Havok says GPUs are fine for physics

[Ubermann]

Havok vice president Jeff Yates talks about why the GPU, not the PPU, is the better option for physics processing.

Q&A:
http://www.gamespot.com/news/6136639.html

[DilTech]

Havok has reason to make this claim...

If companies go with the PhysX ppu, then that means they won't be using the Havok physics engine. Considering Havok's money all comes FROM that engine, they have all the reason in to world to say we don't need to use a PPU....

Now, if game developers like Epic make this statement, then we'll have reason to believe it. Currently though, I say it's just them trying to persuade people not to go with Ageia.

[Ubermann]

Future will tell i guess.
If GPU can do it then fine with me, i dont have to spend alot of cash on a PPU.

[einCe]

using a new Shader Model 3 graphics card, we think there is definitely room for basic physics effects with no appreciable impact on the graphics.
isnt the purpose of a ppu to use an amazing level of physics?
i think i remember something like 25,000 objects at once or something

[DilTech]

Yes eince, you're correct...

Basically, havok wants to run basic physics on the videocard, and major ones still on the cpu. There's just no way that could keep up with the physx chip.

[Ubermann]

Wish we had some numbers what a GPU can perform so we could compare.

[J-Mag]

All I know is that even with dual 7800GTX's, I don't have any spare cycles so I would essentially have to reduce resoltion or settings in order to allow physics to be calculated. SCREW THAT

I want a dedicated PPU.

[alexio]

We just need multicore support in game engines/games. A dedicated PPU is nice untill we are seeing more powerfull cpu's (quad-core, etc) and games supporting muliple cores.

[MaxxxRacer]

You guys all know whats funny?

The PhysX board uses a nvidia GPU.. atleast the first proto's did.

[IYP]

i want my independant PPU, in logical sense, whatever the GPU can calculate the PPU can do more....just like SLI....the single card can handle stuff....but SLI can handle more stuff

[texuspete00]

Wow, with all that power to burn in FEAR this sounds awesome! :rolleyes:

[nn_step]

Wow It Sounds just like what I heard about the CPU when the GPU came out :rolleyes:
It's doable but it isn't extremely cost effective...

[Entity_Razer]

I to believe PPU is the way to go. Simply because this free's up GPU resources and will result in better performance.

It'd be stupid to have a PPU ona GPU and every time you upgrade your GPU you need to pay more for this PPU feature which is basicly a software based Physics engine. While with a PPU card you could just keep the card and spend a little more on a better GPU. Makes perfect sence + is cost effective. Not to mention a PPU card is also good for the multimedia nuts amongst us like me who like to make shortmovies or modeling in programs to make movies with those. graphical designers could make their demo reels and their overall quality of work a lot better when using a PPU not to mention it will cut down rendering times.

[slider99]

PPUs all the way. Havok might soon be just another havoc.

[Ubermann]

http://www.havok.com/%20index.php?option=com_content&task=view&id=187&Itemid=77

[Tim]

They will probably only delay the inevitable... :lsfight:

let's say that a PPU will last quite some time let's say 2,5 years...what if all that logic has to be implemented on a GFX card? That would mean more money if you buy a new one....if you are upgrading that means 2-3 videocards in 2,5 year (if you are like me upgrading once a year or so)

But if you had a PPU you wouldn't need that....I dunno sometimes I don't make sense...never mind :p:

[s81o]

surely ati and nvidia could add a ppu to future graphics cards :)

[Thorry]

surely ati and nvidia could add a ppu to future graphics cards :)

They could but the intended life-cycle of physics cards and the projected future life-cycle of graphic cards don't match up, so they prolly won't put them on there. Most likely they will add some extra pipelines designed for physics so people without a physics card would still have a little bit of the same, without becoming cost-ineffective.

[slider99]

Why must the life cycle of a PPU be so long?

I mean, why won't there be a new PPU out every 6-12 months? Just as in graphics, with more pipeilines, ram, clock speeds etc... able to handle more and more physics (particles, object, whatever) per second? A higher physics "fill rate"? Just wondering? It shouldn't be as dull as sound cards? :confused:

[Ubermann]

It can handle 25.000 object and i think that is enough for awhile =)

[Tim]

It can handle 25.000 object and i think that is enough for awhile =)

lol yeah I was just going to say that :D :D :slobber:

[Ugly n Grey]

DO you really think that the vid card makers are going to support someone wedging into their territory even if it is on an angle? I don't expect any support from ATI or Nvidia for anything to do with games beyond their own cards and AMD/Intels CPU's. I even expect it to be a case of drivers might somehow happen to conflict with a physics cards.

Just me looking at the industry with 2 dominant players in CPU's and GPU's, any competion won't last long. I expect any product that is not made by one of the big four I named to be an absolete and useless product as soon as the named company goes bankrupt. Which will be rapidly.

Spend you money on Intel shares, you will be happier.

[Frisch]

DO you really think that the vid card makers are going to support someone wedging into their territory even if it is on an angle? I don't expect any support from ATI or Nvidia for anything to do with games beyond their own cards and AMD/Intels CPU's. I even expect it to be a case of drivers might somehow happen to conflict with a physics cards.

Just me looking at the industry with 2 dominant players in CPU's and GPU's, any competion won't last long. I expect any product that is not made by one of the big four I named to be an absolete and useless product as soon as the named company goes bankrupt. Which will be rapidly.

Spend you money on Intel shares, you will be happier.

The power of Software.

[Thorry]

Why must the life cycle of a PPU be so long?

I mean, why won't there be a new PPU out every 6-12 months? Just as in graphics, with more pipeilines, ram, clock speeds etc... able to handle more and more physics (particles, object, whatever) per second? A higher physics "fill rate"? Just wondering? It shouldn't be as dull as sound cards? :confused:

Very easy, it takes time for developers to integrate the physics into their games. It isn't a 9th generation engine like Direct-X or OpenGL, it's brand new.

When the first physics cards are released the engine will be full of bugs and games will be slow to adopt it. After maybe 2-3 years a next generation will be developed. Then after 1 year another.

Just because GPU's now have a life cycle of 5-6 months doesn't mean they always have. When GPU's were brand new the life cycle also was very long.

Besides, there is only so much you can do with physics, to integrate new features into the engine is a tough job. Also you need to provide backwards compatibility. Because the techniques used are borrowed a great deal from GPUs the first products will be of such a level they will provide enough performance for the first couple of years.

[JuanFlaiter]

All modern GPU's are programable, AGEIA's first Physix boards looked like a GF3. ATI or NV can implement them as soon as the want, just compare the size of the companies, do you think the big whales would let a new company enter their market?