NVIDIA has reportedly moved up the launch of its upcoming entry-level graphics card to July 1st. The GeForce RTX 5050 will cover the lower end of the Blackwell GPU series and is expected to be available on the market for between 200 and 260 euros. The GB207 chip, used for the first time, has 2,560 CUDA cores of the latest Blackwell generation and is expected to ship with 8 GB of memory connected via a 128-bit interface.
Most likely, unlike its larger siblings, the card only uses GDDR6 instead of GDDR7 memory, which is why the data transfer rate should be around 288 GB/s. The computing power is roughly estimated to be around 12 FP32 TFLOPS.
The new RTX 50x0 cards finally offer 10-bit 4:2:2 decoding.
In contrast to the higher-priced GPUs of the Blackwell generation, the GeForce RTX 5050 will not be released as a Founders Edition. Instead, the release will be exclusively with custom models from board partners.
The RTX 5050 should be particularly exciting for price-conscious video users who want to use the new 
Nvidia hardware decoder for 10-bit 4:2:2 codecs under Windows. Anyone looking for a GPU with low power consumption may also find what they are looking for here, as the GeForce RTX 5050 is said to have a TDP of 100 watts. However, like any card over 70W, it will probably require an additional power connection from the power supply.
The cheapest Blackwell card to date is the RTX 5060 for currently around 300 euros, which then offers 448 GB/s memory throughput at 20 TFLOPS with 8 GB. For "normal" video effects up to 4K resolution, 8 GB of GPU memory is usually sufficient, but for very complex, locally calculated AI models, at least 16 GB should be the necessary lower limit in the near future. This requirement currently costs at least 450 euros at Nvidia with the 5060 Ti.
;)