Table of Contents
Introduction
Ageia impressed everyone with demonstrations of what their product could do. Fluids flowed with more ripples, sparks were much more abundant, and the classic explosion was much more satisfying. Enthusiasts were excited about the new technology and were eager to employ this new hardware onto their own systems. But the enthusiasts were not alone – Nvidia was also interested in this new technology. In 2008, Nvidia acquired Ageia and their PhysX products. The PhysX engine is now known as Nvidia PhysX.
After several revisions, Nvidia completed their project and released their first driver package that included PhysX compatibility. By accessing the Nvidia control panel and enabling the option, a supported Nvidia card could now perform both functions at once. Nvidia continued to work on their PhysX platform, which now has the ability to offload the physics processing onto a second supported Nvidia card.
One of the biggest questions consumers have about this sort of new technology is “how much will this really make a difference?”. We decided to independently test the Nvidia PhysX platform, and share our results.
Methods
Test System
Asus P6T Deluxe BIOS Version 1402
Intel Core i7 965 Stock
6GB Kingston HyperX 1600Mhz CAS8
Western Digital 640GB Black Edition
Windows Vista 64-bit
Video Cards
nVidia 186.18 Driver Package
Geforce 285 GTX 1GB
Geforce 250 GTS 1GB
Geforce 9600 GT 512MB
Game Settings
Monitor Resolution at 1680 x 1050
H.A.W.X at Max Settings with DX10 enabled
Far Cry 2 at Max Settings with DX10 enabled
Cryostasis at Max Settings with DX10 enabled
H.A.W.X. Results
The clear winner in the above comparison is the GTX 285 SLI setup with the PhysX disabled. When using a single GTX 285, enabling PhysX gives yields almost no performance change. There is not a significantly large impact on performance by adding a dedicated PhysX card and H.A.W.X looks to run better with PhysX disabled.
The GTS 250 as a standalone card has a lower average FPS, but nearly equal maximum FPS scores. With the GTS 250’s in an SLI configuration we find a better mean score and it performs better with a 9600 GT for a dedicated PhysX card compared to a dedicated GTS 250.
The 9600 GT placed on the lowest tier of the three. It did have much better maximum FPS as compared to the GTS 250 but with lower average scores, several sections of the benchmarks slowed to a crawl. It also performed best as a standalone card, when compared to its SLI setup.
Farcry 2 Results
In Far Cry 2, one very noticeable finding is that all three dedicated configurations gave us very similar results. All three configurations performed just like the single GTX 285 in each mode. A pair of GTX 285’s with PhysX enabled out-performs all the other setups.
The GTS 250’s gave us very consistent scores. Both SLI modes showed similar scores. Single card configurations, as well as having a dedicated PhysX card were very similar in benchmarks. This was almost a mirrored pattern that we saw with the GTX 285 configurations.
Like the other two configurations, the 9600 GT’s gave us similar results. Summary: Far Cry 2 benefits the most from an SLI setup with PhysX enabled.
Cryostasis Results
The pattern is pretty straightforward. With PhysX enabled we get decent scores and smooth game play. Turn off PhysX and the scores drop significantly. The SLI with PhysX configuration took high score but across the chart we saw similar results. [Note: The high maximum FPS is reflective of the ten seconds where physics are not in play, and thus maximum FPS scores shoot up.]
The GTS 250 continues the pattern – PhysX is a massive boost in performance, and for the first time we see a dedicated physx card give a significant performance boost – although it’s still inferior to a SLI setup.
Unlike the other two setups the 9600 GT could not yield a decent score with PhysX enabled until it was paired with a dedicated PhysX card. This is the only benchmark that gave us a better overall score with a dedicated PhysX card.
Conclusions
Single card setups perform very similar with or without PhysX enabled and the user shouldn’t hesitate to turn on the function. As for dedicated PhysX card configurations, benchmarks show us that the single card counterpart performs just as well in most scenarios. In many cases, dedicated cards just do not have much of a performance benefit, and will simply add expense, heat and complexity to a system. Until more games can be heavily optimized for these types of special effects which utilize the PhysX engine, there is just not much benefit in dedicating a video card to handle the physics workload.
The potential for PhysX is great, but will depend on whether developers will want to embrace this technology. It’s no coincidence that the PhysX engine has the biggest impact in Cryostasis, which has been coded specifically to take advantage of PhysX. If more developers start to take advantage of this in a future, expect to see PhysX become a more important feature. For now, our purchasing advice is to enable PhysX on any compatible Nvidia card, but hold off on purchasing a dedicated card for the task.