Okay, so the NVIDIA GeForce RTX 5090 SE. Yeah, it’s coming. Reports from GameGPU, picked up by TechPowerUp and others, confirm this new card. It slots right in there, between the RTX 5080 and the flagship RTX 5090.

This is a mainline gaming GPU product, not some region-specific thing like the 5090D.It’s based on the Blackwell architecture, specifically a cut-down version of the GB202 die. The GB202 is the same silicon found in the standard RTX 5090. This GB202 die is a large chip, 750 mm² in area, with 92,200 million transistors. The RTX 5090 SE is rumored to feature 14,080 CUDA cores.

That’s a lot, like 33 percent fewer than the 21,760 CUDA cores in the RTX 5090. But it’s also a significant jump from the RTX 5080’s 10,752 cores. The card is tipped to include 32GB of GDDR7 memory. This memory is connected via a 384-bit memory bus. Now, this 32GB on a 384-bit bus, that’s interesting.

It would require a mixed density of 24 Gb and 16 Gb memory chips. NVIDIA has generally avoided mixing chip densities due to validation issues. It makes the alleged specifications difficult to slot into NVIDIA’s current lineup. Other rumored specifications include 110 streaming multiprocessors, 110 RT cores, and 440 Tensor cores. The total graphics power is expected to be around 500W.

This is about 12 percent lower than the RTX 5090’s 575W rating. A 1,000W power supply would likely be needed for a system with this card. The rumored MSRP for the RTX 5090 SE is around $1,500. That’s roughly 25 percent lower than the original RTX 5090’s $1,999 price tag. But, you know, whether gamers can actually buy it at MSRP is another question.

The RTX 5090 rarely sells for less than $3,000, sometimes over $4,000, because of high demand and limited supply. Even the RTX 5080, with its $999 MSRP, often goes for between $1,200 and $1,600. NVIDIA has really shifted focus to the AI accelerator market. This leads to persistent shortages of consumer GPUs. It’s frustrating for gamers.

The company’s stock, NVDA, increased by 4.03% today, July 10, 2026. This followed reports of relaxed US export controls to the UAE, which should help the UAE get advanced technologies, including AI semiconductors. That’s a big deal for them. There are also rumors about other Blackwell cards, like the RTX 5080 Super, the 5070 Ti Super, and the 5070 Super. These Super variants are expected to get a huge increase in memory capacity, with 3GB GDDR7 chips instead of 2GB chips.

The RTX 5080 Super is rumored to have 24GB of GDDR7 memory on a 256-bit bus. Its TGP could be 415W. The RTX 5070 Ti Super might get 24GB of GDDR7 memory and a 350W TGP.I mean, will this RTX 5090 SE actually help with the supply issues? Or is it just another high-end card that will be impossible to find at its suggested price?

It’s all about allocation, right?My personal trade, I bought NVDA on January 31, 2024, at $61.42. I’m planning to sell when it hits $300, or if there’s a major market correction and I need to rebalance. It’s been a ride. The Blackwell architecture was officially announced at NVIDIA’s GTC 2024 keynote on March 18, 2024. The focus there was mostly on datacenter accelerators, like the B100 and B200.

Gaming wasn’t really mentioned. The RTX 5090 was launched on January 30, 2025, with a 512-bit memory interface. The RTX 5080 was indicated to be the first 50 series card to reach consumers, with a planned release date of January 21, 2025.