Nvidia GeForce RTX 4090 vs Nvidia GeForce RTX 3070 Ti
92
out 100
VS
49
out 100
Compared 2 processors: Nvidia GeForce RTX 4090, vs Nvidia GeForce RTX 3070 Ti. Below are specs, tests, pros and cons for each device
Review
Games
Performance in DirectX, OpenCL, Vulkan
AI/ML
AI workload testing
Professional Apps
3D modeling, video editing, rendering
Energy Efficiency
Power usage in various scenarios
Tests & specs
Specs & test comparisonGames
| Game title | 1080p (High) 1080p (Ultra) 1440p (Ultra) 4K (Ultra) | |
 Forza Horizon 5 | 277 199 191 154 | 156 102 93 68 |
 The Witcher 3 | 288 265 187 100 | 177 148 105 59 |
 Counter-Strike 2 | 329 244 185 91 | 261 198 140 81 |
 Far Cry 6 | 203 182 167 104 | 134 124 102 64 |
 Hogwarts Legacy | 201 164 132 76 | 104 85 65 36 |
 COD Modern Warfare 3 | 274 258 221 149 | 133 120 93 60 |
 Ghost of Tsushima | 183 154 137 91 | 94 79 63 36 |
 Cyberpunk 2077 | 227 208 143 68 | 115 102 63 29 |
 Shadow of the Tomb Raider | 304 277 215 119 | 178 165 125 68 |
| 1080p High | 254 | 150 |
| 1080p Ultra | 217 | 125 |
| 1440p Ultra | 175 | 94 |
| 4K Ultra | 106 | 56 |
| Deviation | Low | Low |
Benchmarks
| Time Spy | 36322 | 14873 |
| Solar Bay | 186930 | 69827 |
| Port Royal | 26115 | 8880 |
| Fire Strike | 72253 | 37227 |
| Night Raid | 195300 | 133622 |
| Background Blur | 318.9 frames/s | 197 frames/s |
| Face Detection | 222.8 frames/s | 128.1 frames/s |
| Horizon Detection | 14.2 Gpix/sec | 6.56 Gpix/sec |
| Edge Detection | 21 Gpix/sec | 8.81 Gpix/sec |
| Gaussian Blur | 23.9 Gpix/sec | 6.47 Gpix/sec |
| Feature Matching | 2.38 Gpix/sec | 1.26 Gpix/sec |
| Stereo Matching | 1950 Gpix/sec | 609.3 Gpix/sec |
| Particle Physics | 48090.4 fps | 18205 fps |
| API | OpenCL | OpenCL |
| G2D Mark | 1297 | 1051 |
| DirectX 11 | 329 fps | 192 fps |
| DirectX 12 | 150 fps | 91 fps |
| GPU Compute | 27111 Ops/s | 11708 Ops/s |
| Aztec Ruins (Normal) | 1617.6 fps | 865.5 fps |
| Car Chase | 1228.6 fps | 699.5 fps |
| Manhattan 3.1.1 (1440p) | 1234.6 fps | 610.7 fps |
| T-Rex | 6286 fps | 3191.9 fps |
| ALU 2 | 15073.4 fps | 4526.7 fps |
| Driver Overhead 2 | 303.7 fps | 288.3 fps |
| API | OpenGL | OpenGL |
| OS | Windows | Windows |
| Tessellation | 12119.1 | 4520.6 |
Tech Specs
Specs & test comparisonGeneral Info
| Manufacturer | Nvidia | Nvidia |
| Build | Discrete | Discrete |
| Announced | September, 20, 2022 | June, 10, 2021 |
| Form Factor | PC use | PC use |
| Use case | Gaming | Gaming |
| Segment | High-end | Mid-range |
| Architecture | Ada Lovelace | Ampere |
| Codename | AD102 | GA104 |
| Price | 1599 | 599 |
GPU
| GPU Frequency | 2235 MHz | 1575 MHz |
| Boost GPU Frequency | 2520 MHz | 1770 MHz |
| Shader Units | 16384 | 6144 |
| TMUs | 512 | 192 |
| ROPs | 176 | 96 |
| Execution Units | 128 | 48 |
| Ray Tracing Cores | 128 | 48 |
| Tensor Cores | 512 | 192 |
| L1 cache | 128 KB/core | 128 KB/core |
| L2 cache | 72 KB/core | 4000 KB/core |
| Instructions per Clock | 2 IPC | 2 IPC |
Performance
| Pixel Fill Rate | 444 Gpix/sec | 170 Gpix/sec |
| Texture Fill Rate | 1290 GTex/s | 340 GTex/s |
| FLOPS | 82.6 TFLOPS | 21.7 TFLOPS |
Process
| Interface | PCIe 4.0 x16 | PCIe 4.0 x16 |
| TGP | 450 W | 290 W |
| Process | TSMC | Samsung |
| Process tech. | 5 nm | 8 nm |
| Transistor Size | 609 mm² | 392 mm² |
| Transistor count | 76.3 billion | 17 billion |
| Transistor density | 125.29 MTr/mm² | 43.37 MTr/mm² |
Memory
| Interface | GDDR6X | GDDR6X |
| Memory Frequency | 2625 MHz | 1188 MHz |
| Bus | 384 bit | 256 bit |
| ECC support | No | No |
| Effective Memory Speed | 21000 MB/s | 19000 MB/s |
| Memory Size | 24 GB | 8 GB |
Max Throughput
API
| DirectX | 12 | 12 |
| Vulkan | 1.3 | 1.3 |
| OpenGL | 4.6 | 4.6 |
| CUDA | 8.9 | 8.6 |
| DisplayPort | 1.4a | 1.4a |
| Ray Tracing | Yes | Yes |
| DLSS | DLSS 3 | DLSS 2 |
Results
After analyzing the data, we consider Nvidia GeForce RTX 4090 the clear winner in this comparison.
Survey
Total Votes: 0
