• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

Epic talks Samaritan and the future of graphics.

Mik2121

Member
Straight from Epic's tweeter.
EpicGames
Epic Games talks Samaritan and the future of graphics. http://ow.ly/4Rao5

At this year’s Game Developers Conference, attendees were wowed by Epic Games’ Samaritan presentation, a sneak peak at the visual fidelity achievable with Unreal Engine 3 and the horsepower of three NVIDIA GeForce GTX 580 graphics cards. Offering a world first look at a real-time demonstration constructed entirely in DirectX 11, Epic’s Samaritan demo set tongues wagging, and served as a wake-up call for those in the industry still languishing in the dark ages of DirectX 9.

In the opening frames of the demo, one of the most talked-about advancements is shown. Known as Bokeh Depth of Field, it is an upgrade to the standard out-of-focus blurring seen in games of recent years, but has in fact existed for over a decade, having been demonstrated by the now-defunct 3dfx using their famed Voodoo T-Buffer in 1999. The Bokeh component of the new technology relates to the out-of-focus, definable shapes seen in television and film, which are typically used to enhance the mood or visual quality of a scene. Derived from the Japanese word "boke," which literally means "blur," Epic Games’ Martin Mittring believes that the effect is far more than mere obfuscation, in that "it allows you to lead the viewer’s attention and creates depth," an effect that is "very important for storytelling, which is an integral part of today’s games." Mittring, Epic Games’ Senior Graphics Architect, also believes that "this effect can be distracting," but, "in some situations it can be used to improve aesthetic quality, like when aiming through a weapon."

http://www.geforce.com/#/News/artic...dia-talk-samaritan-and-the-future-of-graphics


I would post pics, but they are added as Flash and the links take you to HUGE versions of those pics only...


Some highlights:

n the opening frames of the demo, one of the most talked-about advancements is shown. Known as Bokeh Depth of Field, it is an upgrade to the standard out-of-focus blurring seen in games of recent years, but has in fact existed for over a decade, having been demonstrated by the now-defunct 3dfx using their famed Voodoo T-Buffer in 1999.

The technology for the Samaritan demo took eight months in itself to develop and implement, work having first begun in June 2010

As already mentioned, the demonstration ran in real-time on a 3-Way SLI GeForce GTX 580 system, but even with the raw power that configuration affords, technological boundaries were still an issue [...] but with enough time and effort, could the Samaritan demo run on just one graphics card, the most common configuration in gaming computers? Epic’s Mittring believes so, but "with Samaritan, we wanted to explore what we could do with DirectX 11, so using SLI saved time.

I won't be posting more because then you just wouldn't need to read the actual thing :P
 
I'll be honest, while it's interesting, I don't think anyone in their right mind wants to low 2k$ and buy 3 560GTX to get that kind of fidelity.

While the proof of concept is nice, it has next to no application gaming wise and won't for a at least 3-4 more years for single cards to catch up (and that's in the PC field, don't expect home consoles to do anything like that before PS5/Wii3/X1080)
 
I'm interested, and I'm hoping the May UDK comes out in the next couple of days, but I don't see anything new from what you quoted. It sounds like the same demo, same effects, and same concepts from back at GDC 2011.

I would love clean, anti-alised graphics and better performance from the UDK. Bokeh is somewhere near the bottom of my feature wishlist, I think. :P
 
Fularu said:
I'll be honest, while it's interesting, I don't think anyone in their right mind wants to low 2k$ and buy 3 560GTX to get that kind of fidelity.

While the proof of concept is nice, it has next to no application gaming wise and won't for a at least 3-4 more years for single cards to catch up (and that's in the PC field, don't expect home consoles to do anything like that before PS5/Wii3/X1080)

Well said.
 
while the tech behind it is interesting, I can't say the art style is appealing. granted, that's not the point of the demo in the first place
 
Only thing im excited about UDK is I can develop for Mobile games. So yeah, asides from that nothing else. Looking foward to render some characters, but first in my list is the mobile developing, then off to build up my portfolio.

$$$ first
 
I saw the demo that people kept talking about just a couple of days ago and it was impressive. It doesn't matter much to me since I won't see that kind of fidelity until some time after 2016 on the PC at best.

