Ageia's PhysX PPU and first board exam
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.
It's ahead of Havok FX™ in what it can accelerate, the PhysX PPU able to process more than just large-scale collision or 'effects' physics via its API. There's support in the API for limited fluid dynamics sims, vehicles (wheel, torque and tire simulation), object raycasting and more, which the PPU can fully or partly accelerate (with Ageia moving more onto the hardware as time goes by).
At its core it's just a wide parallel stream processor with a command core (sometimes called the control engine) to run it all and a memory controller to move data onto and off of the chip during processing.
The parallel elements are themselves made up of multiple fully FP32 processing units in MIMD, each with a SIMD array of vector units. Think of the hardware as an NxM array of vector units in that sense (if the patents are to be believed), likely 4x4 for this first iteration of PhysX hardware, which fits in with their likely makeup and implementation in silicon.
The units aren't generally programmable though, at least in the way you might be used to thinking about with a GPU. You can't (easily) 'shade' the physics interactions that take place, and be certain the hardware will (mostly) execute an instruction stream you can further control. There's no compilation of physics programs as you would a shader program on a programmable GPU.
More on the programming of the thing shortly.
In terms of its data rate, the hardware is supposedly capable of 6 5D vector MADDs per cycle, per vector unit. In the 4x4 design we suspect first hardware to have, that's a near 50Gflop (all FP32 flops) rate when fully utilised, at 250MHz.
Of course we could (and may well) be entirely wrong; Ageia are reluctant to tell us how it works and we're left poking at patents and developer information to glean our ideas about the hardware.
Board
Our sample is one of BFG Tech's PhysX boards, sold by the mainly NVIDIA AIB partner into system builders, SIs and retailers as we speak.
Measuring 167x99mm, the PCB for these first BFG PhysX boards is pretty much identical in dimension to a modern soundcard. Indeed your author's Audigy 2 ZS is much the same size. I've yet to come across a soundcard that needs an active cooler and auxiliary power input, though.
The board supposedly needs a maximum of 2.3 amps under maximal load (28W total via the 12V supply you feed it) although I've yet to measure it taking more than ~1.8A amps (~21W) under any load condition.
The 40mm fan and heatsink combo reminds your author of those fitted to graphics boards in days gone by. It spins devoid of any thermostatic control and rotational regulation. It's loud enough to intrude on a silent PC and we estimate something in the order of 32-36dBA (yes, it's just a guess and we know it's logarithmic but we don't have the tools to accurately measure).
The PhysX PPU's memory controller supports the same GDDR3 memory that current graphics hardware does, and the BFG sample is equipped with four 32MiB 500MHz Samsung DRAMs. Each populates a 32-bit channel and they combine to give the chip a maximum read bandwidth of 16GB/sec and 128MiB total on-board storage.
And that, dear reader, is just about it. The heatsink gets pretty damn hot over time, indicating the aluminium heatsink and fan aren't the best pairing to get rid of the current PPU's heat, but a well ventilated chassis should see you right. Oh, and it lights up.
