GeForce RTX 5070 vs GeForce RTX 5090
GeForce RTX 5070
56#3 of 3 desktop GPUs · bottom 34%
Strong ray tracing at 250 W; 12 GB is tight for 4K textures.
vs
GeForce RTX 5090 wins by 42 points
GeForce RTX 5090
Winner98#1 of 3 desktop GPUs · top 34%
Fastest consumer GPU by a wide margin; 575 W and $1,999 say it all.
Overview
Key Differences
GeForce RTX 5070
- $1,450 cheaper at launch: $549 vs $1,999
- 57% lower rated power: 250 W vs 575 W
- ≈ estimated4 points higher Efficiency score: 96 vs 92
GeForce RTX 5090
- ≈ estimated142% higher 3DMark Speed Way (ray tracing): 14,500 vs 6,000
- ≈ estimated141% higher 3DMark Port Royal (ray tracing): 35,000 vs 14,500
- ≈ estimated140% higher Blender 4.x samples per minute: 12,500 vs 5,200
- ≈ estimated109% higher 3DMark Time Spy graphics: 47,000 vs 22,500
- ≈ estimated59 points higher Ray tracing score: 100 vs 41
Review
GeForce RTX 5090 scores 98 overall to GeForce RTX 5070's 56: #1 against #3 of 3 desktop GPUs.
GeForce RTX 5070 leads on Efficiency by 4% (96 (estimated) vs 92 (estimated)).
GeForce RTX 5090 leads on Raster by 108% (100 (estimated) vs 48 (estimated)) and Ray tracing by 144% (100 (estimated) vs 41 (estimated)).
GeForce RTX 5090 is ahead in 4 of 4 benchmarks, the widest gap in 3DMark Speed Way (ray tracing) by 142% (14,500 (estimated) vs 6,000 (estimated)).
GeForce RTX 5070 is rated at 250 W against 575 W; launch prices $549 and $1,999.
- Raster≈ estimated48≈ estimated100 +108%Traditional rendering: most games at high frame rates.
- Ray tracing≈ estimated41≈ estimated100 +144%Ray-traced lighting and reflections.
- Efficiency≈ estimated96 +4%≈ estimated92Raster performance per watt of board power.
- Overall5698
Benchmarks
- 3DMark Time Spy graphics≈ estimated22,500≈ estimated47,000 +109%
- 3DMark Speed Way (ray tracing)≈ estimated6,000≈ estimated14,500 +142%
- 3DMark Port Royal (ray tracing)≈ estimated14,500≈ estimated35,000 +141%
- Blender 4.x samples per minute≈ estimated5,200≈ estimated12,500 +140%
Specs
| Spec | GeForce RTX 5070 | GeForce RTX 5090 |
|---|---|---|
| CPU | ||
| Architecture | Blackwell (GB205) | Blackwell (GB202) |
| Shaders | 6,144 | 21,760 |
| RT cores | 48 (4th gen) | 170 (4th gen) |
| Tensor cores | 192 (5th gen) | 680 (5th gen) |
| Base / boost | 2.16 / 2.51 GHz | 2.01 / 2.41 GHz |
| GPU | ||
| L2 cache | 48 MB | 96 MB |
| ROPs | 80 | 176 |
| TMUs | 192 | 680 |
| Memory | ||
| Type | GDDR7 | GDDR7 |
| Capacity | 12 GB | 32 GB |
| Bus width | 192-bit | 512-bit |
| Bandwidth | 672 GB/s | 1,792 GB/s |
| Connectivity | ||
| Interface | PCIe 5.0 x16 | PCIe 5.0 x16 |
| Outputs | 3× DP 2.1b, 1× HDMI 2.1b | 3× DP 2.1b, 1× HDMI 2.1b |
| Power connector | 1× 16-pin | 1× 16-pin |
| Power | ||
| Total board power | 250 W | 575 W |
| Recommended PSU | 650 W | 1000 W |
| Misc | ||
| Node | TSMC 4N | TSMC 4N |
| Launched | March 2025 | January 2025 |
| Launch price | $549 | $1,999 |