Edit: I saw the boke effects a few times, but I'd never seen it before so I was a bit confused. It will certainly have a place in the future.
 
This is interesting as well:

"We're extending Unreal Engine 3 to remain the most powerful and feature-rich engine for the duration of this hardware generation." The next generation will arrive one day, and on that point Sweeney says, "as we move to the next hardware generation -- thinking beyond current DirectX 11 hardware to future consoles and computers with entirely new levels of performance -- we'll combine the latest and greatest of Unreal Engine 3 with significant new systems and simplification and re-architecting of existing systems to become Unreal Engine 4."

samaritanbillboarduelogk1d.jpg

samaritanbokehie5ke2.jpg


samaritanconceptarthjpy.png
 
Fularu said:
I'll be honest, while it's interesting, I don't think anyone in their right mind wants to low 2k$ and buy 3 560GTX to get that kind of fidelity.

While the proof of concept is nice, it has next to no application gaming wise and won't for a at least 3-4 more years for single cards to catch up (and that's in the PC field, don't expect home consoles to do anything like that before PS5/Wii3/X1080)

In 2 years time that will be one card's worth of performance. So it is interesting in that regard, and that is ignoring optimization that inevitably happens with new techniques. We haven't seen mass adoption of even directx 10 in the pc space, so I'd imagine there is quite a bit of optimization that will occur on these effects that are quite graphically intensive to perform right now.
 
Fularu said:
I'll be honest, while it's interesting, I don't think anyone in their right mind wants to low 2k$ and buy 3 560GTX to get that kind of fidelity.

While the proof of concept is nice, it has next to no application gaming wise and won't for a at least 3-4 more years for single cards to catch up (and that's in the PC field, don't expect home consoles to do anything like that before PS5/Wii3/X1080)
I think that the next gen consoles will be able to achieve similar results with different techniques.
 
Fularu said:
I'll be honest, while it's interesting, I don't think anyone in their right mind wants to low 2k$ and buy 3 560GTX to get that kind of fidelity.

While the proof of concept is nice, it has next to no application gaming wise and won't for a at least 3-4 more years for single cards to catch up (and that's in the PC field, don't expect home consoles to do anything like that before PS5/Wii3/X1080)
The problem isn't going to be people affording the hardware to run it. In three years it'll be as cheap as modern hardware is today.

The problem is production budgets, which are going to skyrocket for AAA games. If everything except Call of Duty has to fear selling *gasp* less then two million copies this gen, then how bad could things get when budgets are fifty, sixty, eighty million.
 
The_Technomancer said:
The problem isn't going to be people affording the hardware to run it. In three years it'll be as cheap as modern hardware is today.

The problem is production budgets, which are going to skyrocket for AAA games. If everything except Call of Duty has to fear selling *gasp* less then two million copies this gen, then how bad could things get when budgets are fifty, sixty, eighty million.
There's nothing "cheap" about a 600$ GTX 580, you won't see that much power into a console by 2016, let alone 2013 or 2014.

Optimization aside and there's only so much you can do), this was nice but remains pointless for the foreseable future.
 
The_Technomancer said:
The problem isn't going to be people affording the hardware to run it. In three years it'll be as cheap as modern hardware is today.

The problem is production budgets, which are going to skyrocket for AAA games. If everything except Call of Duty has to fear selling *gasp* less then two million copies this gen, then how bad could things get when budgets are fifty, sixty, eighty million.

Budgets are not really going to get much bigger , assets already get done better than anything they can do this gen take GT5 cars for eg.
Making some stuff higher res, adding some stuff the GPU can do not going to raise things by a huge amount.
 
felipepl said:
I think that the next gen consoles will be able to achieve similar results with different techniques.

