GeForce RTX 5070 vs RTX 4090
VS
Nvidia GeForce RTX 5070
GB205-300
Nvidia GeForce RTX 4090
AD102
We compared two discrete desktop gaming GPUs: the GeForce RTX 5070 12 GB with 48 pipelines and 6144 shaders against the 2 years and 4 months older RTX RTX 4090 24 GB that utilizes 128 pipelines and 16384 shaders. Here you will find complete details about specs, efficiency, performance tests, and more.
Review
Gaming
Performance in DirectX, OpenCL, and Vulkan games
Workstation
Perf. in 3D modeling, video editing and rendering apps
AI/ML
Capabilities for machine learning and AI-related tasks
Energy Efficiency
Power consumption efficiency in different scenarios
NanoReview Final Score
Overall video card score
Value for money (Beta)
Key differences
Reasons to consider the GeForce RTX 5070
- Manufactured using a more efficient 4 nm process technology
- Supports Nvidia DLSS 4 technology
Reasons to consider the GeForce RTX 4090
- Performs significantly better (up to 82%) in 3DMark Steel Nomad Lite
- 2.7x higher maximum theoretical performance (82.6 vs 30.9 TFLOPS)
- Shows 38% higher average frame rate in modern games at QHD resolution – 177 vs 128 FPS
- Includes 12 GB more video memory
- Features 320 more tensor cores for effective ML and AI workloads
- Achieves 66% more points in the GeekBench 6 Compute test (316K vs 190K)
- Has 50% higher memory bandwidth: 1010 vs 672 GB/s
- Has 2.7x more shading units (16384 vs 6144)
Gaming Performance
Frame rate comparison across popular AAA titles at different resolutionsGames
| Forza Horizon 5 | 197 140 130 97 |
275 199 193 155 |
| The Witcher 3 | 216 183 131 71 |
292 268 191 101 |
| Counter-Strike 2 | 293 210 160 81 |
339 238 188 90 |
| Far Cry 6 | 164 145 128 82 |
203 179 170 104 |
| Hogwarts Legacy | 140 117 92 51 |
200 163 132 74 |
| CoD: Modern Warfare III | 178 171 136 89 |
266 263 219 151 |
| Ghost of Tsushima | 123 107 89 58 |
188 152 137 91 |
| Cyberpunk 2077 | 164 144 99 43 |
226 208 143 66 |
| Shadow of the Tomb Raider | 253 238 189 98 |
302 284 222 118 |
| 1080p High | 192 | 255 |
| 1080p Ultra | 162 | 217 |
| 1440p Ultra | 128 | 177 |
| 4K Ultra | 74 | 106 |
| Margin of Error | Low | Low |
Benchmarks
Graphics cards’ performance in recent benchmarking apps3D Mark
Steel Nomad Lite Score
| Time Spy | 22708 | 36299 |
| Solar Bay | 107004 | 187525 |
| Port Royal | 14498 | 26127 |
| Fire Strike | 58324 | 72432 |
| Wild Life Extreme | 43879 | 85218 |
| Night Raid | 172064 | 195813 |
GeekBench 6 OpenCL
GB6 Compute Score
| Background Blur | 254.4 img/sec | 318.9 img/sec |
| Face Detection | 180.6 img/sec | 222.8 img/sec |
| Horizon Detection | 7.91 Gpixels/sec | 14.2 Gpixels/sec |
| Edge Detection | 10.2 Gpixels/sec | 21 Gpixels/sec |
| Gaussian Blur | 10.7 Gpixels/sec | 23.9 Gpixels/sec |
| Feature Matching | 1.84 Gpixels/sec | 2.38 Gpixels/sec |
| Stereo Matching | 975.7 Gpixels/sec | 1950 Gpixels/sec |
| Particle Physics | 27816.5 FPS | 48090.4 FPS |
| API | OpenCL | OpenCL |
Cinebench 2024 GPU
Cinebench 2024 GPU
