NVIDIA is reportedly working on a new GeForce RTX 4070 SKU which would feature GDDR6 memory instead of the faster GDDR6X.
Last week, we heard reports that NVIDIA's GeForce RTX 4070 GPU SKUs were facing shortages for several reasons. One could say that these are attributed to the production and supply cuts on the RTX 40 "Ada" SKUs but there's also an impending GDDR6X shortage due to a quality check failure among one of the G6X batches from Micron.
There was speculation that NVIDIA might opt for GDDR6 memory to resolve the supply issues with GDDR6X memory and it looks like that might indeed be the case as reported by MEGAsizeGPU. The leaker states that NVIDIA's GeForce RTX 4070 will come in a new SKU with the PG141 SKU 347 PCB which will feature GDDR6 memory. All specifications will remain the same except the VRAM which might be configured at 20 Gbps.
As for the specifications, the NVIDIA GeForce RTX 4070 GPU features the AD104-250-A1 GPU with 5888 CUDA cores, 184 TMUs, and 64 ROPs. The GPU is clocked at a base frequency of 1920 MHz & a boost frequency of 2475 MHz. The GPU is configured with a 192-bit bus interface which features 12 GB of VRAM, up to 504.2 GB/s bandwidth (21 Gbps G6X modules). The card has a rated TDP of 200W.
The NVIDIA GeForce RTX 4070 and the GeForce RTX 4070 SUPER feature 21 Gbps GDDR6X memory, so 20 Gbps GDDR6 memory is a downgrade. If NVIDIA sticks with 20 Gbps speeds, then we are looking at 480 GB/s of bandwidth which is a 5% reduction over the GDDR6X models.
Bandwidth is crucial for the GeForce RTX 4070's Ada GPU; even slight reductions like these can impact gaming performance. The good thing is that RTX 4070 GPUs come with a full 48 MB L2 cache on the AD104 chip & alongside featuring several memory compression techniques, also allow decent memory
Read more on wccftech.com