They will definitely be able to achieve the same results - and probably more (providing we get a normal generational leap from Microsoft and/or Sony, as I think we will). The whole thing would be useless without hardware to support it and Epic is not making these changes just to show off.
 
i tried to read this article, i really did. but after about the 3rd bokeh example i started skimming. by the bottom of the page i was just looking for neat pictures.

like others have said, this is a glimpse in the future of what gaming will be later this decade. i guess for me, a non-master race, it doesn't hold much relevance at the moment. especially when some of the best games of the moment do not come anywhere near this level of graphical fidelity.
 
Original-Blue said:
We need a UDK thread on here to discuss UDK, how to use and to post various things that people have done and created using it.
I'm all for this. I know Mik and I have messed with it. Is anyone else interested? I was hijacking a different thread for a while with UDK-related stuff. Of course, the official forums would be better places for asking how to do things in general, but it would be neat to have UDK thread here if the mods don't mind and a few people are interested.
 
Unless next gen cards are beastly. There is a very slight chance new consoles could touch this. 3 580s is a fuck ton of power just one of them is what a normal generational leap above the console are now.
 
I am excited/scared what new workflow artists have to do to create "next gen" graphics. It took a long time for a lot of artists to adopt zbrush, normal maps, specular maps, and move from simple polygonal models to building a high rez, low rez and LOD models.
More realtime shadows, realtime colored lighting, bigger texture memory and more complex shaders would be a welcome change.
 
Krauser Kat said:
Unless next gen cards are beastly. There is a very slight chance new consoles could touch this. 3 580s is a fuck ton of power just one of them is what a normal generational leap above the console are now.

But this is not a normal console generation, it'll be like almost two standard generations in length. Also, bear in mind that hardware does not produce the same results on consoles and PC (it makes sense to optimize much more on fixed spec machines, and the OS overhead is much lower) and that consoles often launch with hardware that is ahead of the PC high end at the time, at least in some regards (as seen in Cell and Xenos this gen).


MiniBossBattle said:
I am excited/scared what new workflow artists have to do to create "next gen" graphics. It took a long time for a lot of artists to adopt zbrush, normal maps, specular maps, and move from simple polygonal models to building a high rez, low rez and LOD models.
More realtime shadows, realtime colored lighting, bigger texture memory and more complex shaders would be a welcome change.

Indeed, only a few big studios could afford that at the moment. That's one of the reasons why all the speculation about Xbox 360's and PS3's successors launching before 2013 is crazy talk.
 
dionysus said:
In 2 years time that will be one card's worth of performance. So it is interesting in that regard, and that is ignoring optimization that inevitably happens with new techniques. We haven't seen mass adoption of even directx 10 in the pc space, so I'd imagine there is quite a bit of optimization that will occur on these effects that are quite graphically intensive to perform right now.

You're not going to see a card beat 3s GTX 580 in 2 years time. It'll be closer to 3-4 years.
 
gundamkyoukai said:
Budgets are not really going to get much bigger , assets already get done better than anything they can do this gen take GT5 cars for eg.
Making some stuff higher res, adding some stuff the GPU can do not going to raise things by a huge amount.

this, plus i believe as we get closer to poly/texture saturation, devs will get huge libraries to pick assets from.
They'll be able to license perfect car models, or trees, or brick walls, or faces, or buildings and tweak them instead of building them from scratch in house.
I actually think development costs will go down, albeit slightly, in a few years through assets sharing.
 
"As already mentioned, the demonstration ran in real-time on a 3-Way SLI GeForce GTX 580 system, but even with the raw power that configuration affords, technological boundaries were still an issue [...] but with enough time and effort, could the Samaritan demo run on just one graphics card, the most common configuration in gaming computers?"

This seems a bit contradictory to me. One the one hand they admit that even the power provided by 3 580 wasn't enough (which i assume is what they meant by having issues), but on the other hand they think they can optimize it to run on a single 580? That'd be one hell of an optimization.
 
