The Ada Lovelace-based NVIDIA RTX 40 series will include 5th Gen Max-Q technology, such as tri-speed memory controllers and ultra-low voltage GDDR6 memory. The architecture will additionally handle the most recent advancements in AI, namely DLSS 3 upscaling.
The business will introduce enthusiast RTX 4090 and RTX 4080 GPUs first, followed by mid-range RTX 4070, 4060, and 4050 graphics.
The laptop GPU for NVIDIA RTX 4090 has an AD103 GPU with 9728 CUDA cores. Desktop RTX 4080 card already uses this configuration. The RTX 4090 Ti GPU is a replacement for the Ampere-based RTX 3080 Ti GPU, which was limited to a 150W maximum TGP (note this does not include Dynamic Boost).
This card will provide 38.9 TFLOPS of single-precision computation capability, which is almost on par with desktop RTX 4070 Ti, according to NVIDIA. However, it is not even close to the twice as fast desktop RTX 4090.
The AD104 with 7424 CUDA cores and 12GB GDDR6 memory will be found in the RTX 4080 GPU. This model will be restricted to 150W TGP, just like the RTX 4090, but the entry-level version might come in at 60W TGP (so make sure to check the laptop specs before buying).
The AD106 GPU in the NVIDIA RTX 4070 Laptop GPU will have 4608 CUDA cores and run at up to 2.175 GHz. The operating range of this card is 35W to 115W, and the same range is applicable to laptops with RTX 4060 and RTX 4050 class graphics cards. The 4070 has 8GB of GDDR6 memory spread across a 128-bit memory bus, and the RTX 4060 also uses this setup.
The AD107 GPU will be found in the lowest tier of RTX 40 laptop GPUs from NVIDIA. While the RTX 4050 will launch with 2560 cores, the RTX 4060 will have 3072 CUDA cores. The latter has a 96-bit memory interface and 6GB of GDDR6 memory.
The price of the entry-level RTX 40 laptop GPUs was today disclosed by NVIDIA. Beginning on February 22, the less expensive 4070/4060/4050 class GPUs will be available, while the RTX 4080/4090 series will debut two weeks sooner (February 8th). The enthusiast laptops start at a cost higher than $1999.