Last week Nvidia announced three major graphics card . RTX 3070, 3080, 3090.
Spec :
RTX 3070 :
Gpu core : Cuda core = 5888 and base frequency 1.50GHz / Boost clock 1.73 GHz
Memory : 8GB Ram GDDR6
RTX 3080 :
Gpu core : Cuda core = 8704 and base frequency 1.44GHz
Memory : 10GB Ram GDDR6
RTX 3090 :
Gpu core : Cuda core = 10496 and base frequency 1.40GHz / Boost clock 1.70 GHz
Memory : 24GB Ram GDDR6
Note :
CUDA – Compute Unified Device Architecture
These cores are developed by the Nvidia for parallel computing .
why we need 10496 cores for GPU , while modern CPU has 8/16/32 cores is enough for computing?
During the graphics intensive work the computer need to work on complex calculation simultaneously so these huge core in graphics card will resolve the problems in parallel. compare to Intel or AMD Cpu architecture CUDA core are designed smaller.
- Both RTX 3080,3090 has 28 billion transistors . while 3070 was built by 17.4 transistors.
- All the RTX models are built by samsung 8nm node
- Nvidia also stated that RTX 3090 will gives the 2x perfomance when compare to RTX2070 GPU
- First Gpu that supports HDMI 2.1 .
- All the latest GPU Support 8k monitor and we can connect upto 4 monitor.
- These GPU supports 3 gen Tensor cores so AI execution will be 2 x faster compare to previous gen processors.
official press release : https://www.nvidia.com/en-in/geforce/news/introducing-rtx-30-series-graphics-cards/