I have to wonder if the Ubuntu/Redhat type of Linux is really the right target for Valve. Those tiny Android on a stick and Ouya type of systems are just going to keep increasing in power. I guess you could dual-boot Android/Ubuntu but I see Android winning against desktop Linux for most consumers.
The obvious reason is that Ubuntu has proper support for x86_64, and Nvidia graphics cards. You can't play AAA games on an ARM SOC (yet). Plus Ubuntu (w/ Nvidia drivers) supports the latest OpenGL, and the native display stack is quicker than Dalvik + Native on Android.
>I see Android winning against desktop Linux for most consumers
Erm... maybe, but not for the consumers who are in Valve's target demographic. Valve sells AAA video games, and while tiny systems might have the hardware to be able to play today's TF2 five years from now, their games will also have moved on.
Android only works on PC at all through third-party forks, and certainly doesn't have GPU support. Whatever else you think of Canonical, they've been taking point on getting graphics driver support for Linux.
Valve Time is a popular joke because of Half-Life, but if you set that aside they actually release things pretty consistently. With all the engine improvements and content updates, TF2 in 2013 is only barely comparable to TF2 in 2007 in terms of gameplay, and not at all in terms of system requirements to play at high quality. And apart from that, since the Orange Box in 2007, they've released major games in 2008 (L4D), 2009 (L4D2), 2011 (Portal 2), 2012 (CS:GO), and 2013 (DOTA 2). The only year they missed was 2010, but that was the year they released the OSX ports of all the Source games. They also rolled out Linux support this year, although I imagine the transition from OSX to Ubuntu was much less difficult than from Windows to OSX.
Actually ARM SOC faces some serious issues about increasing performance. Especially with memory on chip. None of the ARM fabs are able to produce at the size needed to get further speed increases easily. Intel on the other hand is still driving their size down.
Most of the performance increases in ARM recently have been from multiple cores. That ship has sailed. The low hanging fruit has been gathered. Anyone expecting performance increases similar to what Intel managed on the x86 over time is going to be disappointed.
Intel will probably get isn't offerings down to compete with ARM before ARM is able to boost performance to compete with Intel.
> Actually ARM SOC faces some serious issues about increasing performance. Especially with memory on chip. None of the ARM fabs are able to produce at the size needed to get further speed increases easily. Intel on the other hand is still driving their size down.
Didn't apple just announce their A7 chip (64bit, double registers, etc), supposedly based on the arm8? Rumor is that it is 20nm process with TSMC as the fab.
TSMC isn't supposed to hit 20nm wide production until later in 2014 last I heard. Haven't seen anything about Apple using those chips, but that could be the case. Expect most of the performance gain to be from increased memory instead of actual CPU.
But it isn't just a matter of raw size. The performance you are getting out of it at a given size is an issue too.
ARM may take the world of smartphones by storm, but it really hasn't made significant inroads into areas where performance matters.
Much in the same way that mobile/social gaming didn't displace "hardcore" gaming, ARM hasn't displaced Intel. It just found another audience with different priorities.
The big thing missing from that discussion was any real discussion of input devices, other than a mention that these are important. The CPU & GPU and other stuff in the base box are relatively easy to handle with an arbitrary version of Linux or Linux derivative. The interesting action is in coming up with innovative input devices that will appeal to a broad enough main seam audience that game writers will support them.
I recommend you to read again the article. Newell here makes some interesting points that directly address your concerns. Briefly:
(a) Games of tomorrow will be made to host creative communities, which almost mandate those games to be for desktops. It's just too awkward to do serious modding work in a tablet.
(b) What is really worrying about the PC is not the format being outdated -- Minecraft, LoL, TF2 and WoW attest to its potential. What's wrong is Microsoft awkwardly replicating mobile ecosystems. Creating videogames now is all about novelty, taking ideas to market fast, and growing cool communities around these games. This just can't happen, when your platform maker is Big Brother's whiny, lazy cousin.
