The NVIDIA GeForce RTX 4070 Laptop GPU, codenamed GN21-X6, emerges as a high-end graphics solution tailored for laptops. Operating on the AD106 chip within the Ada Lovelace architecture, the RTX 4070 showcases 4,608 cores (a subset of the 7,680 in the AD106) and employs a 128-bit memory bus. The mobile variant is equipped with 8 GB GDDR6 dedicated graphics memory, boasting a clock speed of 16 Gbps (effective). The Total Graphics Power (TGP) is configurable within a range of 35 to 115 Watts, with an additional 15-25 Watt Dynamic Boost from the CPU, introducing diverse performance levels. For comparison, the desktop RTX 4070 Ti utilizes 285 Watts. The boost clock spans from 1230 MHz (at 35 Watt TGP) to 2175 MHz (115W), showcasing a performance dependency on the power settings employed for the RTX 4070 mobile in the laptop.
Expected to deliver performance below that of the desktop RTX 4070 Ti, the RTX 4070 is ideally suited for 2k gaming at maximum details, catering to the demands of 2022’s most demanding games.
Synthetic benchmarks of an average RTX 4070 mobile indicate a performance slightly below that of a mobile RTX 3080 Ti. However, reduced power limits significantly impact performance, with instances like a 60W TDP version achieving only 70% performance in Time Spy. In gaming scenarios, a robust RTX 4070 excels in Quad HD (QHD) gaming, where some highly demanding titles may fall short of the 60 fps threshold, potentially necessitating adjustments like running at 1080p, reduced settings, or utilizing DLSS. Refer below for detailed gaming benchmarks across various cards.
The RTX 4070 boasts 144 tensor cores (4th gen), enabling DLSS 3, including frame generation for supported games. Additionally, it features 36 raytracing cores (4rd gen) for immersive raytracing effects in supported games.
Manufactured on the 5nm process at TSMC (4N process), the AD106 chip showcases technological advancements, contributing to the performance and efficiency of the RTX 4070 Laptop GPU.
| Graphics Processor | |
|---|---|
| Chip (AD106) |  |
| Architecture | Ada Lovelace |
| Foundry | TSMC |
| Process Size | 5 nm |
| Transistor Count | 22.9 Billion |
| Density | 121.8M / mm² |
| Die Size | 188 mm² |
| Memory | |
|---|---|
| Memory Size | 8 GB |
| Memory Type | GDDR6 |
| Memory Bus | 128 bit |
| Bandwidth | 256.0 GB/s |
| Graphics Config | |
|---|---|
| Shading Units | 4608 |
| Base Clock/Boost Clock | 1395 MHz/1695 MHz |
| Memory Clock | 2000 MHz = 16 Gbps effective |
| TMUs/ROPs | 144/48 |
| SM Count | 36 |
| RT Cores/Tensor Cores | 36/144 |
| L1 Cache/L2 Cache | 128 KB (per SM)/32 MB |
| DirectX/OpenGL | 12 Ultimate (12_2)/4.6 |
| Theoretical Performance | |
|---|---|
| FP32 (float) | 15.62 TFLOPS |
| Pixel Rate | 81.36 GPixel/s |
| Texture Rate | 244.1 GTexel/s |
| 3DMark Time Spy Graphics | 11957 |
|---|---|
| 3DM11 Performance GPU | 41027 |
| Score | 52,984 |