Gigabyte's Windforce X5 GTX 680 SOC and A Quick Retrospective on nVidia's WHQL Drivers

I recently picked up one of Gigabyte's GTX 680 Windforce X5 SOC cards for the low price of $155, and being the tweaker I am I decided to benchmark it instantly. I've not owned a Kepler GPU since [regretfully] selling my GTX 690 over a year ago and I wanted to feel that GeForce power. The drivers I installed were the WHQL 341.44 from February of 2015. First test I ran was Heaven 4.0 with lowered settings (as per the settings used on the GPU Benchmark Thread).

A very quick aside, I have heard word that nVidia's drivers took a steep decline for Kepler owners after the launch of the full Maxwell lineup, and was sort of expecting it.

Now back to this first test. The card ran flawlessly, never exceeding 60C with the fans on a default speed of 35%. Stutter was minimal, framerates were good. However, I had previously been using a DEV Sample Radeon HD 7950 with an identical operating hardware layout to a Radeon HD 6970, and was expecting the GTX 680 to leave it in the dust.

It didn't.

The overclocked DEV Sample Radeon HD 7950 outperformed the higher clocked GTX 680 SOC by 3.2% or +2.4FPS average. Here is the 7950's score, and here's the first GTX 680 score.

After checking that the card was indeed boosting properly, and that there was no power limitation I ran the test again and got the same result. It was not until I had overclocked the memory by 95MHz and pushed the core up by 52Mhz that the scores began to match. I don't have the validation image for that run though, as the drivers crashed.

Now moving away from comparing the two cards directly, I moved to assessing whether there was indeed a driver limitation as I have heard in the past. I downloaded three drivers from various dates, starting from the launch of the GTX Titan Black in 2014 and ending on the most recent WHQL certified driver. Not a huge spread, but it should show if nVidia stunted Kepler after Maxwell.

The Drivers:

  • WHQL 334.89 - Released 2/18/2014
  • WHQL 352.86 - Released 5/18/2015
  • WHQL 358.91 - Released 11/9/2015

Results from Heaven 4.0:

Not exactly what I was hearing. I was sure the older driver would perform better compared to newer drivers, and yet there are the scores. I am still confused as to why the card performs so low in this benchmark, when older generation GeForce cards such as the 560 Ti run just as well as they are expected to.

I'll continue playing with it and will try many other benchmarks and games before I'm done just to find where the performance went off to, but until I find out I'll just be confused and run the card with an insane OC, because in the end that's what the card was meant to do.