GeForce RTX 4070 SUPER VS GeForce RTX 3090
We are comparing the GeForce RTX 4070 SUPER (NVIDIA) with the GeForce RTX 3090 (NVIDIA). Scroll down to the pros and cons of each graphics card, in-depth specs, and gaming benchmarks.
Last Updated: November 20, 2026
Overall Comparison
KEY DIFFERENCES OF GeForce RTX 4070 SUPER AND GeForce RTX 3090
Here’s how GeForce RTX 4070 SUPER and GeForce RTX 3090 stack up against each other in their most impactful areas.
Reasons to consider the GeForce RTX 4070 SUPER
-
3% higher 4K Ultra FPS (71 vs 69)
-
46% higher Boost Clock (2475 MHz vs 1695 MHz)
-
5% higher 3DMark Time Spy score (20885 vs 19898)
-
13% higher PassMark G3D Mark score (29962 vs 26570)
-
5% higher Blender score (6139 vs 5864)
Reasons to consider the GeForce RTX 3090
-
4% higher 3DMark Port Royal score (13633 vs 13122)
-
86% more memory bandwidth (936.2 GB/s vs 504.2 GB/s)
-
100% more VRAM (24GB vs 12GB)
-
100% higher Memory Bus Width (384 Bit vs 192 Bit)
-
1.5× higher Tensor Cores (328 vs 224)
SPECIFICATION TABLE COMPARISON OF GeForce RTX 4070 SUPER AND GeForce RTX 3090
Below is the detailed specification table comparing GeForce RTX 4070 SUPER and GeForce RTX 3090.
Note: Highlighted specs mean that GPU wins that spec row.
If neither side is highlighted, it is a tie or there is not enough data for both specs.
| Specification | GeForce RTX 4070 SUPER | GeForce RTX 3090 |
|---|---|---|
|
General Info
|
||
| GPU Name: | GeForce RTX 4070 SUPER | GeForce RTX 3090 |
| GPU Type: | Discrete | Discrete |
| Codename: | AD104 | GA102 |
| Vendor: | NVIDIA | NVIDIA |
| Foundry: | TSMC | Samsung |
| Architecture: | Ada Lovelace | Ampere |
| Generation: | GeForce 40 | GeForce 30 |
| Release Date: | January 8, 2024 | September 1, 2020 |
|
Board Design
|
||
| Bus Interface: | PCIe 4.0 x16 | PCIe 4.0 x16 |
| PCIe Version: | 4.0 | 4.0 |
| PCIe Lanes: | x16 | x16 |
| Process Size: | 5 nm | 8 nm |
| Transistors: | 35.8 billion | 28.3 billion |
| Density: | 121.8 MTr/mm² | 45.1 MTr/mm² |
| Die Size: | 294 mm² | 628 mm² |
| Board Length: | 267 mm | 336 mm |
| Board Width: | 42 mm | 61 mm |
| Board Height: | 112 mm | 140 mm |
| Total Graphics Power (TGP): | 220 W | 350 W |
|
Graphics Processor
|
||
| Compute Units: | 56 | 82 |
| Shading Units: | 7168 | 10496 |
| Texture Mapping Units (TMUs): | 224 | 328 |
| Render Output Units (ROPs): | 80 | 112 |
| Tensor Cores: | 224 | 328 |
| Ray Tracing Cores: | 56 | 82 |
| Base Clock: | 1980 MHz | 1395 MHz |
| Boost Clock: | 2475 MHz | 1695 MHz |
| L1 Cache: | 128 KB per SM | 128 KB per SM |
| L2 Cache: | 48 MB | 6 MB |
|
Raw Performance
|
||
| Pixel Fill Rate: | 198.0 GPixel/s | 189.8 GPixel/s |
| Texture Fill Rate: | 554.4 GTexel/s | 556.0 GTexel/s |
| FP32 Performance: | 35.48 TFLOPs | 35.58 TFLOPs |
|
Memory
|
||
| Memory Type: | GDDR6X | GDDR6X |
| Memory Size: | 12 GB | 24 GB |
| Memory Bus Width: | 192-bit | 384-bit |
| Memory Clock: | 1313 MHz | 1219 MHz |
| Effective Memory Speed: | 21000 Mbps | 19500 Mbps |
| Bandwidth: | 504.2 GB/s | 936.2 GB/s |
|
APIs
|
||
| DirectX: | 12.2 | 12.2 |
| Vulkan: | 1.4 | 1.4 |
| OpenGL: | 4.6 | 4.6 |
| OpenCL: | 3.0 | 3.0 |
| CUDA: | 8.9 | 8.6 |
| Ray Tracing: | Yes | Yes |
| DLSS: | DLSS 3 | DLSS 2 |
| DisplayPort: | 1.4a | 1.4a |
| HDMI: | 2.1 | 2.1 |
|
Gaming FPS
|
||
| 1080p Ultra (FPS): | 153 | 152 |
| 1440p Ultra (FPS): | 121 | 119 |
| 4K Ultra (FPS): | 71 | 69 |
|
Benchmarks
|
||
|
3D Mark
|
||
| Steel Nomad Lite | 21542 | 21870 |
| Time Spy | 20885 | 19898 |
| Solar Bay | 98157 | 97359 |
| Port Royal | 13122 | 13633 |
| Fire Strike | 50115 | 47513 |
| Wild Life Extreme | 41644 | 43662 |
| Night Raid | 165076 | 139745 |
|
GeekBench 6 OpenCL
|
||
| Compute Score | 206598 | 211669 |
| Background Blur | 359.5 | 320.8 |
| Face Detection | 210 | 207.2 |
| Horizon Detection | 7.61 | 10.1 |
| Edge Detection | 10.4 | 14.4 |
| Gaussian Blur | 11.3 | 11.5 |
| Feature Matching | 1.97 | 1.48 |
