| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Comments:
<0> 1 thread for rendering and it doesn't max out the usage?
<1> I said 100% of core #1
<1> 100% = max no? :)
<0> EwIck says 74
<1> I suppose he's talking about overall
<2> the non-rendering threads actually
<2> you'd have to ask my guy for the deep technical details though
<2> he's brilliant
<1> Where did you get that 74% from?
<2> profiling tools that come with the 360
<3> this is a good example of Cray's business model. Which is build the hardware to fit the software rather than vice versa.
<1> Not that much, I got rendering multithreaded ;)
<1> I don't have physics though :(
<1> Well, I have a namespace called physics but it's empty :D
<0> lol
<2> just buy havok, or something similar
<2> a lot of work saved
<0> yeah, everybody's doing it
<1> I'm in contact with Ageia
<1> But I don't want to limit myself
<1> Thing is Ageia is more "accessible" than Havok
<1> Thanks to all those companies we'll go back to 1995 when it was Glide vs OpenGL vs Direct3D
<1> And some day Microsoft will make a DirectPhysics (when they start to care)
<0> hehe
<2> ouch. Glide.
<2> I had finally forgotten about it, thanks a lot
<3> how does one even develop for console systems?
<1> With the development kit
<3> I have no idea how that works
<1> Heheh
<0> someone at the GDC (I think it was the Havok guys) showed off a physics simulation running all calculations on a GPU
<2> Twister: you buy the dev kit and start working
<3> So, you burn the source and test it per each dev iteration?
<4> you pay a rediculous amount of money and bribe people to get approved for a dev kit
<1> rdragon: yes, that's Havok
<1> rdragon: Havok use shaders (partnered with nVIDIA), Ageia uses a PCI card, ATI announced they'll kick both ***es but they haven't released anything
<0> ah
<2> Twister: no you don'T need to burn a CD/DVD during development
<1> That'd be evil
<2> CD protection is disabled on dev kits, and you just put your stuff there on the hard drive
<1> Each time you compile you burn a CD heheh
<3> thats what I meant heh
<2> yeah.. I had to do that with EEPROMs for a while
<1> "oops I forgot to type a line => compile => burn (5 minutes) => run"
<2> before my boss finally purchased an eeprom emulator
<5> who do you work for, if you don't mind me asking
<3> for people like me who compile like once every 5 lines of code, that would not be cool ;)
<1> F7 => compile => upload through network => run
<1> Yay
<1> (well, for Xbox anyway)
<1> Others are a bit more... painful
<3> ah, uploading would not be bad
<2> well, that's how it is
<1> Especially when you have to use Codewarrior
<2> you just upload your stuff and run
<1> And everybody knows Codewarrior is evil
<0> I don't care for the idea of buying a physics card
<4> who's developing a physics card? ive thought that would be a neat idea for a while
<1> Ageia
<2> it's not a particularly bad idea
<4> any developers actually supporting it?
<1> There's a downside to it
<1> Graphics is not that much part of the gameplay
<1> Physics can be
<1> If all players don't have the card/shaders (for Havok)
<1> What do you do?
<4> I suspect doing physics on the GPU would be more efficient in terms of cost to the end user though
<5> software physics :P
<2> software?
<1> Can you still do software rendering as fast as hardware?
<1> (with the same end result)
<2> didn't we go through the same thing when 3d cards first were introduced?
<5> and yea, the GPU should just be replaced with a general purpose float vector math card
<1> The GPU is already a general purpose float vector math thing
<1> That's why Havok can run on it in fact heheh
<5> not as general purpose as i'd like it to be.
<6> I don't agree the GPU should be replaced.
<1> Which shader version did you use?
<6> However, I doubt there should be a "physic" component to the GPU.
<5> i for one can't figure out how to even multiply two float vectors and get the product vector back out.
<1> That's just mul dest, src0, src1
<5> _m_ : it's almost the same thing, functionally.
<6> Such a component would better be placed on a separate card.
<5> physics, graphics, any number of other uses....
<5> all you're doing is absurd amounts of floating point operations.
<6> NineVolt: yes, it is. However, the GPU has better access to the video RAM.
<5> true.
<3> have one or more specialized processors for each type of task, and let the compilers decide which one best handles the job?
<6> I doubt such access would make sense for that physic component.
<1> EwIck: if the guy playing Counterstrike was playing software and another guy playing with hardware acceleration, they can play just the same way; with physics, if it's part of the gameplay, that just won't work, if one of them can't do the same physics simulation
<5> it wouldn't hurt.
<1> So it's not general purpore enough but you didn't find the multiply
<1> purpose too
<1> Then I can understand why you feel it's not general purpose
<1> There's add, sub and rcp too
<5> i thought the physics in CS was server-side
<1> If it's server side why would you need a physics card
<2> ... for the server?
<5> heh :P
<1> Physics runs client side in CS
<1> For prediction
<6> The physics of splatter could be done on the client side :)
<2> but CS is gaey
<1> True, CS is for nazi 14 years old germans
<5> i wasn't the one that brought it up.
<1> So player #1 plays TheSupahGame, he's got this brand new PhysX card that's <marketing>100000000 more powerful than a Pentium 4</marketing>
<1> TheSupahGame's gameplay is all about physics
<1> Other people can't play with him if they don't have the card
<1> Cause they can't run the simulation
<5> they can, just at a really slow speed. in software.
<7> Sure they can... they just get significantly lower FPS.
<1> Unless they get 100000000 Pentium 4s I guess
<1> Physics don't affect FPS
<5> same argument as 3d cards.
<1> And 3d cards don't affect the gameplay
<5> ...
<1> Having high resolution textures don't change anything to the gameplay
<5> so playing at 1fps, your gameplay isn't affected.
<5> right.
<1> Frames per second are not part of a game
<5> people with high end gaming systems don't have any sort of advantage against people playing at 1fps.
<1> You're trying to sound like a moron again or something?
<1> A box falls
<1> Physics handles it
<1> In a very smart way
<1> Cause the card's so powerful
<1> The other guy has software physics
<1> It's too slow to run the same simulation
<1> It can't know where the box will go
<5> it can get handled the -exact- same way.
<1> That's gameplay
<1> No they can't
<5> yes it can.
<5> !!?!
<5> why not?
<1> Just like software rendering can't do what hardware rendering does
<1> In realtime
<5> is there some functionality in the physics card that can not be emulated in software?
<5> no.
<5> software rendering -can-.
<5> it just takes a lot longer.
<1> Nope
<5> meaning you run at 1fps.
<5> yes. i ***ure you.
<1> Pixel shaders
<1> You run at 0.0000000000001fps
<1> So it can't be emulated
<5> EXACTLY>
<1> Eventually you'll get it
<5> you just contradicted yourself.
Return to
#c++
or
Go to some related
logs:
#skype
www-€6-kont
#windows
#chatzone
miss tiina
#chatzone
ghbhgbj
ubuntu autoreboot
#c++
openwrt fail2ban
|
|