PhysX layer running on AMD Radeon 3870, utility available "soon"

As if overnight, Eran Badit of NGOHQ.com has PhysX running on the AMD Radeon 3870. Badit said that the hack was "easy," and NGOHQ.com will distribute the utility after a bit more testing. As for performance, he hit a 22,606 CPU score in 3D mark Vantage, which is nothing to sneeze at. He swears that AMD isn't involved in any of this, and that the utility release will be entirely independent.

[Via TGDaily]

[LC @ Jun 29th 2008 2:12AM]

Which is nothing to sneeze at?!?!

5Ghz Quads get about that much. While it is about the same compared to a Nvidia card running PhysX it's amazing compared to CPUs.

[Bunson @ Jun 29th 2008 2:21AM]

Impressive... Most impressive.

[ByronGman @ Jun 29th 2008 2:34AM]

Yeah, all the benchmarking is impressive but I installed PhysX today and didn't get anything except a hotter card with the same performance....

Of course, I do have a laptop and 8600m GT and I heard somewhere that PhysX was only useful for Desktop cards....

Can someone set me straight on all this???

[DefPo3t @ Jun 29th 2008 2:48AM]

8600 is your problem only for 8800 cards and up

[TavisJohn @ Jun 29th 2008 2:59AM]

I thought that it was going to be for all Nvidia 8000 series cards...

If it is for the 8800 and up, than that just sucks!

[Daza @ Jun 29th 2008 3:09AM]

This might sound like bs, but as I was reading "This is nothing to sneeze at", I actually launched a big wet sneeze.

I'm not reading your posts any more, might make me do other things you write =/

[tha-don @ Jun 29th 2008 3:19AM]

@Daza

at least they didn't say "which is nothing to crap your pants at."

[DefPo3t @ Jun 29th 2008 3:27AM]

@bryon after looking into it your card should be supported with a future firmware update
http://www.nvidia.com/object/cuda_learn_products.html
all of these theoretically should be able to since the physx coding layer is based on cuda

[E71 @ Jun 29th 2008 9:30AM]

So many redundant cores on our single/dual quad-core setups yet we're putting this PhysX crap in there to boost physics...

[shadowarmy75 @ Jun 29th 2008 1:24PM]

I sneezed also.

[helloUser @ Jun 29th 2008 2:33AM]

.....anyone know if this sort of thing will be available for OS X (or, Snow Leopard?)

[DefPo3t @ Jun 29th 2008 2:46AM]

yeah sory but apple doesnt let you use the required cards that are required.

[DefPo3t @ Jun 29th 2008 2:47AM]

"required cards that are required" gawd i suck at grammer

[Pochi @ Jun 29th 2008 2:50AM]

The required cards that are required.

lol.

[phanbouy @ Jun 29th 2008 2:56AM]

It's the art of fighting without fighting.

[heffeque @ Jun 29th 2008 6:03AM]

8600m GT is Cuda compatible, and so is he 8800 for Mac Pros.

[DefPo3t @ Jun 29th 2008 6:26AM]

OSX isnot cuda compatible

[thethirdmoose @ Jun 29th 2008 12:08PM]

department of redundancy department

[DWells55 @ Jun 29th 2008 12:50PM]

Obvious statement is obvious.

[Hamidxa @ Jun 29th 2008 1:40PM]

lol...good one.

Apple users wont be seeing this for God knows how long, 1+ year at best.

Even then, so what?
You want to enable physics in Quake 3? Zork perhaps?

[loosely_coupled @ Jun 30th 2008 12:42AM]

CUDA is the name for Nvidia's C language extensions and SDK for programming their GPUs. AMD has a similar system. I'm not sure if the CUDA SDK is currently available for OSX.

Apple's announced OpenCL framework is going to provide an abstraction layer over the top of both nVidia's and AMD's system so future OSX Snow Leopard software is able to take advantage of the GPU (and multi-core CPUs) for parallel processing. By Integrating this functionality directly into the operating system as an API, Apple is making it MUCH easier for developers to take advantage of this technology. This should mean that a lot more professional and consumer software will be able to take advantage of GPU processing.

[Vanillacide @ Jun 30th 2008 6:47AM]

Using Google or simply going to NVIDIA's CUDA website provides links to the OSX versions CUDA driver, CUDA toolkit, and CUDA SDK.

