NVIDIA GeForce RTX 3070

The Nvidia GeForce RTX 3070 is a fast desktop graphics card based on the Ampere architecture. It uses the big GA104 chip and offers 5,888 cores and 8 GB GDDR6 graphics memory. 

The performance in games and 4k resolution is slightly below a RTX 2080 Ti but clearly faster then the old RTX 2080 Super. For most games, the performance is sufficient for 4k and maximum details. Enabling raytracing however (without DLSS) will lead to a significant lower performance (although the RTX performance was clearly improved compared to the RTX 2000 series).

The GA104 chip is produced in 8nm at Samsung and the power consumption of the RT 3070 is specified with 220 Watt (RTX 2070 Super - 215 Watt).

Manufacturer	NVIDIA
GeForce RTX 3000 Series	NVIDIA GeForce RTX 3090 10496 @ 1.4 - 1.7 GHz 384 Bit @ 19500 MHz NVIDIA GeForce RTX 3080 Ti 10240 @ 1.37 - 1.67 GHz 384 Bit @ 19000 MHz NVIDIA GeForce RTX 3080 8704 @ 1.45 - 1.71 GHz 320 Bit @ 19000 MHz NVIDIA GeForce RTX 3070 5888 @ 1.5 - 1.73 GHz 256 Bit @ 14000 MHz GeForce RTX 3080 Laptop GPU 6144 @ 0.78 - 1.25 GHz 256 Bit @ 14000 MHz NVIDIA GeForce RTX 3060 Ti 4864 @ 1.41 - 1.67 GHz 256 Bit @ 14000 MHz GeForce RTX 3070 Laptop GPU 5120 @ 0.78 - 1.29 GHz 256 Bit @ 14000 MHz NVIDIA GeForce RTX 3060 3584 @ 1.32 - 1.78 GHz 192 Bit @ 1875 MHz GeForce RTX 3080 Max-Q 6144 @ 0.78 - 1.25 GHz 256 Bit @ 6000 MHz GeForce RTX 3070 Max-Q 5120 @ 1.29 GHz 256 Bit GeForce RTX 3060 Laptop GPU 3840 @ 0.9 - 1.28 GHz 192 Bit @ 14000 MHz GeForce RTX 3060 Max-Q 3840 @ 1.28 GHz 192 Bit GeForce RTX 3050 Ti Laptop GPU 2560 @ 1.22 - 1.49 GHz 128 Bit @ 12000 MHz GeForce RTX 3050 Laptop GPU 2048 @ 1.24 - 1.5 GHz 128 Bit
Codename	Ampere GA104
Architecture	Ampere
Pipelines	5888 - unified
Core Speed	1500 - 1725 (Boost) MHz
Memory Speed	14000 MHz
Memory Bus Width	256 Bit
Memory Type	GDDR6
Max. Amount of Memory	8192 MB
Shared Memory	no
DirectX	DirectX 12_2, Shader 7.0
Transistor Count	17400 Million
technology	8 nm
Date of Announcement	16.09.2020

Benchmarks