NVIDIA GeForce RTX 5070 vs NVIDIA GeForce RTX 4070 SUPER
NVIDIA GeForce RTX 5070
NVIDIA GeForce RTX 4070 SUPER
NVIDIA GeForce RTX 5070 and NVIDIA GeForce RTX 4070 SUPER perform within a few percent of each other for gaming. Both sit in the same class, and our framerate data has each of them limited by the processor rather than the card itself, so neither holds a decisive lead. Choose on price, power draw and the features you actually use.
NVIDIA GeForce RTX 5070 vs NVIDIA GeForce RTX 4070 SUPER at a glance
Each axis is scored against the better of the two, so the outer edge is always the stronger part on that measure, not an absolute scale.
- Raster
- Memory
- Bandwidth
- Efficiency
- Value
NVIDIA GeForce RTX 5070 vs NVIDIA GeForce RTX 4070 SUPER benchmark scores
What is the difference between the NVIDIA GeForce RTX 5070 and NVIDIA GeForce RTX 4070 SUPER?
4 specifications separate the two cards:
- Memory 12 GB GDDR7 12 GB GDDR6X
- Bandwidth 672 GB/s 504 GB/s
- Board power 250 220
- Architecture Blackwell Ada Lovelace
Which one should you buy?
Choose the NVIDIA GeForce RTX 5070 if
- Cheaper at launch ($549)
Choose the NVIDIA GeForce RTX 4070 SUPER if
- Lower power draw (220 W)
How many FPS do the NVIDIA GeForce RTX 5070 and NVIDIA GeForce RTX 4070 SUPER get?
Estimated framerates with both parts measured against the same reference processor (a AMD Ryzen 7 9800X3D), so the graphics card is the only thing changing. Figures are ranges, not exact numbers. Measured at 1440p because at 1080p even a fast processor caps both cards and the columns come back identical.
NVIDIA GeForce RTX 5070 vs NVIDIA GeForce RTX 4070 SUPER: full specifications and benchmarks
| Measure | NVIDIA GeForce RTX 5070 | NVIDIA GeForce RTX 4070 SUPER |
|---|---|---|
| Headline | ||
| G3D Mark | 27,105 | 30,050 |
| Faster than | 94% of 268 | 96% of 268 |
| Performance | ||
| 3DMark Steel Nomad | 5,300 | 4,650 |
| 3DMark Time Spy | 22,267 | 20,671 |
| 3DMark Speed Way | - | 5,179 |
| Memory | ||
| Memory | 12 GB GDDR7 | 12 GB GDDR6X |
| Memory bus | 192-bit | 192-bit |
| Bandwidth | 672 GB/s | 504 GB/s |
| Clocks and power | ||
| Boost clock | 2512 MHz | 2475 MHz |
| FP32 | 30.9 TFLOPS | 35.5 TFLOPS |
| Board power | 250 W | 220 W |
| Suggested PSU | 650 W | 650 W |
| Power connectors | 1x 16-pin (12V-2x6) | 1x 16-pin / 2x 8-pin |
| Platform | ||
| Architecture | Blackwell | Ada Lovelace |
| Process | 4 nm | 5 nm |
| Price and release | ||
| Launch price | $549 | $599 |
| Released | 2025-03-05 | 2024-01-17 |
Which is better value?
At launch pricing, the NVIDIA GeForce RTX 4070 SUPER delivers more benchmark performance per dollar, by about 2%. Street prices move, so check current cost before deciding: a 2% value gap disappears quickly on discount.
Will the NVIDIA GeForce RTX 4070 SUPER bottleneck your build?
These figures hold one component constant. Your own build decides the answer: a faster graphics card buys nothing if the processor is already the limit.
How we compare graphics cards
Benchmark scores come from PassMark, 3DMark, Geekbench and Cinebench. Each of those suites is read from a single source across every part we compare, so a percentage gap here is a difference between the two chips rather than between two testers. Framerates come from our own model, built from measured GPU baselines and processor ceilings, and it declines to produce a number for hardware it does not cover rather than estimating one.