physx chip

Only thing standing in its way is developer support. It's most likely capable of all that it claims.

You are quite right about that, game desgin has nothing to do with the effect that this chip will do it comes with software for the enabling of the chips effects which in my opinion will help any graphic card perform better by allowing the physx chip to handle all the effects and special effect I think I want one as soon as it comes to the market also the triple- to-go would make a awsome gaming machine.
The pic they have of the effects is so interesting.

I can't find any real specs on the AGEIA PhysX chip, makes me wonder.
whats the power use and how hot is this thing?
is it 130nm or 90nm?
can I OC it and what do we bench it with?
Will this work with GPU based physic or do we get to pic?
What I do know is.
the chip is Over 125 million transistors and 182mm.
It needs power from the psu
After seeing the demos on the AGEIA site I want one.
It's costs isn't that bad at 100 and up.
but what do I get at the low end?
I did find this article on ati's plan http://www.pcper.com/article.php?aid=226 it seems like a go idea if it ever gets here. regardless it's a go read, just to learn more on how it works. I would love to upgrade and keep my x1800 as a physic processor. save some cash and know I can OC it.

This was discussed in another thread. Nvidia actually is looking into 'emulating' this system in its SLI solutions (instead of having 2 cards rendering the scene, one would compute physics and the other do the actual rendering).

Still, Ageia's solution has that little interesting bit about only requiring a PCI slot... Or better, a single PCI-eX (single line) slot...

this could be a new start of something like the voodoo and voodoo2 cards where you had your main video card, and an extra tag along card, but I think Nvidia and ATi might do a better job tho cause of driver support and the fact that SLI/Crossfire is already implemented.

Guys isn't it possible for NVIDIA or ATI to incorporate a physx chip in their upcoming graphics cards so that we all avoided buying a separate one physx card?Would this be technically impossible?What do u think?

Actually its not a performance enhancer it is a effect chip the pic looks awsome, it is designed to enhance the effect of the current graphic effects. with this chip installed one can go to a rock and pic it up or what ever else is in the world, the new cards nvidia is global rendering and wont be allowed by law to use this desgin with out buying a share of the supply at any cost this will greatly improve the graphics for all.

Think about this you are playing your best fisrt person shooter, blaster your buddys head off and than pick it up and tossed it.

Dude thanks for the pics and links I got so excited seeing the effects I felt like I was actully in the game.

I am quite sure they are working on what your are speaking about I hope this is as good as it sounds if amd nvidia ati do this we all will see gaming at its best specially with global rendering coming in for the info go here so real dudes its comming to us
http://www.ageia.com/products/physx.html

Thanks for clearing that up mitch sort of sounds like a math coprocessorl

Buying an additional card which costs over 100$ is just plain silly.
Nvidia and ATI will eventually implement physics in theyr GPUs(some say that they can already,but the software development hasn't been totally finalized yet)and some software developers are offering theyr help...like NVidia working with the Havok FX.
So I'll wait for that instead of spending money on a card that will probably become useless. GPUs are already powerfull enough and the raw calculations that they can do compared to a CPU is amazing.
More and more technologies which are much or less scams in my oppionions to atract people from the gaming industry...I already read something about network cards that will reduce lags in games....lol...try getting a bigger bandwidth instead,to see improvements.
As for physx ...from my point of view the Havok physical engine is much more flexible,powerfull and can do better things in a more cost efficient way.

It seems theres two diffrent ways to do this, the gpu is most likely the less efficient one. I just wonder if it can make up for it in pure power. For me and alot of us the gpu may be a better option, we all have a vid card.
We really need a tool to see which one is the best way for us to go. The agiea site is all hype with very little real info and no specs. It's easy to say your the best when no one else is in the game. Now the big boys are in the game and they have the money and R&D behide them. I think the agiea is a great leap in gaming but why so tight lipped on the specs. All said cellfactor, looks killer and thats really what will sell it.

I think we need toms tech crew to take a look for us can a mod email them on this, myself and I'm sure others do too. This is why I posted too good to be true

http://www.xbitlabs.com/news/multi [...] 24421.html
or
http://www.havok.com/content/view/187/77/
And I am not only refering to SLI setups...it's the same thing as in video quality...want to run a game at over 100fps at high resolutions...buy a SLI configuration...if you're more than satisfied running a game at 1024 at a decent frame rate you can stick to a single card. Want to use the last drop of procesing power from your GPU why not enable physics support.
And usually people who can aford throwing 100$ or more just for physics performance have the resources to buy a SLI configuration.
I prefer Havok's aproach on this topic instead of AGEIA...which requires buying new hardware...the thing is that the diferences , and I mean ,performance wise from a X1900XT and a 7900XT are small....while the X1900 has 48 pixel pipelines and the Geforce has 24...so what happens to all the raw power from the X1900XT?!...Wouldn't it be great if you could use it to power the physics engine also.
I think this opens up very interesting oportunities for the GPU...

Like mitch said not everyone has sli or pci x or e so the physx chip would make a difference specially for us who dont plan on upgrading untill the newer generations are well in production.

Oh crap... didn't mean to post this twice... I've got a ton of tabs open, and one had this post waiting to be submitted... so I did... again. I am such a dumbass.

To ChipDeath: I'm sorry you don't get it... you're missing out.