| Stereo Matching | 969.3 | 984.2 |
| Particle Physics | 29061.4 | 27764.9 |
|
Passmark Graphics
|
||
| G3D Mark | 29962 | 26570 |
| G2D Mark | 1193 | 1069 |
| DirectX 11 | 271 | 219 |
| DirectX 12 | 110 | 110 |
| GPU Compute | 16999 | 15155 |
|
GFXBench 5
|
||
| Aztec Ruins High (4K) | 220.10 | 247.50 |
| Aztec Ruins Normal | 1132.40 | 978.40 |
| Car Chase | 981.80 | 854.10 |
| Manhattan 3.1.1 (1440p) | 840.60 | 747.40 |
| T-Rex | 4325.30 | 3732.10 |
| Tessellation | 7090.80 | 6622.80 |
| ALU 2 | 6890.20 | 6650.90 |
| Driver Overhead 2 | 290.10 | 296.00 |
| API | OpenGL | OpenGL |
| OS | Windows | Windows |
|
Blender (Rendering)
|
||
| Blender Score | 6139 | 5864 |
|
GeekBench 6 ML
|
||
| Single Precision | 30706 | 27237 |
| Half Precision | 43851 | 41500 |
| Quantized | 23776 | 15927 |
| Image Classification | 12146 | 10665 |
| Image Segmentation | 23932 | 28963 |
| Image Super Resolution | 36321 | 22697 |
| Face Detection | 52836 | 55232 |
| Pose Estimation | 179788 | 128155 |
| Text Classification | 3371 | 3152 |
| Machine Translation | 5017 | 5088 |
| Object Detection | 15164 | 13891 |
| Depth Estimation | 62500 | 37984 |
| Style Transfer | 357489 | 272000 |
| Framework | ONNX | ONNX |
| Backend | DirectML | DirectML |
|
FurMark
|
||
| 1080p Vulkan | 14071 | 11604 |
| 1440p Vulkan | 9732 | 8504 |
| 2160p Vulkan | 4530 | 4684 |
| 1080p OpenGL | 14401 | 11467 |
| 1440p OpenGL | 10189 | 8557 |
| 2160p OpenGL | 4962 | 4804 |
Tip: ▲ indicates a higher-is-better spec advantage, while ▼ indicates a lower-is-better advantage.
PERFORMANCE CHART COMPARISON OF GeForce RTX 4070 SUPER AND GeForce RTX 3090
The charts below visualize how GeForce RTX 4070 SUPER and GeForce RTX 3090 stack up in key performance metrics.
Radar Chart Comparison
Bullet Chart Comparison
Value For Money Calculator
Prices use available launch or current data. Enter your own prices to see which GPU gives better value.
GeForce RTX 4070 SUPER
GeForce RTX 3090
Calculating...
Domain Score Summary
Scores are generated by Comparigon's scoring algorithm and are intended as a high-level guide. They may not map 1-to-1 with the individual spec-by-spec comparisons shown in the table.
Our Verdict
In conclusion, the GeForce RTX 4070 SUPER is clearly better overall compared to the GeForce RTX 3090. Its strongest advantages are in AI & Machine Learning, Hardware Architecture and Energy Efficiency. The GeForce RTX 3090 is the better fit if your top priority is Workstation.
Try to Compare Again One of These GPUs With Another GPU
Pick which GPU to keep, then search for the other GPU to compare.
Frequently Asked Questions about GeForce RTX 4070 SUPER vs GeForce RTX 3090
GeForce RTX 4070 SUPER scores 5% higher than GeForce RTX 3090 in 3DMark Time Spy (20,885 vs 19,898).
GeForce RTX 3090 scores 3.9% higher than GeForce RTX 4070 SUPER in 3DMark Port Royal (13,633 vs 13,122).
GeForce RTX 3090 scores 2.5% higher than GeForce RTX 4070 SUPER in Geekbench 6 OpenCL (211,669 vs 206,598).
GeForce RTX 3090 has 100% more graphics memory than GeForce RTX 4070 SUPER (24 GB vs 12 GB).
GeForce RTX 3090 has 85.7% higher memory bandwidth than GeForce RTX 4070 SUPER (936.2 GB/s vs 504.2 GB/s).
Both list GDDR6X for memory type.
GeForce RTX 4070 SUPER lists 220 W for graphics power, while GeForce RTX 3090 lists 350 W for graphics power.
GeForce RTX 4070 SUPER scores 4.7% higher than GeForce RTX 3090 in Blender (6,139 vs 5,864).
Both list PCIe 4.0 x16 for bus interface.
GeForce RTX 4070 SUPER was released on January 8, 2024. GeForce RTX 3090 was released on September 1, 2020. GeForce RTX 4070 SUPER was released 1224 days after GeForce RTX 3090.
Disclaimer: Specs, pricing, and benchmarks are sourced from official manufacturer data and verified test results. While we strive for precision, product details and pricing may change. Comparigon may earn an affiliate commission on qualifying purchases.