Scan's TekSpek

Our Aim
To provide you with an overview on New And existing technologies, hopefully helping you understand the changes in the technology. Together with the overviews we hope to bring topical issues to light from a series of independent reviewers saving you the time And hassle of fact finding over the web.

We will over time provide you with quality content which you can browse and subscribe to at your leisure.

TekSpek GPU - Graphics
NVIDIA GeForce GTX 680

NVIDIA GeForce GTX 680


Date issued:

NVIDIA is today releasing the GeForce GTX 680 graphics card. It has been designed to be the fastest single-GPU (Graphics Processing Unit) card in the world, and this TekSpek explains the technology behind NVIDIA's flagship consumer GeForce product.

Known by the codename of Kepler and based on a heavily-tweaked GeForce GTX 580 card, the main design goal for this release has been to construct a genuinely high-end GPU that focuses on energy efficiency as much as on performance. NVIDIA has achieved this aim by moving the manufacturing process from 40nm to 28nm - smaller processes can fit in more transistors into a given piece of silicon - and by redesigning the processing cores for lower-power usage.

In a secondary effort to reduce power, NVIDIA has reworked the memory-controller section of the chip, as well as instigated numerous behind-the-scenes improvements that are outside the remit of this TekSpek.

Perhaps the simplest method to understand just what has changed from GeForce GTX 580 to GTX 680 is to look at a comparison table.



GPU GeForce GTX 680 (2,048MB) GeForce GTX 580 (1,536MB) Radeon HD 7970
(3,072MB)
Radeon HD 7950
(3,072MB)
DX API 11.1 11 11.1 11.1
Process 28nm 40nm 28nm 28nm
Transistors 3.54bn 3.0bn 4.3bn 4.3bn
Die Size 294mm² 520mm² 352mm² 352mm²
Processors 1,536 512 2,048 1,792
Texture Units 128 64 128 112
ROP Units 32 32 32 32
GPU Clock (MHz) 1,006+ 772 925 800
Shader Clock (MHz) 1,006+ 1,544 925 800
GFLOPS 3,090 1,581 3,789 2,867
Memory Clock (MHz) 6,008 4,008 5,500 5,000
Memory Bus (bits) 256 384 384 384
Max bandwidth (GB/s) 192.3 192.4 264 240
Power Connectors 6+6 8+6 8+6 6+6
TDP (watts) 195 244 250 200
mGPU 4-way 4-way 4-way 4-way


The commentary also includes AMD's two finest GPUs: the Radeon HD 7970 and Radeon HD 7950. Writing from a technical standpoint, NVIDIA has increased the number of cores - the building blocks for GPU performance - from 512 to 1,536, or a 3x increase, though, now, the cores all operate at the same speed as the rest of the GPU, not double, per GeForce GTX 580. Meanwhile, while the transistor count has increased from 3bn to 3.54bn, but the die size has dropped by almost half, mainly due to the space-saving nature of the 28nm process.

NVIDIA, too, has decided to run with a high-speed 256-bit memory bus, again to save power, and the overall card bandwidth is the same as the GTX 580. Indeed, the card only draws 195W at maximal load, or almost 50W less than the GPU it replaces as the flagship GeForce offering. One can translate the technical facets of the card into a sentence that goes thus: GTX 680 is better than GTX 580 in every way.

Non-architecture improvements

GeForce GTX 680 also has a specific feature called GPU Boost. Put simply, a small chip on the board controls the core frequency of the card. During gaming the card increases its speed to 1,006MHz. However, the card automatically boosts its speed further if there is scope to do so. This scope is determined by just how close the game is able to push to the card's 195W power rating. Most games aren't that GPU-intensive, so the card boosts frequency and voltage safely. NVIDIA reckons the average speed boost increases the core clock to 1,058MHz, or about five per cent higher than normal. The user cannot turn this feature off, by the way.

The GTX 680 now supports up to four display outputs from one board. This means that each card can run up to four monitors. It is also possible to run 3DVision Surround from a single GTX 680 card - a technology that previously required two GeForces running in SLI. The standard outputs are two dual-link DVI ports, HDMI, and DisplayPort. As each output can feed a monitor, no special adapters are needed to operate three screens in tandem; they plug straight in to the back of the card.

NVIDIA is also ushering in a new form of image improvement called TXAA. This technology is reckoned to improve image quality by using advanced anti-aliasing methods - getting rid of jagged lines - and is the highest quality yet seen from any graphics card. Implementing TXAA requires games developers to support it, and large studios such as Crytek and Epic have already signed up to use it in their forthcoming titles. This is a GTX 680-only feature, mind.

Another feature that's due to be implemented across all modern GeForces is called Adaptive VSync. Switched on in the control panel, this technology reduces the tearing and stuttering that is sometimes observed when playing PC games. Often caused by a mismatch between the graphics card's output and display refresh rate, Adaptive VSync locks the framerate at 60 frames per second, just as regular VSync does. However, if it drops below this figure, which is often in demanding titles, Adaptive VSync better controls the framerate than the large drops seen with regular VSync, which can fall to 30fps or 20fps in one jump. Adaptive VSync brings these drops down in much smaller increments, leading your eyes to see a smoother, artifact-free picture. Think of it as smoothing out the way the image is displayed.

Performance

A new card and associated features wouldn't be worth much if performance was lacking. NVIDIA's GeForce GTX 680 is the world's fastest single-GPU graphics card, beating out competitor AMD Radeon HD 7970 by around 15-20 per cent, according to results published at independent hardware review websites. Meanwhile, it's around 35 per cent faster than the GeForce GTX 580.

It is also comparatively power efficient, pulling fewer watts than both aforementioned cards, and NVIDIA has designed the card to be very quiet when under load. The reference card is 10in length, powered by two six-pin PCIe connectors from your PSU, and takes up two slots. We advise SCAN readers to have a decent-quality PSU with a 600W rating and decent amperage on the 12V line(s).

All factors considered, at the time of writing, the GeForce GTX 680 is the very best graphics card that you can buy. UK pricing is due to start at around £400 for basic models, rising as partners such as EVGA, ASUS, Gigabyte, et al, release custom-designed models that, going by recent history, are clocked in at higher speeds and have better heatsink-and-fan units for even quieter gaming.

A full range of NVIDIA GeForce GTX 680 cards are stocked by SCAN. Please head here to peruse our catalogue.