Posted at 12:14 by Mark Mackay with 48 comments
During October of last year, in an interview with CVG, Valve’s vice president of marketing, Doug Lombardi stated that he believes PC gaming graphics are close to being as good as they’ll get.
Lombardi told CVG that he believes that ‘graphics have started to top-out. We've got really great-looking games but what we want are more intelligent, more visceral games and the multi-core processors are going to be the way that we get there on PC.’
Some of you might feel that this is nonsense; others might feel that it’s true but you wish it wasn’t because you want to be dazzled by graphical feats of unimaginable excellence. Personally, not only do I believe that Lombardi’s statement holds a good deal of truth but I also actually hope that he’s right.
The reason I do is because of something that happened the other day in the labs. Antony was testing an awesome watercooled rig built around an overclocked Core i7-920 and overclocked Nvidia GTX 295. I watched in awe as the machine played Crysis Warhead in DirectX 10 mode at 2,560 x 1,600 with 4x AA at a 24fps minimum frame rate.
If Crysis is indeed the pinnacle of PC graphics then in a few years or maybe less than that, we’ll be able play our games at the native resolution of a 30in display with all settings on max using mid-range hardware. There are a few rich persons out there that can afford such luxury but I’d never seen a game that looked so good being played so smoothly on such a large display and OMG do I want one!
In my opinion, the often reality-checking detail in Crysis with all the sauce turned on looks just as good as Left 4 Dead at max settings even though the latter game is built on a graphics engine that is the better part of five years old and will run on a pocket watch. As Joe wrote in his Graphics versus Presentation Blog how good a game looks isn’t about how technical the graphics are.
For this reason and the fact I want my future mid-range gaming PC to run my games on a 30in display, I really hope that graphics advancements slow down some and gives the hardware a chance to catch up.
Lombardi told CVG that he believes that ‘graphics have started to top-out. We've got really great-looking games but what we want are more intelligent, more visceral games and the multi-core processors are going to be the way that we get there on PC.’
Some of you might feel that this is nonsense; others might feel that it’s true but you wish it wasn’t because you want to be dazzled by graphical feats of unimaginable excellence. Personally, not only do I believe that Lombardi’s statement holds a good deal of truth but I also actually hope that he’s right.
The reason I do is because of something that happened the other day in the labs. Antony was testing an awesome watercooled rig built around an overclocked Core i7-920 and overclocked Nvidia GTX 295. I watched in awe as the machine played Crysis Warhead in DirectX 10 mode at 2,560 x 1,600 with 4x AA at a 24fps minimum frame rate.
If Crysis is indeed the pinnacle of PC graphics then in a few years or maybe less than that, we’ll be able play our games at the native resolution of a 30in display with all settings on max using mid-range hardware. There are a few rich persons out there that can afford such luxury but I’d never seen a game that looked so good being played so smoothly on such a large display and OMG do I want one!
In my opinion, the often reality-checking detail in Crysis with all the sauce turned on looks just as good as Left 4 Dead at max settings even though the latter game is built on a graphics engine that is the better part of five years old and will run on a pocket watch. As Joe wrote in his Graphics versus Presentation Blog how good a game looks isn’t about how technical the graphics are.
For this reason and the fact I want my future mid-range gaming PC to run my games on a 30in display, I really hope that graphics advancements slow down some and gives the hardware a chance to catch up.
Comments (48)
Discuss in the forumsI see your arguement, but i think you'll go down in history with all the other people that think technology has stopped.
Sorry, I think you must have commented on the wrong article there. Don't worry, it's easily done.
Graphics have a bit of a way to go before they are indistinguishable from real life. But it is possible that in a few years time, we could be playing games that are that good. Then there's 3D Graphics. Then there's holographics. Then there's...Who Knows?
As we enter the nano-technology era, it would be great to achieve truly portable gaming computers, visual output device included. Imagine how we will laugh in 10 years time at the size & weight of our ATX cases and TFT monitors! From your comments, can we infer that you do not think DirectX 11 will bring a major improvement in graphics? And lastly, you are the editor of Bit-Tech, suggest to NEC that you require a longterm test of their 30inch panel, it's common practice in the motor journalism industry, you have to suffer for your art:)
Compare the performance of the 8800GTX, 9800GTX, GTX280 and GTX285, sure they do increase with each iteration but the performance jump has become increasingly smaller each time.
