Sections

Carmack predicts “ten times the performance” from GPUs in five years

Monday, 11th November 2013 00:32 GMT By Brenna Hillier

Graphics tech guru John Carmack sees a very bright future for PC gaming just a few years into the future, predicting huge leaps in GPU output before the end of the coming console generation.

Speaking at a recent Nvidia event lovingly transcribed in full by Eurogamer, the id Software veteran and Oculus Rift developer gave his vision of the near future of PC graphics technology.

“Five years ahead, yeah, we can probably make credible comments about that. You’ll still be able to buy Xbox One or PS4 new five years from now – unquestionably. There will be tons of content developed for that,” he said.

“We’ll have 4K resolution displays on tablets and HMDs, and we’ll have another order of magnitude pretty straightforward on there with Moore’s law. GPUs are great at turning transistors into performance and we’ll have ten times the performance. It means that you can probably run that triple 4K display at double the frame-rate from one GPU. That starts looking pretty impressive.”

Carmack said that some things like display technologies and networkinhg advancements are “technical freight trains”, “going forward whether we’re paying attention to them or not”.

“There’s a trillion dollars in economy pushing these things so a lot of that’s going to keep going, and it’s going to be great to sort of be along for the ride and figuring the kind of interesting systems areas, where there’s a convergence of what becomes possible now, that people wanted before, as the sort of cornerstone of where the real innovation is going to happen,” he added.

“We can always turn the cranks on what we’ve already got and always get better, but the insightful things are when you notice something that you hadn’t even thought about and previously dismissed as impossible is now possible and orders of magnitude just sneak up on you like this.”

Carmack said it’s difficult to make predictions much further into the future, as you “can’t wrap your head around these six or eight orders of magnitude” that occur over a few decades. He used the movie Tron as an example; its rendering was state of the art 30 years ago and is now pathetically simple.

Breaking news

9 Comments

Sign in to post a comment.

  1. POOhead

    4K resolution?

    #1 9 months ago
  2. DSB

    Silly Carmack. That’s not really news to anyone.

    It’s Moore’s Law, kind of. The power of hardware tends to double every two years.

    Alas, there’s a slight gap between doubling the power and making it affordable.

    I think people can live with “flat” graphics on consoles, but I’m pretty sure one of the things they’ll have to change in one of the earliest iterations is replacing those HDDs with SSDs.

    It’s too expensive now, but it’s really too much of a performance boost to just leave people waiting on an HDD. Consoles were supposed to be these convenient machines with all the hassle of a PC sliced away, but my PC is ready to use in half the time it takes my 360 to load.

    I think it would become very noticeable a couple of years down the line when consoles are the slowest machines we own.

    #2 9 months ago
  3. Mr Sparkle

    Wow. Bold!

    #3 9 months ago
  4. Telepathic.Geometry

    Dunno why, but I love that last comment, about how now it’s pathetically simple. ^-^

    #4 9 months ago
  5. debug

    “networkinhg” !

    #5 9 months ago
  6. debug

    What was the most interesting bit of this was the enthusiast of Carmack about G-Sync.

    “”G-Sync will “unequivocally make for better games”", in brief.

    Another good bit is the content creation topic. What they’d like is to simplify the process, in order to deliver more to gamers in a shorter time => reducing cost of games among other things.

    #6 9 months ago
  7. ctankep

    They were more enthusiastic about G -Sync being a quick ., easy solution
    to bridging th’ 30 -60 FPS canyon on th’ desktop. It would then let developers
    achieve more [ or seem to ] with current engines an’ minimal re -engineering.
    Mantle for being vendor -specific seems to be th’ sore point with already a
    different flavour o’ OpenGL / ES / PSSL for each platform.

    4K is great but requires so much more o’ your asset pipeline to pull it off.
    I recently saw “Madagascar” Bluray demo’d on 4K an’ you could see all th’
    haloing ., blending artifacts from compositing around characters. This will
    improve but there’ll also be a kind o’ dissonance from games that
    appear to be hyper realistic yet with th’ same basic game systems an’
    mechanics underneath.

    -

    You already see this happening on current consoles with AC4 an’ other
    open world games which struggle with anything beyond driving or shooting
    because hit detection & measuring proximity are easy. Fleshing out these
    brave new worlds requires real work an’ innovation in systems planning .,
    narrative structures an’ especially interaction design.

    Perhaps some o’ that will come from meshing what were once “static”
    systems with online using stuff like web API’s to pull states ., behaviours
    an’ personalisation into games. I don’t mean cosmetic appearance but
    maybe using something like sentiment analysis to find -out a players mood
    based on th’ tone o’ frequency o’ their emails ., then for th’ game director to
    be able to re -structure it’s content or pacing.

    -

    Once VR really gets going other aspects o’ play will become more
    apparent or worthwhile o’ exploration. Th’ largley linear structures we have
    in games now are less than optimal means to explore stuff beyond skill an’
    dramatic intensity. A more network -like topology seems much more apt
    for facilitating player consequence ., emergence an’ how to give players
    more tools for creativity or expression.

    Games to become less like a laundry -list o’ things to do an’ more like an
    environment that can support different intersecting play -styles with an
    emphasis on communities much like “Eve” online. In th’ 50s television
    was general purpose [ talk + variety shows ] then branched out to form
    different genres. Now with YouTube niche content is king though there’s
    no sideways context or linking between all o’ this.

    -

    Finally there’s no reason for screens to be flat anymore. Flexible OLED
    should finally supplant th’ form factor o’ CRT an’ glass an’ displays to be
    based on more spherical designs if not entirely virtual. Th’ act o’ looking
    an’ being immersed comes from moving your head around in space an’
    our displays should reflect this.

