Eastern slave harem and is commonly known as the Lunix Colonel.
The real Linus Torvald's has not left his mothers basement for 25 years. Nor has he shaved in this time. In March 1994 the Kernel was released as version 1.0.0 to celebrate Torvald's beard reaching 1 foot in length. 2 years later, largely due to a healthy diet of lutefisk they celebrated the milestone of 2 feet.
Unfortunately due to interference by corporate actors such as the Soviet conspiracy, Red Hat, the numbers became stagnated and no longer accurately reflected the true length of Torvald's beard. He was forced to trim.
This event caused Torvald's great sadness and resulted in him spiralling out of control into deep depression like parts of Mark Shuttleworth's Challenger spacecraft. He stopped showering for several years and this corresponding time period contained the greatest number of bugs in both the Linux kernel, and his beard.
The depression and lack of hygiene was contagious and spread to Open Source Wizard Richard Stallman who became known for his podiatric-auto-canibillia and was more likely to be associated with sores and sauce than source. The rival HURD kernel will never be completed as Stallman has forgotten how to program.
Torvald's mean while continued coding until his fingers bled, pushing code into his git under pseudonyms of various nerds around the world who paid the open source conglomerate to keep the sole Linux Mainframe online.
In 2011 Torvald's was able to wrestle control back over his versioning system and matched the released to the length of his beard for the 3rd time. This greatly improved the kernel and led to the development of some of the key technology of the 21st century: System D.
Seeing that his kernel was getting bigger, Torvald's began researching peer to peer Bitcoin block chains and Tor network services as a way to revolutionise the kernel for the first time since Al Gore invented the internet. System D was to use the one true linux mainframes hard drive to store pictures of Torvalds Penis, the system D version numbers were to reflect it's size at any time in some of the first research into Quantum computing versioning. After Jarrod from subway the initial angel investor due to seeing how this technology could be useful for his own interests, Google joined the project with the creation of it's D-wave computer - The first self contained and self replicating System D computer.
This caused a further rift between Stallman and Torvald's, as Linus had turned his operating system into a more advanced version of HERDs naming system. Many gnus were killed in the great battle of recursion.
In 2013 Torvald's beard had grown to a staggering 4 feet, as long as Eric S Raymonds was tall. This also marked the first time that Linux and System-D were the same thing as at the time Torvald's penis was 4 feet long.
Torvald's beard is currently approaching 4.3 feet long. He last had a shower this morning when he nearly got an erection and it is currently free from bugs.
I'm not the earlier poster, but you are probably wrong. I have personal experience with this sort of thing. Often when companies go through big changes, including things like mergers, aquisitions, shut-downs, break-ups, etc there are people "from the company" who publicly say all sorts of stuff about the legal entanglements around products/services/employees. Most of the time, these public statements are factually wrong. In most cases, very few people in the business are aware of most (and certainly not ALL) of the contracts that exist between the various business entities involved (not only the business units directly involved in the changes but also contracts in place with suppliers, vendors, and customers).
Just because Joe Shmoe who currently works in some business unit, or perhaps who used to work in some business unit, or who works in a business unit that is part of some merger or separation SAYS that something can go open-source or is not covered by a contract, that just does not make it so. SGI or some former employee could publicly and with honest intent claim there is no problem, BUT Nvidia could be sitting on a binding contract, and worse for them the contract could include a confidentiality clause that prohibits them from explaining this to the public or even refuting a false statement about "no such contract". People in the Linux community should have learned long ago via the SCO lawsuit, if by no other means, that even the worst contract can become a very expensive and years-long major legal fight even it it is with a defunct business. The guys at Nvidia would be total morons if they decided to willfully violate a binding contract "because some guy on the internet said some guy who used to work at SGI says it's fine to blow it off".
If you have not personally gone through all the contracts in Nvidia's legal department, you are in no position to snarkily pretend that they are not serious and honest about the problem.
Let me pose a basic question (admittedly asked several ways) to you that I ask of any conspiracy nut:
Why would the accused conspirator DO THIS? What's "in it" for him? Why is it more worthwhile to him to engage in the conspiracy than not to?
If there's nothing to be lost for Nvidia if they open-source the driver, and there is no legal block, then why don't they do it? They'd certainly get good PR, and extra eyes looking at the code and helping it integrate with every change in Linux would only make the code better and support easier, so WHY NOT? (and please, leave the foil hat off while answering, and come up with something that does not involve cattle mutilations and anal probes)
I hate conspiracy theories and am not engaging in one.
The Linux kernel coders have intentionally taken positive, affirmative steps (and added code that serves no other purpose than) to add code to the Kernel that makes it harder to run closed-source drivers. (Google: "tainted code")
That's an ideological and political act that makes life harder for the users (whose "freedom" this was supposedly all about) to use proprietary video drivers.
It's a second issue that the guys from the distros put in place their own code that also increases the hassle for typical users; they specifically install the Nouveau drivers by default and tie it in in ways they did not originally do so that an average user will find it difficult to root the damn thing out and replace it with a proprietary driver. This was NOT just an incidental effect of their installer code --- earlier versions of Linux did NOT do this. Additionally NOTHING prevents a distro from shipping a CD that also has an AMD, an Nvidia, and an Intel binary blob or installer on it and, at the appropriate point in the install, offering the user a button to select one of those things in place of the default. This would be best for the users, since the proprietary driver on the disc would be one that was tried when the distro was put together.
This is no problem for me, and perhaps not for you. I run my business almost entirely on Linux and need the proprietary Nvidia drivers to support CAD. Over the years I have become quite adept at putting proprietary Nvidia drivers onto Linux.... but then I have written device drivers for Linux and am comfortable building kernels to embed into systems. For a typical computer user, however, this is all a completely unnecessary and "scary" hassle that serves no other purpose that to deter people from Linux, and it's ALL being done a bit maliciously by hard-core activist type programmers grinding their personal political agendas. In engineering, best practice is to put in place exactly what's required to do the job, AND NO MORE. Adding code that only exists to make it hard to use proprietary drivers is BAD ENGINEERING, unless there IS a purpose - making it hard to use proprietary code (which I seriously doubt is a "feature" requested by many users).