ZOTAC GeForce GTX 680 Review

By Nimish Sawant

24-03-2012

Nimish Sawant

Digit Rating: Excellent

4.5/5

Features:

Performance:

Value:

Design:

PROS

Excellent performance
Good pricing

CONS

None

MRP: 32999

Check out hot Parts-and-Peripherals deals on

Summary

The performance scores skew in favour of ZOTAC GeForce GTX 680. When it comes to pricing, the ASUS Radeon HD 7970 is priced at a whopping Rs. 38,000 whereas the ZOTAC GTX 680 comes at a sweet pricing or Rs. 32,999. The huge difference in pricing and performance scores make the winner quite evident. The card to win this battle is the ZOTAC GeForce GTX 680 and for those looking to get the highest end single GPU based card, we feel the choice has already been made to go for the GTX 680.

Review
Specifications & Performance

The Single GPU King: ZOTAC GeForce GTX 680 Reviewed!

Update: ZOTAC GTX 680 pre-bookings have started. You can get it from Aditya Infotech. Phone No:+91 8860661898.

The long wait for the next-gen NVIDIA graphics card series is finally over. NVIDIA announced its flagship single-GPU based graphics card in the new GTX 600 series day before yesterday. The new GPU is codenamed Kepler and it improves upon the Fermi GPUs.

ZOTAC was kind enough to send us the ZOTAC GeForce GTX 680 card for review. We couldn't wait to try it out, specially since the AMD flagship card had scored better than NVIDIA's GTX 580 in most benchmarks.

The GeForce GTX 680 GPU is based on TSMC's 28nm manufacturing process and houses 3.54 billion transistors. The base clock has been set to 1008MHz, with a Boost Clock speed of 1058MHz. We will get to the Boost clock in a bit. The GTX 680 is made of 1536 CUDA cores and the card has 2 GB of fast GDDR5 memory clocked at 6.006GHz. It has managed to keep the maximum TDP at 195 Watts which is much lower than 244 W seen on the GTX 580.

Let us first try to understand what is under the hood of the card.

Looking at the basic architecture of the Fermi GPU, you will notice that it goes from CUDA Cores → Stream Multiprocessor (SM) → Graphics Processing Cluster (GPC). The graphics Processing Cluster has dedicated resources for rasterization, shading, texturing, computer and a majority of the GPU's core graphics related functions are performed within the GPC.

In the GF110 for instance, there are 32 CUDA cores housed in 1 SM; 4 SM make 1 GPC and 4 GPCs make the GPU. In Kepler GK104 GPU, the SM is called SMX and the math goes something like this: 192 CUDA cores housed in 1 SMX; 2 SMX per GPC and 4 GPCs make the GPU.

For the GK104 GPU (Kepler) that equates to 192 x 2 x 4 = 1536 CUDA cores, which is a huge jump from 32 x 4 x 4 = 512 CUDA cores seen on the GF110 GPU on the GTX 580.

One of the main reasons behind a jump in the number of CUDA cores is that NVIDIA have decided to do away with the shader clock unit completely. In the Fermi architecture, the shader clock operated at twice the base clock speed. So instead of having lesser number of higher clocked units, NVIDIA now has more lower clocked units. In Kepler GPU, you have a single base clock called the Graphics clock which at 1008 MHz on which all the functional units within the GPU run at.

Visit page two to read on about the GTX 680's new features, our test setup, and benchmarks..

Graphics Clock: 1006 MHz; Boost Clock: 1058MHz; Memory: 2 GB GDDR5; Memory Bus width: 256-bit; CUDA Cores: 1536; Raster Operators: 4; Max TDP: 195 Watts; Bus Interface: PCIe 3.0