In this respects, I do agree with the article, the technology behind G-Cards does seem to be slowing down somewhat. The question, however, is whether this is because game technology has slowed down, or whether nVidia's relative dominance of the graphics performance sector has meant that they haven't tried as hard between iterations anymore
It's also certainly true that recent games have deviated from the previous "pretty = better" and instead gone more towards "better = better" which, despite being a rather superfluous statement on its own, actually means that they're focussing on making the games more playable, more enjoyable and simply focussing on other areas of gameplay than just graphics. This can only be a good thing, as a pretty game that needs nuclear power and a small loan to play properly doesn't nescessary have the enjoyment value (e.g. Crysis) as something that can be run on a machine 4 years old (e.g. Source engine games)
so wait... your point then is that you'd like 2015's Atom platform to play games at full/full/full/max/max settings?
To me, it seems rather like you are hoping that no further advancements in software technology are madefor games, so you can enjoy high-end gaming on mid-range hardware. It sure sounds nice, but unless i'm wrong *again* about what you are trying to say, i think you'll be disappointed.
If the software side of graphics stops, manufacturers will also stop going bigger. This means that that 2015 GPU (the ATI 9870, or the nVidia GTX1495+GSOX2-9600GT) will be no more powerful then today's generation of cards, just smaller and more economic. Plus, they will still be high-end. The only way for midrange cards to get to that level of performance, is to either play old games (Homeworld for the win!) or for harware and software to get seriously de-synched. You seem to be hoping for that de-synch, but the problem with that is that nobody will buy high-end, and prices of mid-range will go up to the point that they are high-end again.
All that said though, maybe rasterised graphics are reaching its plateau, but not PC gaming graphics as a whole.
I'd have to re-read my blog again but I'm pretty sure I didnt say any of the above.
Quote for truth
If it's not your desire to sit in first class with 3rd class tickets, I'm seriously questioning what it is that you ARE trying to say here...
Well, to be fair, if you're going to put words in peoples mouths then youve got to be prepared to take responsibility for doing so.
It depends if the exponential increase in price/performance ratio of integrated circuits is enough to help GPU speed vastly outstrip gaming requirements. If games graphics plateaued roughly where theyre at now then the amount of money it would cost to actually manufacture a future chip capable of running games smoothly with full sauce at 2,560 x 1,920 with 16x AA would be very low and thus unlikely to cost large sums of money. Especially considering the how fierce competition is in the GPU market.
Thats also before taking into consideration that rasterisation wont last forever. But thats a whole other ball park as there are so many unknowns. I liked Sebbos comment.
A few years of affordable 30in rasterised gaming prior to ray tracing or whatever comes next would be a blessing indeed. Maybe gamers would love it so much that they never bothered the spend the cash to be early adopters of ray tracing and the technology was a flop because we were all loving out 54in displays too much. Though then that would depend on what the game devs were up to. The debate rages on.
before Crysis came out, 8800GTX was sitting at the top for a loooooong time, and people are saying it's enough. only after the release of Crysis, more affordable hardware was released.
I suspect that actual polygon count will not rise a great deal because of the exponential increase in time it takes to make a higher poly-count models.
The 4870 1GB is at $180 now, and it is a decidedly high-end card, perfectly capable of delivering playable rates at 30in (which is 2560x1600 btw, not 1920). It's a pricepoint that only got you midrange stuff two years ago. Generally, i think we can say that prices for similar performance levels have halved in the last year, mostly thanks to the 4870 (introduced at $299), and ATI generally competing again.
I can understand the "want!"-factor of high-end graphics, and I too am hoping for prices to fall even further, i simply disagree with your causality. I'm of the conviction that making software better and better will get increasingly expensive, while making hardware better will get increasingly cheap. Since the two are eachother's incentive to innovate (game with no GPU that can run it is pointless, GPU that no game will ever need is pointless too), i'm hoping for some more software innovations. Both the SourceEngine and the CryEngine are getting a bit dated, and there hasn't been a game release based on a new cutting edge engine for years.
