Based on GB207 silicon, the NVIDIA GeForce RTX 5050 has slower GDDR6 memory.
In July, NVIDIA plans to introduce the GeForce RTX 5050, a new lower mid-range graphics card SKU. The RTX 5050, which is positioned behind the RTX 5060, may aim to sell for less than $250 in order to compete with the Intel Arc B580. As we now know, NVIDIA is making design decisions that allow it to offer this card at a competitive price. One such decision is the use of outdated GDDR6 memory. The card will probably include a 128-bit memory interface with 8 GB of GDDR6 memory. Although we don’t yet know the memory speeds, we could venture that they are 18 Gbps, or 288 GB/s of memory bandwidth.
The smallest chip built on the GeForce Blackwell graphics architecture, the new “GB207” silicon, is also anticipated to make its debut and be fully utilized by the RTX 5050. 20 SM, 2,560 CUDA cores, 80 Tensor cores, 20 RT cores, 80 TMUs, and an unspecified number of ROPs are anticipated to be included in this chip. A total graphics power (TGP) value of 130 W is anticipated for the RTX 5050. Although we anticipate single 8-pin PCIe to be the norm, cards with 6-pin PCIe power connectors (75 W from connector, 75 W from PCIe slot) will be able to be constructed. The 130 W TGP will enable the development of ITX-friendly, small, or low-profile cards.
