PhysX update due?

[silhrt]

"At the very least, unlike with the GPU camp we should have a clear idea by the start of 2008 if the PhysX hardware is going to take off or not. We expect Ageia will be hanging on for dear life until then."

From a post date of July 25th, 2007 on Where's The Physics: The State of Hardware Accelerated Physics.



I was wondering if you plan on updating this topic.

 

[CP5670]

At the very least, unlike with the GPU camp we should have a clear idea by the start of 2008 if the PhysX hardware is going to take off or not.

The answer to that is an obvious "no" by now.

 

[NanoStuff]

I wonder who the tech guru is who decided paying an extra $200 to play at 11 FPS was a good idea.

Their new $400 ultra edition will give you the option of 6 FPS.

 

[Munky]

I think that decision was more along the lines of "Which component of a PC can we exploit to part gamers with their money."

 

[Syntax Error]

PhysX is nothing more than a money sucking gimmick at this point.

 

[Greenman]

I thought Ageia was already down the tubes. Do they have a viable product?

 

[Piuc2020]

What AGEIA needs is a killer app from a high profile developer that NEEDS the PPU to play, only then will they really have a good real-world application of what the PPU can accomplish and gamers may be more interested then.

I'm actually very interested to see incredible physics-involved games that greatly enhance the experience (check out Force Unleashed to see what I mean only on a grander scale), then a $150 (or whatever the Physx costs) addon wouldn't sound too bad.

 

[NanoStuff]

Originally posted by: Piuc2020
What AGEIA needs is a killer app from a high profile developer that NEEDS the PPU to play,

Never gonna happen. That's suicide unless AGEIA is happy to pay full revenue, and I don't think they even have the money to buy their own lunch.

 

[clickynext]

I think they just gambled on new games quickly growing to need more physics processing power, and the need just hasn't been there so far.

 

[Ryan Smith]

Hi all, your friendly neighborhood AT editor here.

When I wrote that article, we had been expecting something decisive out of UT3 when it came to physics, based on the information that had been released up to that point and on the assumption that UT3 would sell like hot-cakes like every other UT game. What we got was entirely different.

First and foremost, UT3 has tanked sales-wise. If its existence was going to have any effect, with the sales that have actually happened there's no way it would significantly buoy PhysX hardware sales in the short-term.

Second of all was that we were looking for something decisive, instead we got yet another mixed bag. The performance on the AGEIA levels is interesting, the performance on the standard levels not so much. Epic did not integrate PhysX in to the engine as we earlier believed, the result was that we got some interesting results on one map (WAR-OnyxCoast, IIRC) and little elsewhere.

This basically puts us in a stalemate. We're not ready to write off the PPU concept yet, even the now-ancient PhysX-1 hardware produces some big performance boosts in AGEIA's tech demos; if it's used correctly there's promise for the hardware. But no one has used it correctly, CTF-Tornado is about the closest anyone has gotten, and that's another tech demo map.

We keep hearing about AGEIA shipping a 2nd-generation PPU this year. If they do that, it'll buy them a bit of time, hopefully in that time they'll find their killer app now that they've made something that gets the idea right and the APEX platform for Unreal Engine 3 will make physics integration easier for the many UE3 licensees. But if we hit 2009 and there's still no killer app, even if there was new hardware we'd be apt to write off PhysX after 4 years of waiting.

 

[NanoStuff]

Originally posted by: clickynext
I think they just gambled on new games quickly growing to need more physics processing power, and the need just hasn't been there so far.

If the need for dedicated hardware hasn't been there thus far, it will never be there because multi-core processors will fill the position, GPUs in particular which are architecturally apt to such a task. The first ones came out when they were not needed, the second ones will come out when they're obsolete. Physics cards are a dead end, I weep for their investors.

 

[silhrt]

Originally posted by: NanoStuff
If the need for dedicated hardware hasn't been there thus far, it will never be there because multi-core processors will fill the position, GPUs in particular which are architecturally apt to such a task.

What your trying to state is that the CPU's will always be able to handle everything... well thats just not correct. GPU's came about for the same reasoning. CPU's coundnt handle the load put on it. Same will eventually happen for the physics cards too.

The size of the CPU's can only grow so far. And at some point the physics portion WILL grow too large and demanding for the CPU to handle. At that point it will become another part of either the motherboard, graphics card or a physics card. The main problem here is the when portion...

Its sort of like the huge internet boom thing that happened in the 90s. Noone bought into it but now its growing fast.

Right now from what i have seen is that it has great potential but not yet realized by people who sometimes put off trying new technologies. ( some people still hold onto cobalt .... "cause if it aint broke, dont fix it" they say.... )

As of right now... It may be ahead of its time.

P.S. Thanks Ryan for the update.

 

[NanoStuff]

Originally posted by: silhrt
What your trying to state is that the CPU's will always be able to handle everything...

Uh, no. Perhaps you should read that again, I don't know how I could have made it any clearer. I said GPUs are ideal for the task. Programmers could sacrifice a few ALUs for physics processing. No developer is going to rely on having a physics card available because very few, even enthusiast users, will have one. Sure the software could check for an alternate ALU source such as a physics card to offset the work from the GPU, but in such case you might as well either buy a faster GPU or another GPU rather than a card that for no good reason is restricted solely to physics processing. There's just no logic in having a GPU-equivalent floating point processor that is artificially restricted to a certain kind of process, doing nothing when it's not needed, and costing much more per floating point unit than an equivalent GPU. And that's before getting into the complications of having another piece of hardware to deal with, including finding space on the motherboard and compatibility issues.

When developers desire a substantial amount of processing power for physics, GPU developers will fill that need very quickly and far more effectively than AGEIA. They already have actually, so again, all the more head scratching trying to figure out what is left for AGEIA to do other than to milk naive investors.