Blizzard said:
I'm all for this. I know Mik and I have messed with it. Is anyone else interested? I was hijacking a different thread for a while with UDK-related stuff. Of course, the official forums would be better places for asking how to do things in general, but it would be neat to have UDK thread here if the mods don't mind and a few people are interested.
I'm all for it too. Don't think we are many using UDK here, but still we have had way less popular threads going on, so I don't see why not.
Hopefully a mod gives us the OK and we can get a small thread going.
 
Angelus Errare said:
Good to see Epic finally got with the times on their lighting because sorry to say. UE3 was very shitty in that regard (shadows too).

Play the Gears 3 beta. It's massive improvement in both visuals and performance, especially in the lighting department.
 
shuyin_ said:
"As already mentioned, the demonstration ran in real-time on a 3-Way SLI GeForce GTX 580 system, but even with the raw power that configuration affords, technological boundaries were still an issue [...] but with enough time and effort, could the Samaritan demo run on just one graphics card, the most common configuration in gaming computers?"

This seems a bit contradictory to me. One the one hand they admit that even the power provided by 3 580 wasn't enough (which i assume is what they meant by having issues), but on the other hand they think they can optimize it to run on a single 580? That'd be one hell of an optimization.

They said just one card, but they didn't specifically say the 580. I took that to mean one card in the future like the IDK 780 or some shit.
 
I'm just hoping that the next generation of consoles can run something close to this fidelity at 1080p. I'd like to see an Assassin's Creed game optimized for this level of hardware. When you'd climb up to the top of buildings, it would be amazing to see NPCs mulling about as far as the eye can see and to see high detailed lighting/shadows applied to all of the buildings in the distance.
 
Angelus Errare said:
Good to see Epic finally got with the times on their lighting because sorry to say. UE3 was very shitty in that regard (shadows too).
UE3 does just fine with shadows and lighting.

It's up to the developer to make it work. Some people just didn't take the time.
 
Blizzard said:
I'm all for this. I know Mik and I have messed with it. Is anyone else interested? I was hijacking a different thread for a while with UDK-related stuff. Of course, the official forums would be better places for asking how to do things in general, but it would be neat to have UDK thread here if the mods don't mind and a few people are interested.

I would love to have such a thread too. I want to learn how to use it :]
 
-PXG- said:
Play the Gears 3 beta. It's massive improvement in both visuals and performance, especially in the lighting department.

Lightmass does wonders, it was painfully overdue.

Izayoi said:
UE3 does just fine with shadows and lighting.

It's up to the developer to make it work. Some people just didn't take the time.

Far from fine, before Lightmass it was horrible.
 
The demo looked nice, but did it look three GTX 580s nice? I'm gonna say no. And what's the big deal with bokeh DOF? Crysis 2 does it, and in DX9 even.
 
Izayoi said:
UE3 does just fine with shadows and lighting.

Eh? The lighting prior to the latest lighting update in UE3 games is just terrible. The only UE3 game I can recall with good lighting is Mirror's Edge, and well, it doesn't use UE3's lighting engine.

Now, the demos we've seen of the new lighting engine actually showcase some pretty great lighting.
 
-PXG- said:
Play the Gears 3 beta. It's massive improvement, especially in the lighting department.
Funny enough,nthe new lighting environment (lightmass) has been around for a while already.

Also, for the people saying that it woud cost much more to make these games.... no. Just no. Most of this stuff are post-processing shaders. That means that you only do it once and it gets applied to everything. Of course, a game would have A LOT of these shaders (or effects) to fit different environments, but it won't be the same time consuming step that using normals (etc...) and complex shaders meant.
Current games' assets are created in both low poly (the in-game model) and high poly (sculpted or subdivided model). Of course, artists will need to sculpt to a high enough level so that when the normals (and the tessellation if they get that to work too) kicks in, there's enough info to read from. But then again, most assets nowadays are created with quite a lot of detail. As for the low poly, it's just a matter of giving it a few extra edges to improve on the silhouette and details of the model.
With tools like Zbrush getting better by the day, and extra tools to help baking maps (RTT, XNormal, etc...) or even creating em in Photoshop (nDo), making quality assets is faster than ever, and it will continue getting even faster, which means less production cost.
 
infinityBCRT said:
I'm just hoping that the next generation of consoles can run something close to this fidelity at 1080p. I'd like to see an Assassin's Creed game optimized for this level of hardware. When you'd climb up to the top of buildings, it would be amazing to see NPCs mulling about as far as the eye can see and to see high detailed lighting/shadows applied to all of the buildings in the distance.
I assume that would need massive amounts of RAM and a powerful CPU to boot.
 
