Nvidia RTX 50 Super GPUs are reportedly ready, but stuck in limbo due to excessive GDDR7 pricing — 3GB GDDR7 module costs triple the price of 2GB
The upcoming Super refresh of the Nvidia RTX 50-series GPU is reportedly on hold due to the high cost of 3GB GDDR7 memory chips. A VideoCardz source confirmed that one board partner already has RTX 50 Super GPUs on hand, but Nvidia has allegedly told the company that the products are on hold because of the price of 3GB GDDR7 memory chips.
This means that the AI GPU giant has already set an internal release date but is reportedly pushing it back because of memory pricing. If the cost of GDDR7 chips becomes too high, then the RTX 50 Super GPUs would either have a selling price that’s way above Nvidia’s targeted MSRP or, if it forces its partners to stick with or remain close to its set prices, GPU board manufacturers wouldn’t just make any units at all, as they’re going to lose money with every sale.
The RTX 50 Super GPUs are rumored to have 3GB GDDR7 chips, which offers 50% more capacity than the 2GB found in current-gen RTX 50-series graphics cards. This would allow the upcoming GPUs to have more memory without needing to increase or change their memory bus configurations.
According to the publication, the cards expected to be released soon include the RTX 5080 Super, RTX 5070 Ti Super, RTX 5070 Super, and RTX 5050 9GB. The first two will each receive 24GB of GDDR7 VRAM with a 256-bit bus width, while the RTX 5070 Super will have 18GB of VRAM with a 192-bit bus width. Unfortunately, these chips cost twice or thrice as much as their 2GB variants, which will likely push the retail price of these cards beyond Nvidia’s envisioned MSRP.
Nvidia used to supply VRAM chips alongside GPU dies to its board partners, but it changed this policy in late 2025 as the memory chip crisis unfolded. Because of this, the companies that complete the final assembly of the graphics cards are forced to source their own memory chips in an increasingly competitive market. SK hynix, one of the big three memory chip manufacturers, even says that 2027 is set to be the “worst year” for the memory shortage and said that the crunch will last until 2030.
Even Nvidia, one of the biggest winners in the AI race, has been affected by the RAMpocalypse, with the company not announcing a new GPU at CES 2026. This is the first time this has happened in five years, with Jensen Huang releasing the 30-series, 40-series, and their respective mid-generation refreshes despite supply chain limitations and several other issues that arose during that period. Its latest AI systems are now more expensive than ever, with memory accounting for 25% of the BOM, as costs have soared by nearly 500%.
It’s still unclear what Nvidia and its board partners plan to do about the memory situation, especially as things don’t seem to be improving. While it could delay the launch of the RTX 50 Super, it can only do so for so long, especially if it’s true that its dies are already in the hands of its board partners.
Follow Tom’s Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.
