Posted by : News Technology
Kamis, 24 Mei 2012
NVIDIA GTX 680 GPU chip architecture successor GK104 with
the latest generation of Fermi name Kepler. GK104 GPU chip consists of four
blocks with two blocks in GPC Streaming Multiprocessor. If the Streaming
Multiprocessor Fermi called BC, when Kepler, Streaming Multiprocessor last
generation called SMX. SMX is claimed to have a performance / watt two times
better than the previous generation Streaming Multiprocessor.
The GF110 GPU chip has on each block 32 CUDA live streaming
multi-processor cores. While GK104 GPU chip has 192 CUDA cores in each block
Streaming Multiprocessor (six times the GPU chip GF110). With as many as eight
blocks of SMX, the CUDA GPU chip GK104 Cores as many as 1536 pieces. GK104 is
the mastermind behind the NVIDIA GTX 680 is built with transistors as much as
3:54 billion. Although an increase in the number of transistors, die within
GK104 was reduced to 294 mm 2 by means of 28 nm, the manufacturing process.
Compared GF110 GPU chip, chip CUDA Core 1536 GK104 (three times) and 128
texture units (twice as much). There was however a reduction in units of 48
pieces of the Rop 32 pieces.
NVIDIA GTX 680 specs nvidia armed clocked higher than GTX
680th Time is at 1006 MHz GPU and memory clock is at 6008 MHz. Besides GPU
Shader is clock, memory and NVIDIA GTX 680 is now positioned to increase the
clock stands at 1056 MHz. Complete discussion can be a boost clock found in the
"GPU Boost" in this article.
NVIDIA GTX 680 is equipped with a memory 2 GB of GDDR5
memory types. NVIDIA GTX 680 now uses less memory than the NVIDIA GTX 580 is
256-bits.