[ GEFORCE ]
Graphics Cards
GeForce RTX 4070 SUPER
Overall Score
62/100
Key Technical Specifications
Graphics (GPU)
Ada Lovelace
[ EXPERT VERDICT · SUMMARY ]
GeForce RTX 4070 SUPER is an engineering-tested graphics cards achieving an overall rating of 62/100. It is positioned for balanced everyday compute and power efficiency.
Benchmark Performance Scores
Normalized 0–100 Scale
Efficiency
Rating: Poor
45/100
Gaming
Rating: Average
64/100
Detailed Technical Specifications
3D Mark
| Specification | Value |
|---|---|
| Fire Strike | 50092 |
| Night Raid | 164614 |
| Port Royal | 13123 |
| Solar Bay | 98188 |
| Time Spy | 20879 |
| Wild Life Extreme | 41625 |
API
| Specification | Value |
|---|---|
| CUDA | 8.9 |
| DLSS | DLSS 3 |
| DirectX | 12 |
| DisplayPort | 1.4a |
| OpenCL | 3.0 |
| OpenGL | 4.6 |
| Ray Tracing | Yes |
| Vulkan | 1.3 |
GeekBench 6 ML
| Specification | Value |
|---|---|
| Backend | DirectML |
| Depth Estimation (Q) | 62500 |
| Face Detection (HP) | 52836 |
| Framework | ONNX |
| Image Classification (SP) | 12146 |
| Image Segmentation (HP) | 23932 |
| Image Super Resolution (Q) | 36321 |
| Machine Translation (HP) | 5017 |
| Object Detection (SP) | 15164 |
| Pose Estimation (Q) | 179788 |
| Style Transfer (SP) | 357489 |
| Text Classification (SP) | 3371 |
GeekBench 6 OpenCL
| Specification | Value |
|---|---|
| API | OpenCL |
| Background Blur | 359.5 img/sec |
| Edge Detection | 10.4 Gpixels/sec |
| Face Detection | 210 img/sec |
| Feature Matching | 1.97 Gpixels/sec |
| Gaussian Blur | 11.3 Gpixels/sec |
| Horizon Detection | 7.61 Gpixels/sec |
| Particle Physics | 29061.4 FPS |
| Stereo Matching | 969.3 Gpixels/sec |
General
| Specification | Value |
|---|---|
| AMD Equivalent | -Radeon RX 7900 XT |
| Architecture | Ada Lovelace |
| Build | Discrete |
| Case | Desktop |
| GPU Codename | AD104 |
| Launch price (MSRP) | $599 |
| Purpose | Gaming |
| Recommended CPU | -Intel Core i7 14700Kor above |
| Released | January 17, 2024 |
| Segment | Mid-range |
| Vendor | Nvidia |
Graphics Processing Unit
| Specification | Value |
|---|---|
| Base Clock | 1980 MHz |
| Boost Clock | 2475 MHz |
| Compute Units (Pipelines) | 56 |
| Instructions Per Cycle | 2 IPC |
| L1 Cache | 128KB per cluster |
| L2 Cache | 48MB shared |
| Ray-tracing Cores | 56 |
| Render Output Units (ROPs) | 80 |
| Shading Units | 7168 |
| Tensor Cores | 224 |
| Texture Mapping Units (TMUs) | 224 |
Memory
| Specification | Value |
|---|---|
| Bus | 192-bit |
| ECC | No |
| Effective Memory Speed | 21000 Mbps |
| Memory Clock | 1313 MHz |
| Memory Size | 12 GB |
| Memory Type | GDDR6X |
Passmark Graphics
| Specification | Value |
|---|---|
| DirectX 11 | 271 FPS |
| DirectX 12 | 110 FPS |
| G2D Mark | 1193 |
| GPU Compute | 16926 Ops/s |
Physical
| Specification | Value |
|---|---|
| Die Size | 294 mm² |
| Fabrication Process | 5 nm |
| Interface | PCIe 4.0 x16 |
| Manufacturing | TSMC |
| TGP | 220 W |
| Transistor Count | 35 billion |
| Transistor Density | 119.05 MTr/mm² |
Raw Performance
| Specification | Value |
|---|---|
| Pixel Fill Rate | 198 GPixel/s |
| Texture Fill Rate | 554 GTexel/s |
Recent User Tests
| Specification | Value |
|---|---|
| Date | Benchmark |
| 📘 2025-11-28(Caleb) | GeekBench |
| 📘 2025-12-08(Rickey) | Cinebench 2024 |
Strengths & Limitations
Key Advantages
- Comprehensive positive benchmark metrics recorded.
Key Considerations
- No major hardware limitations flagged during benchmark run.
Direct Competitors & Alternatives
NVIDIA
Nvidia RTX Pro 6000 Blackwell
98/100
Compare →
GEFORCE
GeForce RTX 5090
96/100
Compare →
GEFORCE
GeForce RTX 4090
89/100
Compare →
AMD
AMD Radeon RX 7900 XTX
85/100
Compare →
Methodology, Provenance & Verification
Score version 2026.1 · Data verified 2026-08-26 · Source: official manufacturer datasheets and standardized reproducible benchmark suites (Geekbench 6, Cinebench R24, 3DMark).