New AGEIA PhysX Reviews: Now in Numbers

First Physics Accelerator Failed

by Aleksey Razin

[ 05/05/2006 | 02:49 PM ]

The first AGEIA tests in actual gaming environment were based on the evaluation of visual quality improvement provided by the physics accelerator. The editors who had their hands on the new PhysX card recorded a few movies and offered the users to be the judges. In the meanwhile, many of computer users are more used to the numeric representation of performance results, so they were longing for new product reviews that would discuss the performance gain or drop in case with and without PhysX solution.

<%BANNER[article_nw]%>

Luckily, two hardware review sites – Anandtech and Hexus - posted articles like that these days. But before we discuss the obtained results let’s once again dwell on the differences between the ASUS PhysX P1 and BFG PhysX cards. As you know, the BFG solution features 128MB of memory vs 256MB of memory on the ASUS card. But besides that, the BFG memory is also slower – it works at 500MHz DDR, while the memory on the ASUS card works at 733MHz DDR.

It is quite natural that the official test program was Ghost Recon: Advanced Warfighter game that is one of the first applications to be optimized for AGEIA PhysX. The guys from Hexus site pointed out one remarkable feature: although the physics accelerator makes very bright first impression, you start noticing that it is all a sort of illusion later on. All fragments that break off the shot objects are designed from the same sprites: they have been turned different ways and scaled up or down to create the impression of uniqueness of each piece. Moreover, once the pieces touch the ground they dissolve spoiling the realistic effect.

Note that optimization for AGEIA PhysX is useful not only for the future owners of extension cards like that. In fact, the same effects can also be calculated by the system CPU or a special PhysX chip, the latter however should do it faster and unload the CPU significantly, because it is specifically designed for tasks like that. ATI and Nvidia on their end suggest using graphics chips to calculate physics effects, including the chips used in SLI and CrossFire configurations. Although in this case they will have to use the competitor’s Havok FX engine that is AGEIA PhysX owners will hardly be interested in this option.

The current implementation of the PhysX technology doesn’t allow the end-user to adjust the physics processing speed by rearranging the workload between the CPU and the so-called PPU. So far the game developers are the ones setting what they think is optimal workload balance, which is not always optimal in the end. Our colleagues from Anandtech site noticed that AGEIA PhysX solution causes the fps rate in the game to drop down dramatically for a few moments especially after completing the calculation of a significant effect, such as an explosion for instance. When the same scene is calculated on the software level, no harm is done to the gaming performance but at the same time, the visual quality is less impressive, because the same explosion generates considerably fewer broken off pieces and other object fragments. As a result, the game slows down because it has to transmit the data to and from the PhysX chip. Maybe it is the narrow PCI bus that is to blame here, so when we see physics accelerators for PCI Express x1 and PCI Express x4 the problem will be eliminated.

Of course, the situation should get better when there appear new games optimized for AGEIA PhysX and when the drives get better. So far the games where we can see the new AGEIA PhysX working look like something fun to watch but not impressive enough to spend extra $300 on :) New gaming hits with high-quality optimization for PhysX that are expected to be released within the next 12 months should push the new technology forward. And in the meanwhile looks like there is no real need for a new unusual piece of hardware.

• Anandtech: ASUS Debuts AGEIA PhysX Hardware;
• Hexus.net: Ageia PhysX PPU.