nVidia GTX680 Review
Introduction and Technical Specifications
Published: 22nd March 2012 | Source: nVidia | Price: £429 |
Introduction
Every new generation of graphics cards has seen the battleground littered with the bodies of the fallen. Whether it's the domination of the 9800 Pro, the GTX280, the HD4870 and HD5870s or the GTX580, every generation has had the winners and the losers.
The original GTX480 was too hot and too noisy, but how power in abundance. This was quickly fixed with the GTX580 which ruled the roost as the fastest single GPU until the release of the awesome HD7970 which has taken up the mantle as the ultimate single GPU card.
Now nVidia are back into the fray with the GTX680, codenamed Kepler. Although it has a new GPU under the hood such was the frantic redesign of the Fermi for the GTX580 that this is actually the first real evolution of the Fermi core ideology.
Has it retaken the crown from the HD7970?
Technical Specifications
Now a lot has changed with the way that nVidia are solving the problem of supplying outstanding performance without breaking the bank in terms of power consumption and heat. Rather than clocking the bejeezus out of each core they have greatly increased the amount of cores available but left them running at a lower clock speed, thus saving power. However rather than just let this spare power go to waste the GPU now automatically overclocks itself up to the TDP limit so you get the dual benefits of low power gaming without compromising the performance. However, let's just look at the numbers for now and we'll cover the rest as we go.
Graphics Processing Clusters | 4 |
Streaming Multiprocessors | 8 |
CUDA Cores | 1536 |
Texture Units | 128 |
ROP Units | 32 |
Base Clock | 1006 MHz |
GPU Boost Clock | 1058 MHz |
Memory Clock | 6008 MHz |
L2 Cache | 512KB |
Video Memory | 2048MB GDDR5 |
Memory Interface | 256bit |
Memory Bandwidth | 192.26 GB/s |
Texture Filtering Rate | 128.8 GT/s |
Process | 28nm |
Transistors | 3.54 Billion |
Connectors | 2x DVI 1x HDMI 1x DisplayPort |
Form Factor | Dual-Slot |
Power | 2x 6pin |
Recommended PSU | 550W |
TDP | 195W |
Thermal Limit | 98°C |
The most immediately obvious changes are the low TDP, down from 250W, the insane memory speed (up from 4000 MHz on the GTX580) and the use of 8 SMX units to give a hefty 1536 CUDA Cores.
Most Recent Comments

My choice was made anyway, I am a faithful and has less than a huge disappointment, I have not bought anything.
Ordered, delivered tomorrow ... quick, quick, quick!!!
