Other SKUs stay at 28 Gbps, but the NVIDIA GeForce RTX 5080 stands out with 30 Gbps GDDR7 Memory.
With its “Blackwell” GeForce RTX 5080 graphics card and its state-of-the-art GDDR7 memory technology, NVIDIA is almost ready to launch. On the other hand, 16 GB of GDDR7 memory operating at a remarkable 30 Gbps is anticipated to be included in the RTX 5080. When paired with a 256-bit memory bus, this setup will provide roughly 960 GB/s of bandwidth, which is 34% more than the RTX 4080’s 716.8 GB/s performance. However, the RTX 5080 will be the only card in the lineup with 30 Gbps RAM modules; other RTX 50 series models will include slightly slower 28 Gbps versions. The significant CUDA core difference between the RTX 5080 and RTX 5090, which is rumored, may be the cause of this strategic distinction.
By introducing a broader 512-bit memory bus, the flagship RTX 5090 is poised to push the envelope even farther and possibly reach bandwidths of over 1.7 TB/s. At least until higher-capacity GDDR7 modules hit the market, NVIDIA seems to be saving larger memory configurations of 16 GB+ for this top-tier model. Even with these remarkable specs, the bandwidth of the RTX 5080 is still about 5% less than that of the existing RTX 4090, which has a physically bigger bus design. To determine the what, how, and why of the Blackwell gaming GPUs, we must wait for the final launch. This performance difference between the 5080 and the projected 5090 indicates that NVIDIA is keeping a distinct hierarchy within its product stack.
