[ultragpu]

Sorry folks, I was just been too lazy. SSGI being screen space Global illumination as linked by Nebula. From my previous post I was just hoping at least one of them gets the go in the next Killzone. But hey, let's look more into it whenever it gets officially announced.

Originally Posted by Nebula

Perhaps for screen-space global illumination?

Link
http://www.crytek.com/fileadmin/user...on_Volumes.pdf

[ihamoitc2005]

Originally Posted by ultragpu

The smoke trail in KZ2 especially the ones from those Tank rounds are most impressive. You can see the dense plumes of particle traveling in great length and duration, observe the slightly curved movements of the rocket trail as it simulates heat tracking effect.

My friend, it is for this kind of thing, that some will not notice, but is amazing for them to have, that I feel Killzone 2 is the best graphics I have seen. The dynamic physics and graphics effects are, for me, really amazing. Some people say this is easy to do but I see no other game that has this kind of technology. GG must have done a lot of research and really love their job.

[TrungGap]

Originally Posted by Dr. Nick

I'm not sure if it is on same thing but I remember smoke in Call of Duty 2 being persistent.

Hmm...what about smoke grenade? Many games have 'em, and they're not like distant fog. They have a volume appearance and dissipate slowly.

[edit] in GeOW2, you can use smoke grenade and stick an enemy...effectively tagging them with smoke. It's pretty cool and cleaver way of showing 'em who's boss.

[Acert93]

So maybe a developer can outline how they would do interactive, persistent smoke? What techniques would they use and what is the expected cost? Is an interactive smoke trail that last 10 seconds [x%] more difficult than a standard non-interactive smoke that dissipates in 2 seconds?

Btw, a couple years back I linked to a tech demo showing off persistent smoke trails that slowly diffused. I cannot find it now, but it was 2005/2006. Neat demo with a novel approach I believe.

[zed]

heres a demo of mine from 4 years ago
http://www.youtube.com/watch?v=St7Ws2mZr7Q (near the end)
scuse the singing doing a dylan -> bowie + also the vid quality

the smoke does dissipate, also receives shadows/lighting

the reason smoke normally doesnt hang around
A/ you cant see things in the game (gameplay downgraded)
B/ standard method with 2d particles equals a lot of overdraw which aint good for performance

I do have an idea which Ive been meaning to try out for a while (so much to do so little time)
involving raycasting (thus decent CPU required) but the results visually will be much superior to whats done now + in fact maybe even quicker since overdraw is nonexistant

[ultragpu]

Originally Posted by zed

heres a demo of mine from 4 years ago
http://www.youtube.com/watch?v=St7Ws2mZr7Q (near the end)
scuse the singing doing a dylan -> bowie + also the vid quality

the smoke does dissipate, also receives shadows/lighting

the reason smoke normally doesnt hang around
A/ you cant see things in the game (gameplay downgraded)
B/ standard method with 2d particles equals a lot of overdraw which aint good for performance

I do have an idea which Ive been meaning to try out for a while (so much to do so little time)
involving raycasting (thus decent CPU required) but the results visually will be much superior to whats done now + in fact maybe even quicker since overdraw is nonexistant

Sounds like it can benefit the Cell, using more CPU cycles instead of the limited bandwidth for overdraw.

[Billy Idol]

Originally Posted by Shifty Geezer

It's being talked about in relation to the consoles' abilities to alpha-blend many particles, and whether KZ2's example is something extraordinary or not. Where it looks 'cool' or not is immaterial for this technology thread.

got it.


Regarding the persistence of the KZ2 smoke. Here is a youtube vid (althoug very bad quality) that perfectly demonstrates the longevity of the smoke:
at the begining, when the HGS launch their rockets from the buidlings...the smoke trail last very long.
Another good example is the smoke after the bridge explodes, starting at about 51sec it lasts over 10 sec.

http://www.youtube.com/watch?v=7Xq3Rslr190

For me this is (not only cool ) but extraordinary as I did not observe this in other console action games (FC2, GeOW1+2, UC2, ...)

[upnorthsox]

Originally Posted by TrungGap

Hmm...what about smoke grenade? Many games have 'em, and they're not like distant fog. They have a volume appearance and dissipate slowly.

[edit] in GeOW2, you can use smoke grenade and stick an enemy...effectively tagging them with smoke. It's pretty cool and cleaver way of showing 'em who's boss.

2 different dynamics though, one is a cloud burst and the other is an emitter.

[catisfit]

Originally Posted by zed

6msec though is a long time, think of it as 60fps == 16.7 msec per frame ( thus youre approaching half your available time!)

For those 2 SPUs.

[catisfit]

Originally Posted by Richard

AFAIK (some correct me please) the dual-layer problem is due to the refocusing of the laser lens AS you transition, not an overhead penalty of the second layer itself. Thus keeping the transitions to a minimum would reduce the penalty.

It's both. There's a transition penalty as you say, but streaming data from the second layer is slower than from the first as well.

I believe there was a thread on here many moons ago with (estimated) figures. EDIT: here I believe.

[brain_stew]

Originally Posted by zed

It depends on what theyre doing
GPUs are fast dumb machines
CPUs are slower intelligent machines

cell is somewhere in the middle

yes for certain grafix things cell can run rings around what a GPU (though the latest + future GPUs from ATI/nvidia are catching up)

6msec though is a long time, think of it as 60fps == 16.7 msec per frame ( thus youre approaching half your available time!)
30fps == 33.3msec

Yeah, but that 6ms is when using just two SPUs and the PS3 has seven (or six depending on how the OS reserves resources) at its disposal. Are we suggesting that developers are using all the resources that the SPUs have to offer in their current engines?