They chose a bad art style/setting to showcase the tech. It's always more impressive to see a new tech mimic photo-realism in say, a grassy field or a jungle.
 
Is it bad that this demo does very little for me? To me, even though I understand it's a complete generational leap and extremely powerful tech, I still had to do a double take to determine it wasn't just a PS3 game or something. I must not pay attention to graphics the way I used to...

Also, things still looked too shiny.
 
herzogzwei1989 said:
hopefully the next-gen Xbox and PS4 will be able to run this using Maxwell generation GPU, and its AMD equivalent.

Don't get your hopes up. The specs for the next generation Xbox and Playstation are both probably locked down by now, and it's extremely unlikely they will have GPUs that are even close in power to a single GTX 580.
 
Zefah said:
Don't get your hopes up. The specs for the next generation Xbox and Playstation are both probably locked down by now, and it's extremely unlikely they will have GPUs that are even close in power to a single GTX 580.
No they're not, the 360 and PS3 specs were changing all the time, devs didn't have final devkits until relatively close to launch. If they're launching in late 2013/2014 then they've definitely not locked down the specs yet.
 
Anyone care to take a guess as to when the prices on 580's will drop? I'm building a high end pc in June and would like to do it with 2 580's for less than 2k.
 
Zefah said:
Don't get your hopes up. The specs for the next generation Xbox and Playstation are both probably locked down by now, and it's extremely unlikely they will have GPUs that are even close in power to a single GTX 580.

We shall see. Both Nvidia's Kepler (2011) and Maxwell (2013) are going to be significant leaps beyond Fermi (2009-2010). The GTX 580 is really 2009 technology released in 2011.

nvidia_roadmap.jpg
 
Mr_Brit said:
No they're not, the 360 and PS3 specs were changing all the time, devs didn't have final devkits until relatively close to launch. If they're launching in late 2013/2014 then they've definitely not locked down the specs yet.

Depends on what you mean by specs. MHz, ram and storage size, sure. Those can fluctuate until they start manufacturing.

Architectures and features of the GPU and CPU are probably close to locked.

You probably need the chips validated at least 3 months before launch. Validation can probably take 3-6 months more if you need to re-spin. Initial manufacture can take 1 to 2 months. Verification of a locked down design as big as these can be be even 6 months, so you looking at around 15 months for all the work after the feature set/architecture has been locked.

If MS is releasing something next year, it's already been designed and locked. If it's 2013, then it will be finalized by the end of this year.
 
McHuj said:
Depends on what you mean by specs. MHz, ram and storage size, sure. Those can fluctuate until they start manufacturing.

Architectures and features of the GPU and CPU are probably close to locked.

You probably need the chips validated at least 3 months before launch. Validation can probably take 3-6 months more if you need to re-spin. Initial manufacture can take 1 to 2 months. Verification of a locked down design as big as these can be be even 6 months, so you looking at around 15 months for all the work after the feature set/architecture has been locked.

If MS is releasing something next year, it's already been designed and locked. If it's 2013, then it will be finalized by the end of this year.
I'm assuming Sony/MS are releasing either late 2013 or early 2014 which means that specs and the architecture of the GPU can still change.
 
Zefah said:
Don't get your hopes up. The specs for the next generation Xbox and Playstation are both probably locked down by now, and it's extremely unlikely they will have GPUs that are even close in power to a single GTX 580.
Generally basic architecture would be locked down 18 months out but I don't think hardware gets locked down until the final 6 months or so before launch. Usually its a balancing act in terms of seeing what the competition is doing and what third parties want.
 
Top Bottom