Looking closer at the Kepler Architecture
nVidia has made a passive-aggressive yet subtle statement with its Kepler architecture, which is clearly evident from the success the GTX 600 series is receiving. Some almost say it is the resurgence of the giant GPU maker, ever since ATI’s been quietly eating up its market share. ATI has been reporting exponential sales growth ever since it introduced the HD 4000 series. And ATI has consistently delivered great graphics solutions ever since, with the high point being the HD 6000 series. Where nVidia was the De-facto leader in the GPU business in the pre-2008 years, the general image of nVidia now resembles that to an underdog playing catch-up. This is about to change now, as it is clearly evident that nVidia has put up an effort on its Kepler architecture. Not only do they perform great, they’ve reduced power draw considerably. The Kepler’s arguable bet on going for Unified Shaders have paid off.
The GTX 660 Ti is based on the GK104 chipset. You can say that it’s a chipset that’s been done-to-death by nVidia. The GK104 is the same chipset used in the GTX 670, 680, and 690. To be honest, nVidia doesn’t have many options after the GK104 and GK107. And that is the actual reason why their GTX 6xx cards are mostly at the upper end of the spectrum. They have moved away slightly, if not completely, from the low-budget/mainstream market. This strategy gives them less to worry about on the production side since they have to maintain only one or two types of chipsets. But I still believe that they need to broaden their line-up, if not now, then soon. It’s true that they receive better margins for their up-scale products, but the mainstream is where the volume is.
GTX 680 | GTX 670 | GTX 660 Ti | GTX 570 | |
Stream Processors | 1536 | 1344 | 1344 | 480 |
Texture Units | 128 | 112 | 112 | 60 |
ROPs | 32 | 32 | 24 | 40 |
Core Clock | 1006MHz | 915MHz | 915MHz | 732MHz |
Shader Clock | N/A | N/A | N/A | 1464MHz |
Boost Clock | 1058MHz | 980MHz | 980MHz | N/A |
Memory Clock | 6.008GHz GDDR5 | 6.008GHz GDDR5 | 6.008GHz GDDR5 | 3.8GHz GDDR5 |
Memory Bus Width | 256-bit | 256-bit | 192-bit | 320-bit |
VRAM | 2GB | 2GB | 2GB | 1.25GB |
FP64 | 1/24 FP32 | 1/24 FP32 | 1/24 FP32 | 1/8 FP32 |
TDP | 195W | 170W | 150W | 219W |
Transistor Count | 3.5B | 3.5B | 3.5B | 3B |
Manufacturing Process | TSMC 28nm | TSMC 28nm | TSMC 28nm | TSMC 40nm |
Market Price (approx) | Rs. 36,000 | Rs. 26,000 | Rs. 20,000 | Rs. 28,000 |
The GTX 660 Ti is very similar to the GTX 670. It is the exact same architecture used. The base clock, boost clock, memory clock and the number of shaders are the same. The only difference lies in the number of ROP/L2/Memory Clusters. Where the GTX 670 uses 4, the GTX 660 Ti makes do with 3. This is also the reason why there is a difference of 20W in the power draw between both the cards. The missing ROP also scrapes 25% of the performance straight away in theory. To put it in other words, think of it as two cars with the exact same horsepower, but different engine capacities.