My understanding of the megatexture tech is that it doesn't make the game look much better - but it makes developing and editing levels much easier...
Yes it is. But when the GeForce 6800 Ultra came out it was about £400 and when the 8800 GTX came out, that was about £400. This is because games were increasingly more and more demanding and the GPU's were more advanced in attempt to keep up.
What I'm saying is...
Me too. Just one more time now
qft.
It will soon be like Photoshop: The program gives you the possibilities to do almost anything and is "plateaued" (efficiency not accounted for). The problem is that most designers don't know what they're supposed to want to do.
Soon, graphics hardware will be enough for our rasterization needs, but our graphics developers must want the right things.
I will give a real world example: Company of Heroes, with a handful of sprites (!), produces the most impressive explosions I've seen for a game (I would post a youtube vid if I weren't at work).
Another example are Hollywood animations. They create fur, metals and clouds with the most surprising fidelity. The awe inspiring visuals stun you! ....Until they try to make a human. It's just mentally hard to produce one, to capture all those skin folds, all those muscles that you subconsciously know of from your life experiences but can't list on a sheet of paper. It's a real challenge that makes "artistic license" a crude excuse for making a game with a comic style like Team Fortress 2. If you're going to draw humans and it's going to look wrong, let's at least make it a styled "wrong"!
Don't get me wrong - I love it's style. But if we were able to draw photorealistic humans with ease, we would see way less comic-styled games. Those that aren't comic styled look bad. You just didn't look so critically at them until now. :P
As for evolution, I would forecast 15 to 20 years until graphics plateau in the way the article says. Like Sebbo said, we still need to make the jump to ray-tracing, and I would grotesquely place that as "half-way there".
In minor steps, we would also need to progress into higher resolutions (up to tiny pixels on wall projections), 3D illusions, angled displays (full domes?), etc. Is it actually a matter of display limitations?
In a later moment we could see more "futile" features that consume lots of compute power, such as rendering independently from the POV (e.g. for spectators watching from other angles" on some wacky omnidirectional display.. or something like that)... Ah, if the future were easier to predict! :P
I wouldn't include physics in graphics, but I guess it does mix up in some ways. For instance, volumetric clouds is considered graphics nowadays, but it's really physics. They're particles and you need to calculate how they move around, disperse and difuse light. Same thing for water. It's a material, so it's physics, but it refracts light, and light is done by graphics. Well... light IS a part of physics, eh?
In a short line, physics still has more than a decade to go, for sure.
But if you do want to stand ground in rasterization, we really shouldn't expect many large steps forward. More particles here, sharper textures there, livelier shadows here, bloomier HDR there and so forth. Shiny materials already have bumbmaps, movement already causes motion blur, round surfaces cast Phong reflections.. it's pretty much done. But rasterization is still well below ray-tracing.
One way is to look at a game screen and compare it with a photo of a similar ambient. Then ask yourself: What's missing to make it look better?
If I do that with Crysis and a jungle photo, I would say "not much".
Then again, if I do that with Left4Dead and a photo of a zombie, I would say "WTFOMGBBQ! A ZOMBIE!!" :D
This will be my last try:
I understand what you are saying, i just disagree. I think it's important for the industry as a whole to keep moving forward, and that this will ultimately yield the best results for us consumers, both in price and performance. I can't say it more plainly then that, and as i sense we are both getting mildly annoyed, i'll stop there. I would like to keep this from spiraling into one of those threads that counts down to hitler. :p
/facepalm
The 295 is the very card that was making Crysis Warhead look so amazing on a 30in display and costs £400.
The 4870 X2 was out before then.
Corrected. But we're digressing and the point of my blog still stands which is...
We're not approaching the end of raster graphics IMO. We may be approaching an age where rasterisation isn't used exclusively for every effect, but there is little point in ray tracing something that doesn't reflect light and isn't also curved. It's very easy to reflect light on a flat surface with rasterisation and it's also very efficient so there's little point changing the way you do something to see no visual benefit.
Where I think we're going is towards a hybrid raster/ray tracing rendering model because using ray tracing exclusively would take us back more than a few years - ray tracing vs rasterisation is not black and white and it never will be, that's for sure. There'll be shades of grey for many years to come.