With most games targetting 30fps, is using less than 1/5th of the resources of just two of those SPUs to add a very impressive and beneficial effect really that expensive? I'd argue the opposite, its probably resources that a lot of engines could give up without any impact on performance, but give a huge benefit to many scenes. Wasn't Killzone 2 only fully using 4 SPUs according to that presentation they released? Surely we'd expect third paty titles to be making less effective use of the SPUs, no?

I dunno, I'm well aware those figure are just guidelines but that sort of performance sounds very encouraging to me.

Edit: Just een the mod note above, apologies, feel free to move this post elsewhere.

[liolio]

Originally Posted by brain_stew

Yeah, but that 6ms is when using just two SPUs and the PS3 has seven (or six depending on how the OS reserves resources) at its disposal.

Actually you're right and wrong
there are seven usable spu in the cell, one is taken for the OS hypervisor, another one can be used by OS/hypervisor from time to time. Devs have 5 (sure) and something to play with.

Are we suggesting that developers are using all the resources that the SPUs have to offer in their current engines?

Well ND stated so even in some PR stunt. Could be true and that would not mean that they can't do better with those resources anyway

[brain_stew]

Originally Posted by liolio

Actually you're right and wrong
there are seven usable spu in the cell, one is taken for the OS hypervisor, another one can be used by OS/hypervisor from time to time. Devs have 5 (sure) and something to play with.

Well ND stated so even in some PR stunt. Could be true and that would not mean that they can't do better with those resources anyway

Oh, I'm sure Naughty Dog are (and SSAO is probably one of the good uses they have for them, heh) but I was talking more in regards to your average third party developer here, who one would expect not to be on quite the same level as Naughty Dog or Guerilla games when it comes to making good use of the SPUs. What are the restrictions on that 6th core then? Would reserving 6ms in a 30fps title (i.e. around ~20% of resources) for SSAO be too much? If so, then yeah that changes things and it definitely makes that performance figure become less attractive. Now, if only developers had full access to the proper 8 SPU chip.............

[marcus_rocks]

One question for Joker454:
With the advantage of high bandwidth, why is Forza 3 limited with transparencies (cannot see through car glass window) whereas GT5P has abundance of transparencies and sports a 4xMSAA?
I think we cannot point our finger at one single factor and declare the winner. There are many ways to work around the problem.

[Neb]

I thought you could see through Forza 3 windows it is just that they are very dark to hide "lack of detail inside" (as in missing drivers or such?). And then it has pretty much rock solid 60fps. Also is tiremarks and smoke a factor here?

[Crossbar]

Originally Posted by marcus_rocks

One question for Joker454:
With the advantage of high bandwidth, why is Forza 3 limited with transparencies (cannot see through car glass window) whereas GT5P has abundance of transparencies and sports a 4xMSAA?
I think we cannot point our finger at one single factor and declare the winner. There are many ways to work around the problem.

This is something I would like know as well. How can GT5P on the PS3 pull off 720p with 4xMSAA at 60 fps without EDRAM, still FM2 on the 360 (with EDRAM and "free" MSAA) is limited to 2XMSAA?

Has PD discovered some clever way of working around the limited memory bandwidth of the PS3, that other developers have yet to discover and/or is the tiling required for 4XMSAA on the 360 to "expensive" for 60 fps?

Any theories about this?

[Rotmm]

Originally Posted by marcus_rocks

(cannot see through car glass window)

Huh?

Originally Posted by Crossbar

Any theories about this?

Rolling starts means fewer cars to render on-screen maybe?

[Crossbar]

Originally Posted by Rotmm

Rolling starts means fewer cars to render on-screen maybe?

Given that GT5P allows up to 16 cars races, I don´t think it´s that difficult to get 8 cars on screen at the same time. If I remember correctly GT5P has races that start from stand still as well, but I am not sure that was 16 cars races. Anyone?

[marcus_rocks]

Originally Posted by Rotmm

Huh?

it uses lower resolution buffer for the transparent objects and it uses dark colors to hide the details of the car's interior.

[Arwin]

Originally Posted by Crossbar

Given that GT5P allows up to 16 cars races, I don´t think it´s that difficult to get 8 cars on screen at the same time. If I remember correctly GT5P has races that start from stand still as well, but I am not sure that was 16 cars races. Anyone?

Online races on Suzuka and Fuji start from stand-still and this includes 16 car races.

[Rotmm]

Originally Posted by marcus_rocks

it uses lower resolution buffer for the transparent objects and it uses dark colors to hide the details of the car's interior.

Huh?

http://www.gamersyde.com/stream_forz...-13275_en.html

[Crossbar]

Originally Posted by Arwin

Online races on Suzuka and Fuji start from stand-still and this includes 16 car races.

Thanks, there you go. GT5P even allows more cars on the screen at the same time. Other theories?

[Arwin]

Originally Posted by Crossbar

Thanks, there you go. GT5P even allows more cars on the screen at the same time. Other theories?

They made a certain choice for Forza 2 that isn't optimal, but didn't radically overhaul their graphics engine? That's my theory anyway. Forza 3 looks better than Forza 2, but it's not a fundamental difference that suggests a complete rendering pipeline rewrite, imho.

[Neb]

About Forza 3, how does tiremarks, smoke, vegetation amount and detail factor up regarding transparencies?

[marcus_rocks]

Originally Posted by Rotmm

Huh?

http://www.gamersyde.com/stream_forz...-13275_en.html

This is what I was referring to.