Industry applications for AGEIA?

[Tetramagic2]

I was wondering if anyone can tell me whether there are any non-gaming industry uses for the Ageia Physx chip?
For example, manufacturing companies use GPUs for 3D CAD/CAM modeling....so I was wondering if there are business uses for a PPU.
For example:
a car company could use it to simulate crash tests;
construction firms could use it to simulate controlled demolitions;
earthquake simulations;
bullet forensics;
weapons/explosive testings?

Its my first post, I was thinking about this topic for a while and didn't see this addressed in any previous thread. Thanks for any responses!

 

[Capeta]

It could be used as an accelerator for scientific computing like the Clearspeed ASICs or GPGPUs.

 

[Techno+]

It would be good for physics simulation, just like you two said, but I think Ageia will only use this as a last resort, i.e only when their physx chip fails in the gaming arena.

 

[Ty]

IMO, Ageia is doomed unless they somehow manage a fairly massive shift in their business plan.

From day 1, I've been fairly pessimistic about them and from what I hear from the inside, things are pretty grim.

 

[Ateo]

http://www.ageia.com/developers/partners.html
http://msdn.microsoft.com/robotics/default.aspx
http://www.amtri.co.uk/amtri_non_asp/press_rel/PhysX.html

 

[Skrying]

The actual chip is nothing special if I recall right. Its more of their software that was useful and then its ability to offload to a secondary processor. As far as I know though scientific applications that this would be used for would already have product available that would handle this.

 

[Ateo]

The actual chip is nothing special if I recall right. Its more of their software that was useful and then its ability to offload to a secondary processor. As far as I know though scientific applications that this would be used for would already have product available that would handle this.

Lets Get Physical: Inside The PhysX Physics Processor

this tidbit is what makes it diffrent form CPU's and GPU's:

The Data Movement Engine (DME)

The PPU has a potentially vast amount of floating point power available, this however is of no use unless all the floating point units can be kept fed with data, the Data Movement Engine is responsible for doing this.

The DME is comprised of 5 memory control units, an external memory controller, a PCI bus interface and a “switch fabricâ€￾.

The switch fabric is a network of switches and busses which allow all the different units to talk to one another. In this case the switch fabric has 7 x 256 bit bidirectional ports, the number of units which can be talking simultaneously is not specified.

The work done in the Data Movement Engine is controlled by a series of 5 memory control units. 4 of these are connected to the Vector Processor Elements in the Floating Point Engine, the 5th is connected to the PPU Control Engine.

Each memory control unit contains a block of RAM and it moves data to and from it. This will mainly involve passing data to and from the external RAM and the vector processor element it is connected to, with the memory controller unit’s RAM acting as a buffer in between. They are not limited to this however and can also move data to and from the other memory controller units and the PCI bus.

You may wonder why it doesn’t just move data directly to or from the vector processors but this is done to make the usage of the external memory bus as efficient as possible. Moving data in big chunks is faster than moving data in small chunks so doing this will increase performance. Keeping data in on-chip buffers also allows data to be moved around the chip without going to main memory, again saving memory bandwidth.

The connections to the Switch Fabric and the Vector Processor Elements are separate so it looks like two types of communication can be operating simultaneously. For example, data could be written to one of the Vector Processing Elements while other data is being read in from external memory.

 

[Ty]

Let's just say that when they got their first silicon back, they were expecting an order of magnitude better performance.

 

[Capeta]

Let's just say that when they got their first silicon back, they were expecting an order of magnitude better performance.

Source?

 

[Ty]

Source?

I personally know two programmers that have worked there.

 

[Capeta]

I personally know two programmers that have worked there.

Didn't they already have a chip that worked and it was only a driver issue that held back performance?

 

[Frank]

For now, you're better off with a Cell card or server. Those are available, and at least twice as powerful. And except if you want to work with assembly, there is probably little difference between them.

 

[Blazkowicz]

Lets Get Physical: Inside The PhysX Physics Processor

this tidbit is what makes it diffrent form CPU's and GPU's:

and this is what makes it similar to Cell, implementation probably differs but what they describe sounds like their version of SPE, local storage and ring bus.

 

[Ateo]

and this is what makes it similar to Cell, implementation probably differs but what they describe sounds like their version of SPE, local storage and ring bus.

Cell CPU in PS3:
http://en.wikipedia.org/wiki/PlayStation_3#Central_processing_unit

The PS3's 3.2-GHz Cell processor, developed jointly by Sony, Toshiba and IBM ("STI"), ...

The preformance of that Cell CPU:
http://en.wikipedia.org/wiki/Cell_microprocessor#Power_Processor_Element

...each SPU can complete two double precision operations per clock cycle, which translates to 6.4 GFLOPS at 3.2 GHz; or eight single precision operations per clock cycle, which translates to 25.6 GFLOPS at 3.2 GHz

The VPE's in the PhysX:
http://www.blachford.info/computer/articles/PhysX2.html

If this arrangement is used in the final implementation the PPU should be capable of a theoretical 208 instructions per cycle (6 multiply-adds + 1 Integer) x 16 VPEs. This is nearly 3 times what the Cell processor in the PS3 can theoretically do and up to 50 times what any current conventional (single core) desktop processor can theoretically do. This equates to 96 GigaFLOPS for each 500MHz, the clock speed has not been announced but should be in the 500MHz - 1GHz range.

I'm not sure about it's clockspeed, this is what I could dig up:
http://www.hexus.net/content/item.php?item=5492&page=2

That chunk of silicon is made by TSMC in Taiwan, on their 130nm major process node for Ageia and Aegia's AIBs. Measuring 14 x 13.5mm (some 188mm²) and comprising around 125 million transistors, it appears to be clocked in the ranges 250-266MHz or 500-533MHz, based on some early in-house codes.

Ageia gotta be make some pretty shitty drivers in order to not being albe to utilze that much computing power, and the bloody thing sucks only 28 Watts:
http://personal.inet.fi/atk/kjh2348fs/ageia_physx.html#generalinfo

Power consumption: chip = 20 watts (peak), entire PhysX card = 26-28 watts

Now if you could link those cards...

 

[SlmDnk]

Some new information for this topic.

Can your PPU make other GPGPU kind of calculation?

The PPU can be used for accelerating other calculations in scientific, engineering and financial applications such as DCC, video encoding/decoding, financial options pricing, CAD applications, medical imaging etc. We are working with partners on each of these applications

If so do you have a software in development like NVIDIAs CUDA that lets you program in C to run GPGPU on G80?

Yes.

Source: http://www.chilehardware.com/articulo.php?sid=1736&pageid=2

 

[Ateo]

Some new information for this topic.

SÃ¥ the rumblings i had heard about PPU "SLI" where true, interesting:

Do you plan to support two or more card working in parallel?
Yes.

 

[Techno+]

there is an interesting article at Xbitlabs saying that AGEIA is already targeting non-gamer apps fir Physx chips

http://www.xbitlabs.com/news/multimedia/display/20070221131421.html

 

[SlmDnk]

there is an interesting article at Xbitlabs...

The Manju Hegde interview, from which Xbit's info is from, was posted earlier in this thread

http://www.beyond3d.com/forum/showpost.php?p=933656&postcount=15

 

[Tim Murray]

You realize that one of the great killers of performance in a GPGPU app is touching the PCIe bus too much, right? PhysX is still on PCI. Have fun with that.

 

[Ateo]

You realize that one of the great killers of performance in a GPGPU app is touching the PCIe bus too much, right? PhysX is still on PCI. Have fun with that.

And you trying to compare the architechture of the PPU to a GPU?
If so, have fun with that...