With the availability of 3GB GDDR7 chips or modules, the original plan for the GeForce RTX 50 Series was for NVIDIA to use these to provide a 50% increase in VRAM capacity with the arrival of the GeForce RTX 50 SUPER Series.
Of course, with the way things have gone regarding memory, storage, and the AI industry gobbling up all capacity, the SUPER series refresh has been put on indefinite hold. But it does look like NVIDIA will use 3GB GDDR7 memory with the RTX 50 Series, albeit not on models that could benefit from a capacity increase.
According to known insider and leaker MEGAsizeGPU on social media, NVIDIA is apparently prepping to release a GeForce RTX 5050 9GB graphics card that will use the new 3GB memory chips. Compared to the current GeForce RTX 5050 8GB, this new configuration would use a 96-bit memory interface and 28 Gbps memory speed. This would deliver a 12.5% increase in VRAM capacity and a 5% increase in overall bandwidth, to 336 GB/s.
Of course, the initial response to this announcement would be wanting to see NVIDIA's two 8GB RTX Blackwell GPUs, the GeForce RTX 5050 and GeForce RTX 5060, become 12GB models using the new 3GB memory dies. This would alleviate VRAM bottlenecks when playing some games at 1080p and 1440p, and make the GeForce RTX 5060 a true RTX 3060 successor.
On the GeForce RTX 5060 front, MEGAsizeGPU also notes that, due to supply constraints, NVIDIA plans to use cut-down variants of the larger 'GB205' GeForce RTX 5070 die for upcoming GeForce RTX 5060 cards. This larger die, with 6144 CUDA Cores, would see that number cut down to 3840 CUDA Cores. It's a strange but potentially understandable solution given current market issues. Still, this move could force NVIDIA's partners to adapt their existing GeForce RTX 5060 designs to accommodate a new PCB, or switch to RTX 5070 designs and adapt them to use the RTX 5060's traditional 8-pin power connector.