    Wearable computing too will take off. Th’ iPad has shown that we don’t
    need a big desktop machine to do 90% o’ our computing tasks. I can see
    a future where all my computing fits into a miniaturised pico -projector .,
    that’s basically a networked dumb terminal running from th’ cloud that
    can talk to different input -> output devices. Still it doesn’t account for
    any behavioural changes in computing that might arise.

    – Chuan

    #7 9 months ago
  8. debug

    “They were more enthusiastic about G -Sync being a quick ., easy solution
    to bridging th’ 30 -60 FPS canyon on th’ desktop.”

    That’s exactly the purpose of G-Sync.

    I’m gonna paraphrase:
    To eliminates screen tearing, VSync input lag, and stutter, G-SYNC synchronizes the monitor to the output of the GPU, instead of the GPU to the monitor, resulting in a tear-free, faster, smoother experience.

    So now game engines won’t have to worry about rendering their world at exactly 30 or 60Hz.

    —–
    Anyway, good comment Chuan, but it looks like you typed this with your phone or smth. Lots of ‘ that makes it weird to read :p

    #8 9 months ago
  9. TheWulf

    @6

    Really nice post there, Chuan. Really nice! And I love how you express yourself, too. Good to have more eccentric, characterful people around, to be honest. Livens the place up.

    Another point I agree on is… well, one I’ve always harped on about myself. And that’s pretty much that fidelity is boring. The only graphical feature I’ve even been remotely interested in the last good number of years is tessellation, because it does add something interesting and new — but mostly, I’m disappointed by how simple (in a bad way) games tend to be.

    I think there’s going to be a degree of laziness and comfort zones to breach with regards to mainstream developers as well. The sorts of simplistic mechanics you speak of are stuck to not solely due to hardware limitations, but because people want to do the least amount of work possible for the most amount of profit.

    So most games are these incredibly simplistic shooty-stabby things. They’re about murder because murder is easy. Worse, it’s become less imaginative because homogenised software solutions allow them to be even lasier. If you can have 20+ guns that fire almost identically (I’m looking at you, modern military shooters), then you can just build a basic gun system and swap some variables around. It’s far more difficult with a game like Unreal Tournament, where each gun is doing wildly different things.

    Some fighting games are similar in that all you have to do is spam buttons to win, missing out on the elegance of older games. You’d never really see something like Lugaru or Overgrowth in the mainstream because no one wants to spend that much time working on systems.

    So what happens is that simplistic systems are created, homogenised, and then spoonfed to the mainstream. And all of these games are ultimately damn near identical with only minor variables flipped in how they play. So instead of being as different than Quake is from Unreal Tournament, you can barely tell them apart. If you’ve played one, you’ve played them all. Developers can put this down to “hardware limitations” and leave it at that.

    I wonder if they’re still going to use that excuse on the newer machines? Whether it’s still just going to be the same old homogeneous mechanics but with shinier graphics?

    I will give BioWare credit where it’s due, though. In both player modes of ME3, they had guns and abilities that all performed very differently from each other. So you could have a bunch of characters that provided very, very different experiences from each other in a way that was reminiscent of TF2. It felt really oldschool, in a good way.

    People actually criticised that for being too arcadey and not realistic enough, which is funny because I often hear ‘realism’ leveraged as a criticism against innovation, good (if stylised) aesthetics, interesting mechanics, unique gameplay elements, and so on.

    So what I’ve come to understand is that this equation is the truest thing of all:

    Realism === Boring ++ Homogeneous ++ Abridged

    Which just plays into what the mainstream developers want to do. Why be anything but lazy when the mainstream audience is going to criticise you for it? Want to make a crazy, colourful superhero game with unique mechanics, something unlike anything you’ve ever played before? Well, that’s just not cricket, because it’s not realistic.

    I think that we’re going to see this homogenisation pushed further as fidelity increases. I don’t really know what word describes it better than homogeneous, because it’s literally ‘everything becoming the same.’

    Mario and Sonic were vividly different games, despite sharing genres.
    Shining Force and Final Fantasy were vividly different games, despite sharing genres.
    Day of the Tentacle and The Pandora Directive were vividly different games, despite sharing genres.
    Unreal Tournament and Quake were vividly different games, despite sharing genres.

    But at some point we went over a tipping point where all of the budget went into graphics and Michael Bay experiences, and less into making something very different, very unique. Yes, there are examples of this that still exist, but Franchise VII is way more common, now.

    And look at the buttonspam of Assassin’s Creed, which proudly displays some of the most simplified game mechanics I’ve ever seen. It plays itself. You just push a couple of things to do stuff, and it’s all the same stuff, for all of the games. The only thing that’s even remotely different or interesting is the naval stuff, and that deserves to be in a different game because you’re still doing more of the assassin stuff than the way more interesting nval stuff.

    I don’t know.

    I just have this horrible feeling that even with improved hardware, we won’t see a change when it comes to complex, satisfying, and diversified gameplay mechanics. That’s something I desperately want. I want AI that can react to me in interesting ways, and I want games that aren’t just shooty-stabby experiences aimed at people who don’t really pay close attention to how shallow their chosen gaming experiences are.

    It’s sad, because there are games from the ’90s whose AI and mechanics could run rings around stuff today. It’s sad, for example, that the mechanics and AI of the original Gothic game are still so far superior to Skyrim (and without using loading doors).

    This is what really grinds my gears about the mainstream.

    The mainstream right now is mostly button-mash controlled psychopathic killing sprees with all of the spectacle.

    And when I think back to the ’90s, and I think of stuff like the Myst games, where I was exploring these brilliant and beautiful worlds, with their stellar writing, and genuinely clever puzzling mechanics. I think of that and I think of how amazing and special that was. Then I look at today and the mainstream can basically be summed up as ‘press A to kill everyone,‘ so it’s hard to not be bitter.

    #9 9 months ago