All in all, I just don't get why the Droid Brigade is so enthused about mobile gaming. It's probably an USian localism. The cons: Android smartphones are slow and jerky and shit; the iPhone is like Manhattan if it were a North Korean gulag; fat-fingering your way around is the shittiest of input methods; and most of the games are preheated gimmicks, tastelessly exploited before the camera until boredom strikes.
TLDR: Linux doesnt require valve to hand over a percentage off of sales, so we will be all self-righteous about how we love open platforms and their evolution... Like we cared before the commission on windows 8 store was announced.
As someone who does most of his work in Linux, I would love to also be able to game in Linux. I see way too many problems for this to be solved anytime soon. Here are a few of the problems I regularly encounter when gaming in Linux.
1) X11 Fullscreen support - Current support is terrible. Basic games like Tux Racer have taken down my system. Maybe wayland or Mir will fix this.
2) Current Compiz / 3D compositing window manager kill 3D performance - The performance hit when running under any of the 3D composting mangers is unacceptable. Maybe wayland or Mir will fix this.
3) Video driver situation is abysmal. Nvidia and AMD open source drivers run at a fraction of speed of the proprietary counterparts and are riddled with bugs. Proprietary drivers only work for latest generation cards. Intel has decent open source drivers, but no true gamer would consider using integrated graphics.
4) Linux Sound subsystem is a disaster. Sure it's fine for listening to music or watching videos, but the latency and CPU usage is unacceptable to real gamers. MS learned their lesson with Vista/Win7 and now have proper hardware acceleration for sound cards in Win8 like they did in WinXP. Linux had their chance to fix this situation with OSSv4 but instead went with the inferior ALSA.
I would love to see Ubuntu and Value clean this mess up, but I'm not sure they have the resources to pull it off. If Sony had chosen Linux instead of BSD, I could see this happening. But the way things are going with Mir in terms of performance and industry support, I don't see it happening.
Real gamer here (well I play games for fun, I would say that makes me real).
I use FOSS AMD drivers, the default sound system for Ubuntu (I presume it's pulse) and the compositing gnome shell desktop environment.
I am able to play Valve's ported games quite smoothly and enjoyably, Even things like alt-tabbing out work flawlessly.
Yes, for causal gaming and general system usage, the open source drivers are usable. But, I play competitively in a clan and take gaming seriously. None of my clan members or gaming buddies would ever sacrifice 10-50% FPS and put up with the micro-stuttering with the open source drivers. I regularly install the gallium3D drivers to see how they are progressing. The current visual glitches and performance hit is unacceptable to me.
Valve and Ubuntu might have popularized Linux gaming, but they also screwed us over.
Every desktop CPU sold in the last decade, at least the ones capable of running games, supports x86-64. The only reason 32 bit installs exist in the wild is because Ubuntu (still) makes it their default download.
Enter Steam for Linux. Targeting 32-bit Ubuntu, it makes this the de facto platform for Linux games. My desktop OS is now cluttered with multiarch libraries, and the top-end games I want to play won't take advantage of either the better instruction set or all my ram.
Linux gaming could have been a fresh new start, a break from the 32-bit instruction set, but instead we're now stuck installing, maintaining and supporting 32-bit libs forever..
Lots of AAA games still ship with only 32-bit executables. If the message is supposed to be "come to Linux, it's easy, we promise," then there's a good argument for making things as similar as possible for devs/publishers who don't want to gamble on a virgin market.
I agree with you, though, it's disappointing that even in abandoning Windows we're stuck with MS-style backwards compatibility.
21 comments
[ 3.0 ms ] story [ 37.8 ms ] threadErm... maybe, but not for the consumers who are in Valve's target demographic. Valve sells AAA video games, and while tiny systems might have the hardware to be able to play today's TF2 five years from now, their games will also have moved on.
Android only works on PC at all through third-party forks, and certainly doesn't have GPU support. Whatever else you think of Canonical, they've been taking point on getting graphics driver support for Linux.
