GeForce RTX 5080 With Blistering GDDR7 Could Top The 4090's Memory Bandwidth

As the end of the year quickly approaches, new tidbits of information start to indicate what NVIDIA's GeForce RTX 50 series may be like. While a 2024 release is unlikely, all hints point to a potential announcement at CES 2025 in January. One interesting rumor relating to VRAM is how the GeForce RTX 5080 will stack up against the current GeForce RTX 4090 flagship. 

Based on the GB203-400-A1 GPU die, the RTX 5080 will likely carry the same 16GB of VRAM as its predecessor, the GeForce RTX 4080 Super. With a 256-bit memory bus, it will purportedly feature GDDR7 VRAM. This is important, since GDDR7 has both speed and efficiency advantages over GDDR6X that can be significant. With a theoretical 32Gbps speed and assuming a 256-bit bus, we're looking at 1,024GB/s of memory bandwidth for the RTX 5080.

In theory, an RTX 5080 with a 256-bit bus and 1,024 GB/s of bandwidth would slightly edge out the RTX 4090. Even though the RTX 4090 has a fatter 384-bit bus, the slower memory (compared to GDDR7) leads to a bandwidth of 1,008 GB/s, just shy of the rumored RTX 5080. Call it a tie, if you'd like, it's still impressive tomorrow's second-highest SKU could match or slightly exceed today's top GPU in memory bandwidth.

As impressive as the RTX 5080 sounds, the RTX 5090 should easily clobber it with a massive 32GB of GDDR7 VRAM and a rumored 600-watt TDP. The RTX 5080 likely may fall around the 400-watt mark, which is an 80-watt increase over the RTX 4080 Super. More importantly, it is less than the 450-watt RTX 4090, a GPU many hope can be beaten by the RTX 5080 in performance metrics. 

According to the latest murmurs, we could see as many as three GPUs announced at CES 2025 by NVIDIA. The RTX 5080 and 5090 are mostly a given, but the GeForce RTX 5070 could also make an appearance. Rumored to have 12GB of VRAM, it will be using the Blackwell GB205 GPU die and likely 672GB/s of memory bandwidth with a 192-bit bus. Power draw should fall under 250 watts, making it more modest to run than its other RTX 50 series brethren. 

According to WCCFTech, "changes to the PCB back-drilling process" should also improve the signal between the GDDR7 VRAM and the GPU. It will be interesting to see how the confirmed specs match up (or don't) to the leaked specs, and just as importantly, what price points NVIDIA will set for its RTX 50 series GPUs.