Nvidia’s leaked “RTX 5090 SE” GPU doesn’t make sense

Nvidia’s reportedly building an RTX 5090 SE GPU, but its specifications don’t make sense

According to a report from GAMEGPU, Nvidia is planning to create a GPU that bridges the gap between its RTX 5080 and RTX 5090. This GPU is reportedly called the “RTX 5090 SE”, and its alleged specifications do not make sense.

While an “RTX 5090 SE” with 14,080 CUDA cores, 32GB of memory, and a 500W TDP for $1500 sounds great on paper, its full specifications do not make sense. Simply put, you can’t make a 32GB graphics card with a 384-bit memory bus. This incorrect detail alone throws this entire rumour into question.

(RTX 5090 SE GPU specifications from GAMEGPU)

Why these specifications don’t make sense

With a 384-bit memory bus and 2GB GDDR7 memory modules, you can build a 24GB GPU. With 3GB GDDR7 modules, you can build a 36GB GPU. There is no possible configuration for building a 32GB GPU with a 384-bit memory bus without mixed memory densities. Since mixed memory densities can lead to an uneven performance profile, it doesn’t make sense for Nvidia to do this with such a high-end gaming product. In fact, Nvidia has never built a GPU like this.

Another thing is that the RTX 5090 SE name is also questionable. Why not call this alleged GPU the RTX 5080 Ti? Nvidia has built RTX xx80 Ti GPUs in the past, but has never created a lower-end RTX xx90 series variant. Why would Nvidia degrade its RTX 5090 branding with a lower-end variant? Between this and the nonsensical memory specifications, I find this GPU rumour hard to believe.

Nvidia is reportedly building RTX 50 SUPER series GPUs, and an RTX 5080 Ti would fill a big gap in Nvidia’s lineup. I just don’t buy this “SE” report.

You can join the discussion on Nvidia’s alleged RTX 5090 SE GPU on the OC3D Forums.