LOL

I'm sure you didn't mean it, but this is a GREAT play on words.
FPU=precise calculations (due to the floating point)
FPU=what is being utilized to calculate the physics by the CPU

sorry couldn't help myself
-mcg

Sure would be a nice feature actually I found something on the final word
http://www.anandtech.com/cpuchipse [...] i=2376&p=3
http://www.bfgtech.com/physx/index.htm

Yeah i know that but its the same scheme as a tag along card.

Not knowing much about the physx chip myself, but I wonder how it would work with Prime95 or other distrubuted networking projects? Would it be possible to offload some work to it so when I'm not playing games I can find the next world record prime?

Just a random thought, we all of course know why we really need a supercomputer's equivalent of computational power on our desktops; that is of course to play games!

Bear in mind my system can do 16 GFLOPS, and that is a peak figure really, over 4 x 2 GHz AMD64 K8 cores (2 x Dual-Core Opteron 270). While using 175+ watts of power for the 2 CPUs at peak load.

A www.clearspeed.com card can do 50 GFLOPs (over 3x) using only 25 watts (1/7th the power consumption). When a given task is offloaded from the CPU (A General Purpose Processing Unit really) to a dedicated Hardware Processing Unit optimised for the given task, you typically get 10 - 20 times the performance at the same, or less, power consumption. (And at the same power consumption in the above scenario we're looking at 21x the mathematical processing power, available, today).

For CPUs to get 21x more power at the same power level would take about 5 - 8 years. (Less for higher end systems, Longer for more mainstream ones). Bearing in mind you're gaining far more than this as once installed the CPU is free, or can also execute 'less complex' functions in tandem, so it is really pushing us 10-16 years forward by using dedicated hardware for a given task. (We'd still be running games with poor detail, at 800x600 or less, if all rendering was done by the processors. My 4-way system would render scenes at 5 fps or less, that my far cheaper video card can render at 80+ fps in higher resolution and detail.... the video card is cheaper than 20 processors, thus it works ).

This frees the CPUs do to execute other code. (Basically run logic loops).

In some cases we are looking at simulation engines running at 2 fps (or cycles per second really at that speed), compared to running at 60-80 fps with the newer hardware.

That is a significant leap forward, and would help with the move to 1024+ player servers, with large island combat, and accurate physics and ballistics. As it is consolidated, uses less power, is affordable (more affordable than 20 processors and easier to code for) it will move us forward in this respect (all over again, just like audio did, just like video cards did, and now... physics acceleration..... next may be dedicated AI processors.... etc).

The CPU is really only 5-10% efficient at executing such things, as it is a general purpose processor, compared to a fully optimized design on the same die size for a given function or collection of similar functions which will be 75%+ efficient at just that task, and far less efficient at 'general purpose' code execution. Although video card GPU / VPU can be used for other tasks, but only ones that are 'similar' in nature.

In 5 - 10 years time the delta between the performance of CPU core(s) executing 'code' and accelerated [function of choice here] hardware processors will be even greater, and we'll all be thankful it happened.

I doubt I could design a processor alone.

The interconnects and 'overall' implementation perhaps.... possible maybe, likely no.

But not a CPU (or anything similar).

It would take 50 lifetimes of human engineering to design such a device, and likely 100's of scientists and engineers working on such a project to have it complete within 5 to 10 years.

Bill Gates, Peter Norton, etc (all software I know, but you get my drift) didn't become rich people overnight, or by working alone.

good choice of word {doubt} if it was can't I would debate that till I turned green.

Well, if you believe my Computer Architecture II professor...

we'll eventually beguin buying FPGA's and IP of this or that card, "flash" it on the FPGA and let it roll.

I don't believe we're quite there yet, but seeing the cyclic nature of technology trends, I don't think it will come to pass.

For those saying that HavokFX is a better way, I think its a compromise. Sure you can get the "effects". But HavocFX is like a $300,000 invesment on a new game, while the PhysX is free of charge for implementation. Also, GPU's are geard toward rendering and display, and while they can perform some pretty intensive FP calculations, they can't match an architecture tailored towards Physical calculations, it escapes their designs, as simple as that.

Can it be enabled on GPU's? Sure. Can it be "theoretically" as fast as a dedicated PPU? Sorry but no at the same price point. I would be really funny for somebody to plunk down $450 for a new GPU, and use their "old" $450 GPU dedicated for some physics calculatios, that can be rivaled or outperformed, or maybe just outdone or be comparable, to a $100 card

Anyway, since in theory things are easy to "dissect" and in practice things are quite muddy (as in not transparent), I believe that the dedicated physics processors are a new piece of hardware "braggin' rights", but if they could help CivIV to run faster... they're welcome

That's what I like to know too!

Please somebody answer to this

Ati and NV both claim the GPU is well suited to do physics and NV put up some numbers that look good.
I have a x1800, in six months i may upgrade. I'm not buying a card to do physics just keeping the card I have[ and it's still a vid card if I need it ].
It's new teck, give it a year to work out the bugs and find it's place. If ATI /NV wanted to they could crush it. In a few months they could have a ppu on the market. They know how to run it on their gpus so they know what it takes to do it right. I don't think they want to though so give it some time and they may work with it to really do some crazy stuff.
Like I said though if it's that good I'm sure I'll get one but not the $100 one if it's like gpus it will suck. I just want to see some charts first before praise it as the best thing since 3D. I want good physics and don't care who I get it.