Nvidia today announced the smallest member of the RTX 50-series, the GeForce RTX 5050. It will start at $249 and will be available in the second half of July.
The RTX 5050 features a new GB207 die based on the Blackwell architecture with 2560 CUDA cores, two thirds of the RTX 5060 that launched in April. It also features 5th generation Tensor cores and 4th generation RT cores like other 50-series cards. This enables it to support features like DLSS4 with multi-frame generation and the upcoming Reflex 2.
The RTX 5050 GPU is clocked at 2.31GHz base and boosts up to 2.57GHz. You get 128-bit 8GB GDDR6 memory, making it the only 50-series card so far without GDDR7. Other features include 3x DisplayPort 2.1b and 1x HDMI 2.1b connectivity, 1x 9th generation NVENC encoder and 6th generation NVDEC decoder. The card runs on 130W of power and can take either a single 8-pin connector or 300W or greater PCIe Gen 5 cable.
Nvidia has a couple of mostly pointless graphs showing the relative performance of the RTX 5050 compared to older Nvidia cards that make it hard to make anything out but the company claims the new card is 60% faster on average in raster than the RTX 3050 that came out over three years ago. The card also seems on average in the ballpark of RTX 4060 performance from two years ago.
The RTX 5050 should be on shelves this time next month. There is no Founders Edition model from Nvidia but the card will be available from all major board partners, including ASUS, Colorful, Gainward, Galaxy, GIGABYTE, INNO3D, MSI, Palit, PNY, and ZOTAC.
There is also going to be a mobile version of the RTX 5050, which is shipping today in laptops starting at $999.