FYI: To PhysX or not to PhysX

Do not read the following if you are not interested in gaming graphics. You will be bored to tears. Remember....I warned you. For all others....well...you too might get bored to tears.....sorry.

I'm currently working on a major system upgrade and while I'm waiting for another part to arrive, I decided to do an experiment with having a dedicated PhysX graphics card. For those who are not familiar with PhysX, it is a graphics engine by Nvidia that renders special effects in games (for more info, you can read about it here).

A big argument has been that having a dedicated graphics card to do just PhysX rendering is not needed when using one, two or especially three high end Nvidia-based cards in SLI...they can do the job just as well or even better. A second point to the argument against using one is that there aren't that many games that utilize PhysX. The other side of the argument has been that a dedicated PhysX-rendering card can relieve the main graphic card(s) to render just "pure" graphics and that there is a growing list of PhysX-optimized games, including the new Cryostasis game, that would justify having a card to just render PhysX (List of PhysX games).

Well...I tested using a PhysX-dedicated card on this new (if unfinished) system and was pleasantly surprised with the results. I used an EVGA 8800 GT 512mb for the PhysX card, along with two GTX 260 Superclocked cards that were in SLI. Just for kicks I also overclocked the GTX 260's up to GTX 285 speeds. For the test I used the Cryostasis Tech Demo...a short demo of the game that tests for speed/framerates and that is optimized for Physx.

In the first setup, I enabled just the two 260 GTX cards to render all graphics, ran the test 3 times and then averaged the results:

Time (to complete the test): 174.296 secs
Total Frames: 9821
Average Frame Rate achieved: 56.3 fps (frames per second)
Minimum Frame Rate achieved: 28.3 fps
Maximum Frame Rate achieved: 143.1

Second test, ran three times with averaged results...the 8800 GT is ENABLED to take over the PhysX rendering (the two GTX's are still running in SLI)...differences in bold:

Time: 132.705 secs
Total Frames: 9821
Average Frame Rate achieved: 74.0 fps
Minimum Frame Rate achieved: 31.0 fps
Maximum Frame Rate achieved: 158.40 fps

That would seem to bear out the argument that when rendering a PhysX-optimized game, having a dedicated PhysX card would give one a little better edge. However, there wouldn't be any dramatic difference with non-PhysX games using the PhysX setup. As soon as the third GTX card shows up, I'll test it both ways with 3-way SLI...that should be interesting.

Should the question of using a PhysX card come up here in the forums, then IMHO it would really be a question of whether the person really desires to play PhysX-capable games enough to warrant purchasing a dedicated card for PhysX (the 8800 GT or 9800 GT are recommended). Having the dedicated card would make a difference in game play...the person would have to determine if that is worth the purchase. But for general gaming though it obviously wouldn't be practical to purchase one (I already had the 8800 GT, so I'm using it). For the general gamer it would probably be a better idea to use their moolah to upgrade the CPU, RAM or graphics card.

Tip: If the person really wants a dedicated card for PhysX, then I would recommend the 8800 GT over the 9800 GT for the simple fact that the 9800 GT is for all practical purposes a re-badged 8800 GT (same speeds). It may have a couple of extra features (runs a little cooler for one thing), but they both will do EQUALLY well for PhysX rendering and the 8800 GT can be had for less money. (But if they are devil-may-care with money like me, then by all means get the 9800 GT....then, like me, sweat it out trying to explain such a purchase to the wife...so THEN try to be devil-may-care and see how far that takes you). :eek::o


Thus endeth the sermon.....just fyi for future reference.