Nvidia has seemingly put its refreshed RTX 50 Super GPUs on hold due to GDDR7 shortages and, presumably, wild pricing. What little capacity is left is likely being siphoned by the RTX Pro and the AI server stack, leaving the consumer segment without a mid-cycle refresh. The hope is that Nvidia does indeed save some memory allocation for the Super Series.
According to @unikoshardware on X, Nvidia has cancelled the RTX 50 Super lineup due to the ongoing shortages of 3GB GDDR7 memory. This includes the RTX 5070 Super, RTX 5070 Ti Super, and RTX 5080 Super. Previous rumours have indicated that each of these is set to offer 50% more VRAM than their non-Super counterparts, without changing the rest of the specs, including the memory bus widths. Thus, the RTX 5070 Super was expected to pack 18GB (+6GB) of VRAM, followed by RTX 5070 Ti Super and RTX 5080 Super with 24GB (+8GB). Enough to ensure future proofness on upcoming games, at least regarding texture quality.
| RTX 5070 Super | RTX 5070 Ti Super | RTX 5080 Super | |
|---|---|---|---|
| CUDA Cores | 6,400 (+256) | 8,960 (+0) | 10,752 (+0) |
| Mem. capacity | 18GB (+6GB) | 24GB (+8GB) | 24GB (+8GB) |
| Mem. clock | 28Gb/s (+0Gb/s) | 28Gb/s (+0Gb/s) | 32Gb/s (+2Gb/s) |
| Board power | 275W (+25W) | 350W (+50W) | 415W (+55W) |
Higher-density 3GB memory chips are the only way to achieve this increase, unless Nvidia feels generous enough to double the capacity, in which case a clamshell design can be used in conjunction with 2GB GDDR7, allowing each two chips to share the data bus. Moreover, seeing how 2GB chips are also becoming expensive, the final price for such a solution will make it unappealing to many customers. As if this weren’t enough, the leaker has also suggested that non-Super cards using 2GB chips are at risk of becoming more expensive.
It is important to note that none of this is official; in fact, Nvidia has yet to announce the RTX 50 Super Series. Yet considering the growing cost of memory, even if they end up being released, they are unlikely to replace their non-Super equivalents’ prices as previously rumoured.
Nvidia is a $4.5 trillion company with a lot of weight in the industry, so it may find a way to get GDDR7 chips more cheaply. But again, Nvidia will likely use any capacity it has on its high-margin data centre AI GPUs. If these gaming cards are destined to see the light of day, we will probably get an announcement during CES 2026, so wait and see.
Regardless of the outcome, we will keep you informed, so make sure to follow the Club386 Google News feed to not miss any updates.