Valve Time seems to disagree.
Most of the performance increases in ARM recently have been from multiple cores. That ship has sailed. The low hanging fruit has been gathered. Anyone expecting performance increases similar to what Intel managed on the x86 over time is going to be disappointed.
Intel will probably get isn't offerings down to compete with ARM before ARM is able to boost performance to compete with Intel.
Didn't apple just announce their A7 chip (64bit, double registers, etc), supposedly based on the arm8? Rumor is that it is 20nm process with TSMC as the fab.
But it isn't just a matter of raw size. The performance you are getting out of it at a given size is an issue too.
ARM may take the world of smartphones by storm, but it really hasn't made significant inroads into areas where performance matters.
Much in the same way that mobile/social gaming didn't displace "hardcore" gaming, ARM hasn't displaced Intel. It just found another audience with different priorities.
(a) Games of tomorrow will be made to host creative communities, which almost mandate those games to be for desktops. It's just too awkward to do serious modding work in a tablet.
(b) What is really worrying about the PC is not the format being outdated -- Minecraft, LoL, TF2 and WoW attest to its potential. What's wrong is Microsoft awkwardly replicating mobile ecosystems. Creating videogames now is all about novelty, taking ideas to market fast, and growing cool communities around these games. This just can't happen, when your platform maker is Big Brother's whiny, lazy cousin.
All in all, I just don't get why the Droid Brigade is so enthused about mobile gaming. It's probably an USian localism. The cons: Android smartphones are slow and jerky and shit; the iPhone is like Manhattan if it were a North Korean gulag; fat-fingering your way around is the shittiest of input methods; and most of the games are preheated gimmicks, tastelessly exploited before the camera until boredom strikes.
Or are we supposed to be self-righteous about loving Linux before someone else came along to try to improve it?
1) X11 Fullscreen support - Current support is terrible. Basic games like Tux Racer have taken down my system. Maybe wayland or Mir will fix this.
2) Current Compiz / 3D compositing window manager kill 3D performance - The performance hit when running under any of the 3D composting mangers is unacceptable. Maybe wayland or Mir will fix this.
3) Video driver situation is abysmal. Nvidia and AMD open source drivers run at a fraction of speed of the proprietary counterparts and are riddled with bugs. Proprietary drivers only work for latest generation cards. Intel has decent open source drivers, but no true gamer would consider using integrated graphics.
4) Linux Sound subsystem is a disaster. Sure it's fine for listening to music or watching videos, but the latency and CPU usage is unacceptable to real gamers. MS learned their lesson with Vista/Win7 and now have proper hardware acceleration for sound cards in Win8 like they did in WinXP. Linux had their chance to fix this situation with OSSv4 but instead went with the inferior ALSA.
I would love to see Ubuntu and Value clean this mess up, but I'm not sure they have the resources to pull it off. If Sony had chosen Linux instead of BSD, I could see this happening. But the way things are going with Mir in terms of performance and industry support, I don't see it happening.
I use FOSS AMD drivers, the default sound system for Ubuntu (I presume it's pulse) and the compositing gnome shell desktop environment. I am able to play Valve's ported games quite smoothly and enjoyably, Even things like alt-tabbing out work flawlessly.
Every desktop CPU sold in the last decade, at least the ones capable of running games, supports x86-64. The only reason 32 bit installs exist in the wild is because Ubuntu (still) makes it their default download.
Enter Steam for Linux. Targeting 32-bit Ubuntu, it makes this the de facto platform for Linux games. My desktop OS is now cluttered with multiarch libraries, and the top-end games I want to play won't take advantage of either the better instruction set or all my ram.
Linux gaming could have been a fresh new start, a break from the 32-bit instruction set, but instead we're now stuck installing, maintaining and supporting 32-bit libs forever..
I agree with you, though, it's disappointing that even in abandoning Windows we're stuck with MS-style backwards compatibility.
And what great 64-bit games are there?