Oh bummer. Is ikanexpress available ? j/k j/k

On the API side of things, I vaguely remember seeing something fomr the MS front on this; would this be the DirectPhysics thing you mentioned above ? Any insider info ? It's not in DX10.1 for sure.

well.. it was rumored to be included in DX11..

When it just came out, it was PCI
Not sure if they now have PCI-E 1x or not

For the graphic card one, it's Havok FX, supposedly using graphic card's GPU to calculate physics simulation. But HavokFX is an API that developers need to pay to use (they can't make money out of hardware because it uses graphic card), so I don't think there's any game out there using HavokFX.

PhysX is the PPU card that users need to buy, but their API is free for developers, and there are some games out there that supports it.

This post has been edited by Hornet: Nov 22 2007, 12:31 PM

MS was/is hiring people for "DirectPhysics", and job description in no uncertain terms said that GPU acceleration was a part of it. And it will probably use a different hook than the regular gfx programs, since that is generally subject to pretty aggressive optimizations that work for graphics but not physics.

---

Added on November 22, 2007, 2:34 pmFrom AMD and Nvidia's (negative) responses to intel's purchase of Havok, i think HavokFX was pretty far along already.

This post has been edited by ikanayam: Nov 22 2007, 02:34 PM

The name said it all. Physic calculations for gaming purpose only.

yup that it
i give u simple example
have u ever play nfs
u go for the highest speed then when in the tunnel suddenly got accident then fly over the tunnel
the card will help the effect on the game 1

Well, to ATi's credit even with CTM's inane level of complexity with ASM, it does support Brook GPU

sorry but... huh?

No, Brook is working on support for it, not the other way around. Brook+ can compile to CAL, which is a higher level abstraction of CTM, but still no one will bother coding with that. ATI made a huge mistake by not providing a high level language for GPGPU, and nvidia basically beat them to the punch and is now even teaching classes at universities about CUDA. No point having good hardware without good software support.

As i recall from reading one of Mike Houston's posts, ATI is working with them (i guess Stanford always had a good partnership with ATi), and if anyone's willing to put the extra time into coding (barring finding people who still actually code in ASM and the tenacity to do so) the raw end performance stands to gain a lot. Anyways, i'll cede that Nvidia has the upper hand here, heck we don't program in machine code anymore for good reasons, but if the basic history of programming language has anything to show, it isn't that hard, the only hard part being standardization and winning over enough people. Last if not least, i wouldn't fret over it too much if ATI has realized their mistake, as long as it compiles and gives you good perf and the results you need, people won't be too picky.

Well, nvidia has had CUDA out for a while now. Many top universities are going to even start teaching CUDA soon, if they haven't already, including Stanford. And if Brook+ is anything like Brook... i'm not sure it's going to be popular. Coding in ASM is already hard enough for CPUs, where the instruction set is the same (x86), coding in asm for GPUs... not so much. As it is the parallel programming concept is already hard enough for most people, parallel programming in ASM would just be more excitement than most people need. Plus, compilers are pretty smart these days, as long as you write decent code, they should do a decent job. For performance critical apps, you write the general code in a high level language then tune the critical parts in ASM. No one writes everything from scratch in ASM anymore, it's impossible to debug and it just takes too long for anything substantial.

True, True, trying to find a bug in an ASM program is like asking for your ass you be handed to you on a silver platter.

But i've heard news that ATI is working on a compiler and C like language of it's own, ok, abit late, but better late than never. But i do see a use for CTM though, as it allows people to get inside the belly of the beast and understand it's inner workings, yes, you don't use it as a practical means of programming but it does wonders in the education department.

That language is called Brook+, and it's not publicly available yet, so can't say what it looks like. But as i said, if it's anything like Brook, it probably won't catch on. And it's not an internal ATI project either. They're still depending on the Brook guys somewhat. Not a good plan if you ask me.

Hmmm, i seriously thought Brook+ and the AMD C like initiative were separate entities. True, depending on outside help which you don't directly control isn't exactly something geared for speed or fast deliverables.

AMD among few companies that considers buying over Ageia
http://www.custompc.co.uk/news/601680/amd-...geia/page1.html


kind of strange with the financial situation AMD is in (though it may benefits them in the future, like integrate PPU in their CPU or motherboard or so?)
It certainly won't be cheap

There's only one way for this card to sell

Release a kickass game that has a VAST difference in experience with the physics card. IMHO its an old technology so not worth it

Is this Ageia physx have to be paired with graphic card for it to function?

Meh, compared to how much AMD has been losing another 50 million won't change much.

I don't think just 50 million. I saw a news (rumor) saying it is 100million. This is huge amount.

Compared to >2 billion? That's nothing lol.