Using Maxwell’s GTX 750 Ti as a dedicated PhysX card

Performance summary chart

Here is the summary chart of five modern games that use PhysX. The highest settings are chosen and it is DX11 when there is a choice; DX10 is picked above DX9, and the settings are generally ultra and maxed as specified on the chart. The benches are run at 1920×1080 and at 2560×1600.

The GTX 680 is shown in the first column and it calculates PhysX besides graphics. The second column features the same GTX 680, but it is now paired with a GTX 750 Ti as a dedicated PhysX card. The third column substitutes a GTX 650 Ti for the 750 Ti. The fourth and fifth columns show a GTX 780 performance without and with a GTX 750 Ti, while the sixth and seventh columns do the same for the GTX 780 Ti.

The final five columns are devoted to the GTX 690 at GTX 680 SLI speeds. First, (Column 8) results are given without a dedicated PhysX card, then (Column 9) uses a GTX 750 Ti, while (Column 10) substitutes a GTX 650 Ti. The second to last (Column 11) uses a GTX 480 as a dedicated PhysX card with GTX 680 SLI, and finally, the last column uses a GTX 680 to calculate the physics.

We see a very clear trend. Pairing a GTX 650 Ti with a GTX 680 brings performance improvements to all games over just using a GTX 680 by itself, except at 1920×1080 in the Cryostasis Tech Demo. Of course, Cryostasis is a very old game and there may also be driver issues.

Some of these gains using a dedicated PhysX card with a GTX 680 are quite significant and make the difference between enjoying a fluid experience and having an unsatisfactory one. However, in every single case, the new GTX 750 Ti is stronger as a dedicated PhysX card than the GTX 650 Ti and we continue to see it bring performance improvements to the GTX 780 and to the GTX 780 Ti.

Using GTX 680 SLI, we can see that it is again held back by using the GTX 650 Ti as a dedicated PhysX card in Cryostasis, yet it brings a performance advantages to the other four games. Upgrading the GTX 650 Ti to a GTX 750 Ti as a dedicated PhysX card, brings further performance increases to all five games without exception.

It is interesting to see that a dedicated PhysX card brings the most needed performance gains to the lesser-performing GTX 680, with a GTX 750 Ti generally making only small gains over the GTX 650 Ti. As the primary graphics card becomes more powerful, the GTX 750 Ti distinguishes itself from the GTX 650 Ti. A GTX 650 Ti brings only modest gains to a GTX 680 SLI set-up, for example, while a GTX 750 Ti gains fps more significantly.

We added a GTX 480 to the benches at the last minute since we have already compared it with the GTX 750 Ti in our original evaluation. We note that its performance as a dedicated PhysX card is better than with using the GTX 650 Ti, and it stands with the GTX 750 Ti and even with the GTX 680 in the more demanding benchmarks. However, even though a card does not use as much power when used as a dedicated PhysX card compared to using it as a graphics card, the original reference GTX 480 adds significant noise and heat and it is far preferable to use a nearly-silent and power-sipping GTX 750 Ti if there is any other choice.

Finally, when we substitute a GTX 680 for the GTX 750 Ti as a dedicated PhysX card, we see that no performance is left on the table except in the Batman: Arkham Origins games at 1920×1080, where there may be driver issues with the new video card. In this particular case, as a dedicated PhysX card, the GTX 480 is faster than the GTX 750 Ti and GTX 650 Ti, and it nearly catches the GTX 680.

Sometimes there are also advantages to using a dedicated PhysX card by leaving it in the “autoselect” mode in the Nvidia control panel instead of dedicating a particular card. We gained 4fps, from 96 to 100fps, in Batman: Arkham Origins at 1920×1080 with GTX 680 SLI by choosing autoselect.

Let’s head for our conclusion.