GeForce GTX 680 - NVIDIA's 28nm Successor to Fermi


Temperature, Power Consumption & Overclocking

Temperature

At load conditions, the the card runs at  69 degrees Celsius. This may be attributed to its lower power consumption (195W TDP) as well as its new finstack of its heatsink. We wonder if the third party card manufacturers would be able to lower the temperature bar further as they roll out their variants of the GeForce GTX 680 in the near future. Compared to the GTX 580, it's a marked improvement.

 

Power Consumption

The GeForce GTX 680 has the third lowest power consumption and it was not as power efficient as AMD Radeon HD 7970 nor, the 7950 even though the new card has a lower TDP rating than the former. One reason for this outcome is due to the GPU Boost working in the background that is always trying to maximize the power budget of the GPU where possible. This means it is better utilized than other graphics cards of the past and as a result of its higher performance, you pay somewhat in its power consumption. At least it's much more efficient than the last generation of products. Overall, given its performance aspects, we think the power consumed is reasonable.

 

Overclocking

For our overclocking efforts, we made use of EVGA Precison X utility software. We managed to push the GPU Clock Offset slider to a value of 130MHz. For its memory subsystem, we managed to garner a headroom of 300MHz. For these settings, we left the voltage of the card at its default value of 0.987mV. NVIDIA GeForce GTX 680 managed to blaze ahead of the dual-GPU cards, besting AMD Radeon HD 6990 by a margin of approximately 4.4% at Performance settings! The other cards were trailing in its wake. At Extreme settings, GeForce GTX 680 was beaten by AMD Radeon HD 6990 by a margin of 3.6%. A close call indeed.

In order to ascertain the nett GPU clock speed we have achieved from our overclocking efforts, we ran the performance monitoring tools in the background. We managed to reach a GPU clock speed of 1240MHz with our GPU Clock Offset set at 130MHz. This meant that we managed to increase the default boost clock of 1058MHz by approximately 17% and we saw a GPU Boost of 9% in action when you consider the new base clock of 1136MHz (which is 1006 + 130).


The Over-Volting Attempt

In our attempt to achieve higher GPU Clock Offset margins, we over-volt the card's GPU to 1.100mV. This allowed us to set the GPU Clock Offset margin by a slight gain of 4MHz to 134MHz with the its Memory Clock Offset at the same headroom level of 300MHz. Our performance took a dip at both settings of 3DMark11, with its Performance score registering at 9121 (a reduction of 4.2% against its earlier score of 9513); while its Extreme settings score was whittled down by a margin of 3.2% to 3150. While your experience may vary, it looks like our advice is to forget about over-volting the card.

Peering into the monitoring graph, we understood better as to why our over-volting attempt didn't go well. Though we managed to increase the card's GPU Clock Offset by a meagre 4MHz to 134MHz, the highest recorded core clock speed was 1154MHz! This was about 7.5% lower than what we had achieved without increasing the operating voltage of the card.

9.0
Performance
9
Features
9
Value
8.5
The Good
Excellent all-round performance
Extremely quiet for a flagship card
GPU Boost in Kepler architecture
The Bad
Temperature and power consumption wasn't low but was justified with GPU boost