And somehow Crysis 1 mods still look better

the visual appeal in end depends more on what the artist can do with what the engine enables him. the deep surface light effects highlighted in new engine apparently would not be possible in crysis 1. at the same time you might like landscapes more, without complex physical effects.

The demo didn't really show much of the terrain stuff, besides most of the advances in engines these days are materials and lighting, physics, that stuff would not really shine through in your comparison which was a lot of terrain...

it will take a long long long time to properly simulate fluids in real time.

O RLY!

Those aren't the most recent real time SPH formulations... Some guys recently incorporated an artificial pressure tern on simplified NS equations in order to make the simulation look like it had proper surface tension. These simulations look good, but that is all.

I recon that SPH simulations are faster than conventional cell based simulations using multiscaling solvers. Nevertheless we are not at the time to accurately simulate the fluid behavior in such complex system in real time using our everyday computers (stick as many Geforce Titan there as you want).

Due to their simpler nature, rigid body dynamics (even soft body dynamics in some cases) can be accurately simulated in real time using our everyday computers, the physics behind are just simpler...

This demo is over 5 years old...
http://www.youtube.com/watch?v=iU3tJ9JmYvA

Fluids can be achieved today if used sympathetically and in controlled areas....

Borderlands 2 has fluids physx implemented but all fluids look the same highly viscous and dense mess (surface tension is a mess too), it feels just wrong. I don't think a cheap implementation (like the one in borderlands 2) would cut it.

Found the new method that I mentioned before:
https://www.youtube.com/watch?v=6WZZARzpckw
^
This looks amazing I must say, nevertheless at some point it just doesn't feel right.

Pre-calculation would be good for cut-scenes etc...

The thing is, when you simulate a flow like a Dam break, you need to not only simulate the water, but also the air!

Well, I have been working on CFD (Computational Fluids Dynamics) for some time now, and that may be the reason that those simulations don't feel right to me.

Yeah it will look like treacle, but I would rather see a bad implementation of something truly next gen than nothing as the next consoles won't be out till 2020 and only then will you have enough power to tackle it properly.

Most people have not seen any fluid stuff in games, so their standards will be quite low.

BL2 fluid looks more like a gimmick, I think you could do more with it than that, but at least its there in principle, it just needs more particles and polygons for the mesh, that will come in time.

A finer mesh wont make up for a bad physics. In fact you can solve the problem in a coarse mesh and then interpolate the solution to a finer mesh, and then solve the problem again using the interpolated solution as the starting point.

I did a simulation of a cube falling through air in a volume of water. It took several days to do 1 second of simulation in a 24 core 32Gb of ram machine. The simulation was physically accurate, but it gives you an idea of the processing power that is needed to properly solver a fluids problem.

I disagree with you on the bad implementation thing. I would like to see the things done right (at least in a convincing way).

I said

"it just needs more particles and polygons for the mesh"
More particles = more physically accurate simulation and more polygons = better representation of the mesh. Yes without more particles you can't have a more detailed mesh as the mesh references the particles for its shape/movement, if they are not there, you won't get a convincing mesh. But I never said that.

"More particles = more physically accurate simulation" This is true only if the physical model behind the simulation is good.

Of course that everything is an approximation and always will, but my point is that compared to the rigid and soft body simulations, the fluids simulations suck.
IMO, if you have very good lighting, soft and rigid body physics, you must have a fluids model to go along with it. I rather have good looking "static" water, that you would expect almost no interaction with, than having water that interacts in a plain wrong way. I understand that computer graphics is a thing that you build upon, but for me everything else is so much in front of fluids simulations, that it wouldn't make sense to have a fluid sim below the standard of the other things. This is how i feel about it.

If you designate them to happen in select areas of the level then you could dedicate a few hundred thousand polygons/particles to the cause.

That should be enough to create something cool.

Also fluids do not have to be water, you could use oil, which would be naturally gloopier and tie in with the tech restraints.

Without talking numbers its pointless, but if a programmer gave me a particle count to play with and poly count for a next gen machine ,I bet I could produce something cool in a short space of time that everyone would agree is better than the usual water you see.

It is not about the fluid or the number of particles, it doesn't work that way... You can set w/e fluids you want, (water and air were just examples), density and viscosity should be enough (disregard surface tension effects to make it simpler).

The domain of the problem is usually meshed with 3D cells, which can be tetrahedrons. For each one of those cells a fraction of both of the fluids as well as an interface are assigned. For each time step, continuity, volume fraction speed and level-set equations must be solved as an iterative simultaneous system. For a domain of like 150k cells, you can compute the solution for one iteration in a reasonable time. The thing is that for each time step you must do around 20-50 iterations (precision can't be set too low, or the solution will diverge). Also the more refined the domain gets, the lower the time step must be, i.e.

Time step = (mininum volume in the domain)^(1/3)/(maximum speed in the domain)

For a coarse mesh it is normal to have time steps of about 1*10^-4s. Consider a target of 60FPS and you can see the amount of compute power that would be needed to properly do such simulations.