Passmark Graphics
G3D Mark Score
| G2D Mark | 1291 | 1303 |
| DirectX 11 | 280 FPS | 323 FPS |
| DirectX 12 | 106 FPS | 151 FPS |
| GPU Compute | 14310 Ops/s | 26022 Ops/s |
Blender
Blender GPU
Recent User Tests
GeForce RTX 5070
No benchmark results yet
GeForce RTX 4090
| Date | Benchmark | Result |
|---|---|---|
| 📘 2026-04-03 (afroman420IU) | Cinebench 2024 | 32578 |
Artificial Intelligence Tests
Performance in machine learning and artificial intelligence tasksGeekBench 6 ML
GB6 ML Single Precision
GB6 ML Half Precision
GB6 ML Quantized
| Image Classification (SP) | 12270 | 15600 |
| Image Segmentation (HP) | 30237 | 36657 |
| Image Super Resolution (Q) | 35585 | 48114 |
| Face Detection (HP) | 62662 | 78111 |
| Pose Estimation (Q) | 145190 | 289639 |
| Text Classification (SP) | 4198 | 4074 |
| Machine Translation (HP) | 6233 | 6349 |
| Object Detection (SP) | 16709 | 21186 |
| Depth Estimation (Q) | 58300 | 75599 |
| Style Transfer (SP) | 246598 | 601428 |
| Framework | ONNX | ONNX |
| Backend | DirectML | DirectML |
Specifications
Technical specifications of GeForce RTX 5070 and RTX 4090General
| Vendor | Nvidia | Nvidia |
| Build | Discrete | Discrete |
| Released | January 7, 2025 | September 20, 2022 |
| Launch price (MSRP) | $549 | $1599 |
| Case | Desktop | Desktop |
| Purpose | Gaming | Gaming |
| Segment | Mid-range | High-end |
| Architecture | Blackwell 2.0 | Ada Lovelace |
| GPU Codename | GB205-300 | AD102 |
| Rival Equivalent | - Radeon RX 9070 | - |
| Successor | - | - GeForce RTX 5090 |
| Recommended CPU | - Intel Core Ultra 5 245K or above | - Intel Core i9 14900K or above |
Desktop GPU rating (21st and 3rd place)
Graphics Processing Unit
| Base Clock | 2160 MHz | 2235 MHz |
| Boost Clock | 2512 MHz | 2520 MHz |
| Shading Units | 6144 | 16384 |
| Texture Mapping Units (TMUs) | 192 | 512 |
| Render Output Units (ROPs) | 80 | 176 |
| Compute Units (Pipelines) | 48 | 128 |
| Tensor Cores | 192 | 512 |
| Ray-tracing Cores | 48 | 128 |
| L1 Cache | 128KB per cluster | 128KB per cluster |
| L2 Cache | 48MB shared | 72MB shared |
| Instructions Per Cycle | 2 IPC | 2 IPC |
Raw Performance
| Pixel Fill Rate | 201 GPixel/s | 444 GPixel/s |
| Texture Fill Rate | 482 GTexel/s | 1290 GTexel/s |
FLOPS (FP32)
Physical
| Interface | PCIe 5.0 x16 | PCIe 4.0 x16 |
| TGP | 250 W | 450 W |
| Manufacturing | TSMC | TSMC |
| Fabrication Process | 4 nm | 5 nm |
| Die Size | 263 mm² | 609 mm² |
| Transistor Count | 31.1 billion | 76.3 billion |
| Transistor Density | 118.25 MTr/mm² | 125.29 MTr/mm² |
Memory
| Memory Type | GDDR7 | GDDR6X |
| Memory Size | 12 GB | 24 GB |
| Memory Clock | 1750 MHz | 2625 MHz |
| Effective Memory Speed | 28000 Mbps | 21000 Mbps |
| Bus | 192-bit | 384-bit |
| ECC | No | No |
Memory Bandwidth
API
| DirectX | 12.2 | 12 |
| Vulkan | 1.4 | 1.3 |
| OpenGL | 4.6 | 4.6 |
| OpenCL | 3.0 | 3.0 |
| CUDA | 12.0 | 8.9 |
| Ray Tracing | Yes | Yes |
| DLSS | DLSS 4 | DLSS 3 |
| DisplayPort | 2.1b | 1.4a |
Cast your vote
Total votes: 186