NVIDIA GeForce RTX 3070 vs NVIDIA GeForce RTX 4050 Laptop GPU

NVIDIA GeForce RTX 3070

The Nvidia GeForce RTX 3070 is a fast desktop graphics card based on the Ampere architecture. It uses the big GA104 chip and offers 5,888 cores and 8 GB GDDR6 graphics memory. 

The performance in games and 4k resolution is slightly below a RTX 2080 Ti but clearly faster then the old RTX 2080 Super. For most games, the performance is sufficient for 4k and maximum details. Enabling raytracing however (without DLSS) will lead to a significant lower performance (although the RTX performance was clearly improved compared to the RTX 2000 series).

The GA104 chip is produced in 8nm at Samsung and the power consumption of the RT 3070 is specified with 220 Watt (RTX 2070 Super - 215 Watt).

NVIDIA GeForce RTX 4050 Laptop GPU

The NVIDIA GeForce RTX 4050 Laptop GPU (codename GN21-X2) is a mid-range laptop GPU. It is based on the AD107 chip and uses the Ada Lovelace architecture. The RTX 4050 Laptop offers 2560 shaders and uses 6 GB GDDR6 dedicated graphics memory with a clock speed of 16 Gbps (effective) and a 96 Bit memory bus. The TGP (Total Graphics Power) can be configured between 35 and 115 Watt (+ 10-25 Watt Dynamic Boost from the CPU) leading to different performance levels. The boost clock depends on the TGP settings and ranges from 1605 MHz (35 Watt) to 2370 MHz (115 Watt). The performance therefore is also highly depending on the laptop and TDP settings.

In our benchmarks, the RTX 4050 stays slightly behind the older mobile RTX 3060 but is clearly faster than the mobile RTX 3050 Ti. Therefore, it is best suited for 1920x1080 gaming at high to ultra settings depending on the game (see benchmarks below).

The RTX 4050 offers 80 tensor cores and therefore supports DLSS 3 including frame generation that can have a big impact in supported games. The raytracing cores however will see less usage, as the RTX 4050 does not offer enough performance for high detail settings and raytracing in demanding games.

The AD107 chip is manufactured on 5nm at TSMC (4N process).