Added: 2 years ago
From: DarkWaterDev
Views: 118,901
Sort by time | Sort by thread (beta)

Link to this comment:

Share to:
see all

All Comments (61)

Sign In or Sign Up now to post a comment!
  • lol locust of planes :D

  • please add collision then see who wins

  • Impressive, most impressive.

  • The Designer of This Software Pokes His Head Out Of His Office Door and Says:

    "Someone call Blizzard and tell em to be waiting at their office with a blank check."

  • UP Tyrone... ! not....

  • There's... To many of them!

  • Zzzzz

  • I jizzed in my pants.... four times.

  • What accent was that? That was possibly the most annoying voice I have ever heard.

  • Im with you on that.

  • Norn Iron

    (i.e. Northern Ireland)

    LOVED the accent, btw.

  • lmao

  • They came from- BEHIND!

  • ...no.

  • That would make an awesome rts.

  • Hey is this that software development company that is based underneath Supervalu in Derry?

  • carrier RUSH!!!!

  • Excellent work!

  • This demonstrates collision avoidance only - is it technically possible to use CUDA for calculating actual collisions and let planes crash into the scenery and into each other and fall to the ground? How would this affect the performance?

  • It is generally possible but conceptually a different things than the demo. NVidia is accelerating PhysX with CUDA, which is basically what you ask for. Performance should be generally much less as the problem can't be parallelized as much.

  • Cuda rocks. For this operations we need cup of CPU's but CUDA + GPU = Sick crazy methods

  • but does it play Crysis 2?

  • you obviously missed the point

  • Do you really test the position of every plane against the position of every other plane? An octree-like data structure would be so much more efficient. Is it really right to go with a less efficient algorithm just to be able to do more things in parallel (or just because CUDA evaluates 'if' branches sequentially)?

  • The showcase demonstrates that you can have computation heavy workloads on the GPU without sacrificing rendering quality too much. You can easily optimize the brute force algorithm used and spend the free cycles for other tasks, but that's a different story.

  • Yes. Another example: Almost all of the processing time in Raytracing is spent on primitive intersections. You can speed it up significantly with a KD-Tree with a depth that easily exceeds 20. But when you do it in CUDA, you should use a flat tree (like 4 or 5 levels) and leave the rest to pure parallel number chrunching power. It's faster that way.

  • Could you have a hidden, separate, simplified collision map to reduce processing? Also, would having the planes only refer to others in their immediate vicinity increase processing or ultimately reduce it with that many planes?

  • They are Hiring!

  • Very cool! Making a really good SDK for CUDA accelerated AI could be a very lucrative and beneficial product!

    The first game that comes to mind that could benefit from really good AI is Thief 4 - maybe you guys could help them with that (might get the game released sooner!) ;)

  • Well, "mass AI stuff" ....... what about individual behaviour ?! I mean ... having a nice ammount of airplanes flying arround you is kinda nice, but is there a possibility that the closest obejcts to the player instantly switch their "mass behaviour" to a more interactive one, for example "hunting" the players airplane in formations, performing coordinated attacks etc. ?

    If your engine would be able to do so, i'd be a very nice technology for big battles, i totally agree.

  • This example looks very promising. I understand the overkill of planes used to demonstrate and calculate the possibilities. Very impressive. In the end, if you develop a WWI flying sim with proper historical planes, realistic flight models for each along with proper gun mounts and damage, you will have an instant audience. There are many WWI historical enthusiasts that have been anxiously awaiting a proper WWI flying sim. Last one was RB3D, which had to be heavily modded for AI, FM's and damage.

  • This could really turn things around for Flock of Seagulls.

  • Also this AI GPU example is not pratical for planes. Never so a 4000 plane airbattle.

    But a locus plage they could do miljoens of.

    Or wild bad's flying out a bad cave. They behave more like this example. In a chaotic way.

  • 4000 planes would be really over the top! But if you add rockets and projectiles (and maybe birds) and increase the "intelligence" of the planes you won't have so many onscreen anymore.

  • I asume they use a GTX285 or GTX275

    Not a 8800GT.

    cuda is a C like dialect specific for nV stream proc with nV specific instruction set. Like x86 AMD65 68000 are instruction.

    Cuda is also used as base for GPGPU part of OpenCL.

    So this is posible with any GPGPU only if nV used OpenCL Wich ATI also support and iNtel.

    And with that give PhysX a harder push if they drop cuda and go for OpenCL.

  • We used follwowing cards for development and testing (Average fps in brackets): 8800GTS (35fps), 9800GTX (55fps) and a GTX260 (45fps)

  • How do u get less fps on gtx260 than 9800gtx?

  • The demo was developed on the 9800 and tested in parallel on 8800. We got the 260 after the project was finished, so we actually only measured on 260 but never really tried to find out where the difference is from.

  • The tech demo looks cool, but I am wondering which ''consumer level single nV GPU'' has been used?

  • Wow. Simply amazing! Not much more to say. Excellent work!

  • Its shows GPGPU potential but I wouldn't call it ai. To be fair though GPGPU will be great for true ai and this is a creative use of its potential much like physx.

  • The demo uses steering behaviours to control the planes which is one AI approach for controlling character movement. You're right that it is not a ready to use, complete solution for a game.

  • Nice, but it seems there isn´t any horsepower left for today´s good graphics.

  • I must admit that the we haven't spent too much time on polishing the artwork for the demo. Don't blame it on the graphics engine. Sorry about that!

  • Oh well, you´re a nice and very capable person, I don´t even deserve your apologies. I´m just a litte frustrated because at this pace, we will need several graphic cards on our pc: one for 3d, one for physics, another one for ai... :P cheers

  • This truly is an impressive technical demo of what can be done nowadays.

    Even if this means getting the most expensive hardware you can get. It's possible! thats an amazing news for gamedevs starting a 2012+ game.

  • Great stuff, although it's sometimes hard to figure out what is being said :)

  • hahha daaaamnn!!!! i really love to see starcraft 2 protoss carrier handling a friggin 4000 interceptors like swarming up on the terran colony!!! yeeeehhhhaaa!!! a true jaw dropping scene!!! XD

    pls Blizzard make good use of this!!! XD

  • In most gameplay scenarios like this instancing would work fine and would be less intensive. How exactly are you going to fight 4000 AI enemies at the same time.

    More impressive would be if you could take that same 4000 AIs and apply it to make 10 brilliant AI driven characters.

  • It depends on the game. I would love to mow down 4000 planes that try to evade my imprevilous stream of high-velocity explosive projectiles. Or a horde of slimy aliens. With polygonized bloodsplatter instead of the dry puff of a textured polygon we have these days.

  • We are using instancing for rendering, but not for steering (which wouldn't work here). You're right that a game mode with fighting 4000 planes doesn't make sense and has never been intended. But the number of planes we can show at max allows us to estimate how many computational expensive blocks can be executed per frame. How these computations are applied in a game is a different story.

  • Wow, thanks for the reply. I hadn't considered that possibility, interesting.

  • Looks like a swarm of insects

  • Thats insane! The sheer amount of processing power packed into GPU`s nowadays is incredible, i cant wait for GPGPU to go mainstream!

  • Amazing stuff!

  • so awesome

Loading...
0 / 00Unsaved Playlist Return to active list
    1. Your queue is empty. Add videos to your queue using this button:
      or sign in to load a different list.
    Loading...Loading...Saving...
    • Clear all videos from this list
    • Learn more