Nvidia has been confusing us with its launch scheme and product specifications for several generations now, and today’s announcement is no different. After throwing an RTX 2060 12GB last week, more than three years after the arrival of the RTX 2000 line and more than one after year the arrival of the RTX 3000 line, the company continues to expand the RTX 2000 line with the appearance of a new RTX 2050, which will come exclusively to laptops and with quite strange specifications.
This RTX 2050 will feature 2048 Cuda Cores, which sounds like it has more Cuda Cores than the RTX 2060, which has 1920 Cuda Cores. However, this is because it is based on Ampere and not Turing, and Nvidia counts Ampere’s Cuda Cores as two instead of one and would equate to 1024 Turing Cuda Cores, effectively being well below the RTX 2060.
A significant limitation that this graphic will have is that it will have a bus of only 64 bits that limits the width to 112GB / s, which will strongly affect the performance of this new graphic. It uses a 64-bit bus, and it only has 4GB of GDDR6 VRAM, which is not bad for low-end graphics.
As indicated by Nvidia, this RTX 2050 has all the technologies of the RTX line, including Ray-Tracing in real-time and DLSS, so it will be interesting to see how it performs with RT activated when having so little bandwidth and a limit of power between 30W and 45W depending on the configuration chosen by the assembler. DLSS support could help get good FPS without RT activated, although, with RT started, the experience may not be the best even if we activate the AI rescaling technology.
This RTX 2050 will reach notebooks in the fall of 2022, so it will be a long time to see it on the market and know its performance. It will also arrive the new MX570 and MX550, whose specifications have not yet been revealed.