Acknowledged, there's a very few titles, that have shown excellent physics aswell lately, but that doesn't solve the real problem of the whole gaming-market....
most games are damn boring to play!
Good point. I wonder how efficient hybrid models will be. Most likely, ATI and NVIDIA optimization efforts will be crucial in accelerating ray-tracing penetration.
I know alot of peeps turned it off.. try at 8q and it looks as good as crysis far as immersion goes (imo)- felt like I was playing a hollywood flick =]
And FC2, not even worth mentioning.
All these games look great, but the lack of intriguing presentation and the lack of any innovation really kills it.
"I hope PC gaming graphics have plateaued"
What the titles pretty much states is that you hope CG stops progressing, flatlining how the top of a plateau is shaped.
Crysis at max/dx10 and Left 4 Dead at max, side by side they are night and day. Screen shot both of them next to each other. Just look at the draw distance.
"The 4870 X2 was out before then."
the 4870 was released mid august, bit-tech reviewed it at the begining of september, it was reviewed 2 weeks after the fact.
1) actual ground textures, not those blurry crap that we still get in 2009
2) characters that don't look like they're made of shiny plastic
otherwise I am satisfied
Can't say I care much for seeing those kind of responses on here from an article author.
Whether or not Xtrafresh got the wrong end of the stick about your article, there's no need to be an ass about it and I think Xtrafresh deserves kudos for not responding in kind.
It may not be my place to say as a nobody-user, but that kind of behaviour isn't bit-tech, by my reckoning.
The main problem that's holding up RT at the moment is the memory bandwidth requirements. The "Raytracing algorithmn" is very simple indeed, just needing a little FP math, the problem is that it takes a LOT of lookups to RAM, and with lots of cores processing at once, RAM speed, latency and bandwidth are going to be the limiting factors.
Get those licked, and it's just a matter of throwing more cores at it 'til it's done, and those cores don't have to be complex at all (they're all doing exactly the same instructions again and again). a little FP math, a lot of reading and writing to RAM, possibly firing off a few more threads (For reflection) and you're done.
http://img26.imageshack.us/img26/8017/25002363.jpg
http://www.horde3d.org/screenshots/alfred.jpg
Both of these were done with a free graphics engine called Hoarde3D and the top one gets 60fps with a 7600GT apparently (with real time shadows).
There's also the leadwerks engine (not free though).
http://www.leadwerks.com/post/island_shot2.jpg
I'm considering using the first one in a project I'm working on. (The only language I really know well enough is Blitz 3D which is DX7 ).
Anyway the point is that everything is getting pretty good now.
there is one example can put forward here: Ambient Occlusion.
Its currently available through an NVidia Beta driver, and when you apply it to Crysis for example, it drops the frame rate down even further, but adds an extra level of reality that really is impressive. So new types of graphical niceness are appearing all the time, making older games look better (think AA on old games that didn't initially support it) and now Ambient Occlusion. Sure things look fine as they are, but compare it to the graphics in Jurassic Park (made in like 1992 or 1993 or something) and it shows just what the graphics of rendered images COULD look like, i think Nvidia and ATI are constantly striving to get to that point eventually, to be able to render like that in real time on end-user hardware and until we get there, this discussion will continue.
Kinda long the same lines, after watching the F1 at the weekend, i noticed how little detail standard TV really gives you, and although we "care" about jaggies and texture filtering in games, we forget that a lap of Washington in Race Driver GRID, actually looks better already than real footage of real cars racing around a real track broadcast on tv.
Certainly seems that way to me, though i'll happily concede that GPU's have hit the similar clockspeed/heat ceiling that CPU's did recently, hence the slowdown while they work around that (with point #1 being a factor, they don't have much need to vastly improve the performance at the moment as well)
Graphics are mostly pegged to what a PS3 or 360 can do due to most games being multiplatform. PC gamers don't get more then higher resolutions and sharper textures really. It does mean things like nvidia's 3d vision (which makes existing engines look better) in some ways have more of a future then DX11 (which requires a new engine to make use of it) in the short term.