Head-to-head
GeForce RTX 4070 SUPER vs GeForce RTX 5070
Key differences
- RTX 5070 launched at $549 in the US, $50 less than the RTX 4070 SUPER ($599).
- RTX 5070 has 33% more memory bandwidth (672 GB/s vs 504 GB/s).
- RTX 5070 is rated for more power: 250 W against 220 W.
Specifications
| GeForce RTX 4070 SUPER | GeForce RTX 5070 | |
|---|---|---|
| Architecture | Ada Lovelace | Blackwell |
| Launch | 17 Jan 2024 | 5 Mar 2025 |
| US launch price | $599 | $549 |
| GPU / chip | AD104 | GB205 |
| Process | TSMC 4N | TSMC 4N |
| Transistors | 35.8 billion | 31.1 billion |
| Die size | 295 mm² | 263 mm² |
| Streaming multiprocessors | 56 | 48 |
| CUDA cores | 7,168 | 6,144 |
| Tensor Cores | 224 | 192 |
| RT cores | 56 | 48 |
| ROPs | 80 | 64 |
| Base clock | 1,980 MHz | 2,325 MHz |
| Boost clock | 2,475 MHz | 2,512 MHz |
| L2 / last-level cache | 48 MB | 48 MB |
| Capacity | 12 GB | 12 GB |
| Type | GDDR6X | GDDR7 |
| Bus width | 192-bit | 192-bit |
| Data rate | 21 Gbps | 28 Gbps |
| Bandwidth | 504 GB/s | 672 GB/s |
| FP32 (shader) | 35.5 TFLOPS | 30.9 TFLOPS |
| AI TOPS (vendor figure) | 568 | 988 |
| Power | 220 W | 250 W |
| Power connector | — | 16-pin 12V-2×6 |
| Host interface | PCIe 4.0 ×16 | PCIe 5.0 ×16 |
| Display outputs | 3× DisplayPort 1.4a, 1× HDMI 2.1 | 3× DisplayPort 2.1b, 1× HDMI 2.1b |
| NVENC | 1× 8th gen | 1× 9th gen |
Values in bold differ between the two. A dash means the spec does not apply to that brand.
Launch price by country
| RTX 4070 SUPER | RTX 5070 | |
|---|---|---|
| United States | $599 | $549 |
| India | — | ₹59,000 |
| United Kingdom | — | £539 |
| Eurozone (Germany) | — | €649 |
| Japan | — | ¥108,800 |
| Australia | — | A$1,109 |
Independent reviews
Not sure which to pick? Try the GPU finder. Full spec sheets: GeForce RTX 4070 SUPER · GeForce RTX 5070.
More comparisons: RTX 5070 vs RTX 5070 Ti · RX 9070 vs RTX 5070 · RTX 5060 Ti vs RTX 5070 · RX 9070 XT vs RTX 5070 · RX 9070 GRE vs RTX 5070