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

Maxwell architecture arrived nearly two weeks ago on 28nm as the GM107 GTX 750 Ti. This new $149 GPU is the most basic GPU of Nvidia’s upcoming Maxwell architecture and it features 640 CUDA cores boasting extra-ordinary efficiency and almost matching a GTX 480 in terms of performance.

Now we want to see how it performs as a stand-alone dedicated PhysX card to accompany a GTX 680, GTX 680 SLI, GTX 780 and a GTX 780 Ti. We are going to compare the GTX 750 Ti to the GTX 650 Ti it replaces and even compare it to a GTX 680 to see if any PhysX performance gets left on the table. We also just added a GTX 480 as a dedicated PhysX card to the mix as we have already compared its graphics performance to the GTX 750 Ti two weeks ago.

What is important to realize is that the $149 GM107 GTX 750 Ti (670 Maxwell Cores) is a much more powerful GPU than the $149 launch-priced GK106 GTX 650 Ti (768 Kepler CUDA cores) that it replaces, but it only requires about half the power and a little more than 25% more die area on the same 28nm process. A GTX 750 Ti does not require more power than what a standard 75W PCIe rail can provide. No external power cable is needed which makes it perfect for use as a dedicated PhysX card!

So as to not bottleneck our graphics nor physics calculations, we will use our Core i7-4770K at 4.0GHz which equalizes very well to our previous Ivy Bridge benches with Core i7-3770K at 4.5GHz. We use 8x2GB of fast Kingston HyperX “Beast” DDR3 at 2133MHz in an ECS flagship Z87 “Golden” motherboard.

We can’t wait to test out our GTX 750 Ti as a dedicated PhysX card, but before we begin the testing, head over to our testing configuration.