Nvidia GeForce GTX 680 – The Nitty Gritty

The first thing to look at on the new GTX 680 is its power consumption. By re-modelling the GPU core architecture, Nvidia have managed to introduce more processing cores into the card and at the same time have improved the cards efficiency doubling its performance per watt of power consumed.

The GTX 580 on its Fermi architecture has 16 streaming multiprocessors, each with 32 cores, giving it a total of 512 CUDA cores overall.

With the Kepler architecture, the streaming multiprocessor has had a design change and the new Extreme Streaming Multiprocessor (SMX), of which the GTX 680 has eight to play with, now has 192 cores giving the card a total of 1536 CUDA cores.  With there now being significantly more cores per SMX, the Kepler design can provide far more performance than before but also at the same time uses far less power.

To show this boost in power efficiency Nvidia have made up some simple charts showing the differences between the GTX 680 and a selection of the other top performing cards on the market today. First off they have shown the comparison against a GTX 580, and we can see that a number of games show a significant boost on performance per watt of power consumed.

Comparing against the Radeon HD 7970 the performance per watt boost is not as great across the board, however the 7970 is a much newer card than the GTX 580. On average there is around a 20% boost in performance per watt.

Taking the power factor out of the equations ad looking at pure performance overall, the GTX 580 with the new Kepler architecture is the new leader as the worlds fastest graphics card at stock speeds.