But they were doing that over 5 years ago on hardware much less powerful than today's.

http://www.youtube.com/watch?v=iU3tJ9JmYvA

I don't believe with today's computing power which is several times more powerful than what you see in that demo they could implement something 2-3 times more detailed than in certain parts of a level.

What you say doesn't make sense in relation to what has already been released years before.

We probably just have different levels of expectations, so will never agree.

But the next console hardware is not till about 2020, so its a long ass wait.

What do you mean by detailed? Finer cells or more detailed physics wise? What you said is true for the first case (finer cells). Physics simulations aren't like the representation of a sphere (throw more polygons and it will look better and will be a more accurate representation of the sphere), in fact if you set ridiculously small cells you start having numerical problems. IMO the detail (accuracy of the formulation) of the physics model is what needs to be improved and that is what is really costly. Getting back to the cube example: It took roughly 72h to simulate 1s:

Lets say we have a target of 60fps (16+2/3ms per frame). If we took 259200s to simulate 1s, simulating a frame would take approximately 4320s, and this was on a machine with 24 cores and 32Gb of ram. Even if you use a GPU just to do part of the solution, that wouldn't be fast enough. I know that games don't need a solution that accurate, but still we are talking about orders of magnitude of computing power until we get real time fluids simulations to the level of light, soft and rigid body physics simulations.
It is possible that someone has something really really fast and efficient in the works regarding fluids simulations that I haven't heard about...

Is there any way to download these?

try that.
http://www.nvidia.co.uk/coolstuff/demos

The current version of CryEngine is much better than the CE2. Looks better and is faster.

Seems like they're making progress all the time.

The last bit was the best.

An fps set in ancient rome would be awesome!

i like this for an idea, i think it was in a movie already

http://en.wikipedia.org/wiki/Hermannsdenkmal
http://de.wikipedia.org/wiki/Hermannsdenkmal
http://en.wikipedia.org/wiki/Battle_of_the_Teutoburg_Forest
http://de.wikipedia.org/wiki/Varusschlacht

The water evaporation was very impressive.

looks good, if only someone would finally make a good game with cryengine

RAGE all over again, lets talk to Bethesda >8-)

Ok, the water drying out was the coolest thing I've seen in a game engine demo in a while.

Looks good but still it seems like the progress that's being made in 3d engines has been slower lately than it used to be like 5 years ago.

Obviously there will be diminishing returns as they get closer and closer to real life , but stil.

its because consoles are market's priority, and because of they limited hardwere, they slow down evolution of gfx.

noone will spend lots of money on a game engine, that would run only on pc.

true. pisses me off a bit that ppl didn't smarten up and switch to pc already.

on the other hand, in the times of pc supremacy, you had to buy new pc every two years. i understand why ppl prefer consoles

Yeah i guess that's the main reason for it.

I actually quite like this upcoming console generation, the cost/gfx quality ratio appears to be very very close to pc gaming.

the differance between next, and current generation of consoles is pretty big.

while ps3/xbox360 were a jump in terms of gfx, and at start it was better than pc, next gen has worse gfx at start. you just cant catch up with crysis 3. and as we see - crytek is working hard on next engine.

I wonder what the pricetag for the ps3 will be once ps4 has been released.

I still need to buy a bluray player and dont rly wanna pay full price for ps4 games anyway, buying a ps3 might make sense for me

considering ps3 will be supported for at least one year, i doubt the prices will drop down as drasticlly, as it was when current gen was released.

The ps3 and ps vita are both being reduced to $199

That will equate to 199 euro at the very least , still expensive for a glorified bluray player :(

yep 199e

in that case meh, did they say anywhere that the ps4 will be backwards compatible?

It's not, only plays ps4 games. Although they will probably release some on PSN down the line.

they didnt even put SSDs into the consoles tho :S

SSD's are too expensive ...

http://geizhals.de/eu/?cat=hdssd&xf=251_2...240#xf_top

http://geizhals.de/eu/?cat=hde7s&xf=958_2...250#xf_top

A stock 120GB SSD would be enough, especially while they're trying to sell Cloud (online storage) services.

life expectancy would be an issue, no?

just an example to compare ... sure u can put in a 1tb hdd wd green

The best part is that as both systems are running practically identical architectures, the multiplatform games will have less compromises and perform far more solidly.

Its gotten to a point when its more important to woo the developers in regards to ease of use

The roman game near the start is 'Ryse' coming to Xbox One... looks amazing yet previews say it feels pretty much on rails.

Looks like one big quick time event unfortunately :(

I heard that the guy playing on the demonstration was really good at it and hit all the QTEs but people who had a chance to play the game didn't get any.

Apparently the game still continues on even if you don't hit them, you just get a reduced score.

Expected a Monkey_Bussiness thread.

All this talk of graphics and engines...

Remember the days of GeForce4 and when the graphics in this ruled? Also, when games were games and not quicktime events.