The specs for NVIDIA’s upcoming midrange Ampere graphics cards, the RTX 3070 and 3070 Ti have surfaced. As per a reliable source, the RTX 3070 will feature 2944 shaders and 8GB of GDDR6 memory while the 3070 Ti will come with 3072 cores and the same amount of GDDR6X memory.
Both the GPUs will be based on the GA104 die, GA104-400 for the RTX 3070 Ti, and GA-104-300 for the 3070. This is the second consumer die we’re hearing about after the GA102 which is expected to power the RTX 3080, 3080 Ti, and the 3090/Titan.
Let’s revisit the rumored specs of the higher-end Ampere GPUs:
| Chip | Model | Cores | Memory | Interface | TBP |
| GA102 | RTX 3090 | 5376 | 24 GB GDDR6X (Double-Sided) | 384-bt | 350 W |
| GA102 | RTX 3080 Ti | 5,248 | 11 GB GDDR6X | 352-bit | 350 W |
| GA102 | RTX 3080 | 4,352 | 10 GB GDDR6X | 320-bit | 320 W |
| GA104 | RTX 3070 Ti | 3,072 | 8GB GDDR6X | 256-bit | 250W |
| GA104 | RTX 3070 | 2,944 | 8GB GDDR6 | 256-bit | 220W |
The RTX 3080 and 3080 Ti sit at the top of the stack with 4,352 and 5,248 CUDA cores, respectively, with the Titan offering slightly more at 5,376 shaders. The latter has a 384-bit bus paired with 11GB of GDDR6X memory while the former comes with 10GB of GDDR6X memory across a 32-bit bus. Both GPUs are expected to have a TBP of 300W+ and the entire family will be fabbed on Samsung’s 8nm node:
- NVIDIA RTX 3080 Reportedly 30% Faster than the GeForce RTX 2080 Ti: Early Benchmarks
- NVIDIA RTX 3080, 3080 Ti, 3090 (Titan) w/ 10GB, 12GB & 24GB GDDR6X Memory Based on GA102 Die Likely
- NVIDIA’s RTX 30 Series GPUs to Leverage Samsung’s 8nm Process
