Nvidia reportedly cancels 24GB RTX 5080 due to high GDDR7 memory costs
NVIDIA is reportedly preparing a significant memory upgrade for the RTX 5080, moving from 16GB to 24GB of VRAM to better support local AI models and high-resolution textures. This shift comes amid a volatile market where high-end Blackwell GPUs are seeing extreme price fluctuations and supply constraints.
The transition to a 24GB RTX 5080 would be achieved by swapping existing 2GB GDDR7 modules for denser 3GB dies. Because the card utilizes a 256-bit bus with eight memory chips, this hardware swap increases capacity by 50% without requiring a wider bus or a complete GPU redesign. However, this increase in capacity does not inherently guarantee higher bandwidth; unless NVIDIA increases memory clocks, the 24GB variant would maintain the same bandwidth as the 16GB model. While some reports suggest this replacement could arrive as early as December, other rumors point toward a Q1 2027 release.
This memory bump appears increasingly necessary if reports regarding the top of the stack are accurate. Rumors indicate that NVIDIA may be discontinuing both the RTX 5090 and the China-specific RTX 5090 D v2, which would effectively position the RTX 5080 as the new flagship consumer card. Closing the memory gap between the 5080 and the 5090's reported 32GB pool would make the 80-class card a more viable long-term option, especially given that the RTX 60 series is reportedly delayed until 2028.
Hardware Specifications and Pricing
The technical leap for the Blackwell flagship is substantial, though it comes with significant power requirements. According to leaked specifications, the RTX 5090 features:
- 32GB of GDDR7 RAM on a 512-bit bus
- 21,760 CUDA cores
- 600W power draw
These performance gains are mirrored by steep pricing. Rumored MSRPs for the RTX 5080 are tipped at approximately $1,499, while the RTX 5090 could retail between $1,999 and $2,499. Some international leaks suggest even higher figures, with Chinese listings pointing toward a price of 18,999 yuan for the 5090—roughly $2,200 to $2,600 depending on VAT calculations—and 9,999 yuan for the 5080.
The discrepancy between rumored MSRPs and current market reality has created a chaotic pricing environment. While the RTX 5080 may have launched at $999, current retail listings on sites like Newegg and Micro Center show prices soaring between $1,899 and $1,979. This inflation extends down the stack; the RTX 5070 Ti, which launched at $749, is reportedly selling for at least $1,299 on NVIDIA's own storefront, while the RTX 5070 has climbed from $549 to $849.
This pricing volatility is compounded by a reported "RAM crisis" in 2026, which has hit high-performance memory supplies particularly hard. This shortage has not only driven up costs but has also led to the disappearance of the RTX 5080 from NVIDIA's GeForce Marketplace. While NVIDIA has stated the card is still available and the absence is due to regional supply constraints, the situation has fueled speculation that the 16GB model is being phased out in favor of the 24GB refresh.
The broader implications of these shifts suggest a strategy of extending the Blackwell lifecycle through incremental refreshes rather than a rapid transition to new architecture. With the RTX 60 series pushed back, NVIDIA is relying on VRAM bumps to keep its high-end cards relevant for AI and gaming workloads. This creates a high-stakes environment for the upcoming CES 2025 event, where NVIDIA's Blackwell updates will likely clash with the debut of AMD's RDNA 4 lineup.