OSX 10.2.5 or greater is required, and the correct hardware too (such as 8600M GT found in a MacBook Pro).

loosely_coupled correctly points out that 'Snow Leopard' (OSX 10.6) will include OpenCL which works on both ATi and NVIDIA GPUs, and is a core technology of 'Grand Central' which is the new thread management system designed to spread load over multiple CPU cores and GPU shaders.

[Aryan @ Jun 29th 2008 3:02AM]

Guys... I think AMD is back...

[DefPo3t @ Jun 29th 2008 3:21AM]

HELLO amd had nothing to do with this
but yes if they can get this to work on a 48XX card then amd is back at least for this generation (not that they went already but this just solidifies it)

[Ihar `Philips` Filipau @ Jun 29th 2008 4:41AM]

Do NOT count on it.

All it takes is for nVidia to pester the PhyX engine with proprietary, geforce-only calls and functions.

[Jon Doe. @ Jun 29th 2008 3:14AM]

Mac user need not apply. But as if they care since they live in their own little bubble world where if it isn't Mac, its irrelevant.

[phanbouy @ Jun 29th 2008 3:28AM]

You're irrelevant. Does that count?

[DefPo3t @ Jun 29th 2008 3:34AM]

mac users can apply if they use bootcamp
os x users need not apply due to CUDA not being available on os x yet
get your facts straight moronic fanboi

[Pochi @ Jun 29th 2008 4:05AM]

@DefPo3t

Could you split those hairs any thinner?

[Ihar `Philips` Filipau @ Jun 29th 2008 4:44AM]

@Pochi

Apple is coming up with OpenCL for next OS release. Let's see what they will do out of it.

[Hamidxa @ Jun 29th 2008 1:43PM]

ihar

That would be bitchin if Apple could get OpenCL out the door.

Physics processing support for Quake 3 can't come fast enough !!

[chansthename @ Jun 29th 2008 3:49AM]

phsx cards are available in laptops but only select high range ones

[ShadowKain @ Jun 29th 2008 12:01PM]

When are people going to support physx completely? It's been available for a long time, but no one cares. Only with these cards and the Nvidia 260 / 280 seem to be pushing it harder. When they added some physx love with Unreal 3's (I think it was unreal 3) in that special level, it wasnt the physx card mostly, but the enormous mad hours to code every brick. Someone explain to me how physx will survive when the newer Nvidia 280's basically contain a GPU and CPU....

[thethirdmoose @ Jun 29th 2008 12:10PM]

Because now lots of people have physx cards?

[Ridgecity @ Jun 29th 2008 2:01PM]

you don't need to "code every brick". Did you know computers are good at repetitive and random things?

[Wwhat @ Jun 29th 2008 9:35PM]

Nvidia bought PhysX you see, and now they made drivers that use the GPU (like the 280's) instead of a dedicated PhysX card, and as a response someone hacked it to also work on ATI cards (allegedly).

[jordan @ Jun 29th 2008 1:59PM]

Unless AMD can get in on the Physx love, don't ever expect it to be completely adopted, as you're just further forking code between the two GPU powerhouses, since you'd have to make separate code for AMD.

I loved CUDA, in theory, but hate the fact that it looks like they're heading down the monopolistic road of making developers go their way or the highway in terms of high-performance/efficiency coding.

[jordan @ Jun 29th 2008 1:59PM]

oops, forgot to hit the reply for ShadowKain

[Wwhat @ Jun 29th 2008 9:40PM]

Yeah it's really annoying how they now separate EVERYTHING, you need to decide what mobo, what RAM, what CPU, if you want SLI or crossfire, if you want CUDA or DX10.1, every day things seem more compartmentalized and it's getting really tiresome, soon they can close down the ISO committee completely, the industry did away with sanity and interoperability.

[bri3d @ Jun 30th 2008 12:56AM]

Bullshit.
CUDA binaries do NOT just fly easily from one TOTALLY DIFFERENT architecture to another.
This is like someone saying OH YEAH I MOVED THIS EXTREMELY TIME SENSITIVE CODE FROM ARM TO x86 WITHOUT AN EMULATOR AND NOW IT RUNS FASTER, IT WAS EASY LOL.
There is a disassembler for CUDA binaries but then they would need an assembler for AMD video cards. And detailed knowledge of EXACTLY how they work (which is NDAed). And since they claim AMD were not involved, somehow I doubt that.