The GeForce RTX 3070 is an upper-middle GPU that uses the GPU core "GA104" manufactured in the 8nm process based on NVIDIA's Ampere architecture. In the GeForce RTX 30 series, it is located below the GeForce RTX 3080.
The GA104 core of the GeForce RTX 3070 has 46 Streaming Multiprocessors enabled, with 5,888 CUDA cores, 184 3rd generation Tensor cores and 46 2nd generation RT cores available.
VRAM is equipped with GDDR6 memory operating at 14 Gbps. It is connected to the GPU with a 256-bit memory interface and achieves a bandwidth of 448 GB / s. The interface is PCI Express 4.0 x16, and TGP (Total Graphics Power), which is the power consumption index of the video card, is 220W.
[Table 1] Main specifications of GeForce RTX 3070 | |||
---|---|---|---|
GPU | GeForce RTX 3070 td> | GeForce RTX 3080 | GeForce RTX 3090 |
Architecture | Ampere (GA104) | Ampere (GA102) | Ampere (GA102) |
Manufacturing Process | 8nm | 8nm td> | 8nm |
Streaming Multiprocessor | 46 | 68 | 82< /td> |
CUDA core | 5,888 | 8,704 | 10,496 |
Tensor cores | 184 (3rd gen) | 272 (3rd gen) | 328 (3rd gen) |
RT core | 46 (2nd gen) | 68 (2nd gen) | 82 units (2nd generation) |
Texture units | 184 units | 272 units | 328 Groups |
ROP unit | 96 groups | 96 groups | 112 groups | tr>
Boost Clock | 1,725MHz | 1,710MHz | 1,695MHz |
Memory Capacity | 8GB (GDDR6) | 10GB (GDDR6X) | 24GB (GDDR6X) |
Memory Speed | 14.0Gbps | 19.0Gbps | 19.5Gbps |
Memory Interface | < td>256bit320bit | 384bit | |
Memory Bandwidth | 448GB/s | 760GB/s | 936GB/s |
PCI Express | PCIe 4.0 x16 | PCIe 4.0 x16 | PCIe 4.0 x16 |
Power Consumption (TGP) | 220W | 320W | 350W |