A couple of months ago we reported on alleged RTX 3080 specs, these following a leak reported on by IgorsLAB. These specs listed the RTX 3080 Ti at 11GB GDDR6X memory, and the RTX 3080 at 10GB, with only the RTX “3090 Super/Ti/Titan” offering more than 20GB memory. Now, however, there’s a rumour that the entire top-end of Nvidia’s RTX 30-series lineup will offer much more memory than this first rumour indicated.
Wccftech reports that its “exclusive” sources have given pertinent information about the RTX 30-series lineup, including different graphics cards’ memory configurations. Because the names of the graphics cards using the different internal Nvidia Ampere GPUs are still unknown, we’re dealing with educated guesses based on previous names in the line.
The ‘long story short’ of this rumour (one that ComputerBase thinks is “credible”) is that we could be seeing 24GB, 20GB, and 10GB variants at the top-end of the lineup, 16GB and 8GB variants in the middle, and an 8GB SKU at the bottom-end. It’s at the top- and upper-middle-end, then, that we could see a significant increase in video memory over RTX 20-series graphics cards.
The board that is claimed to host the top-end GA102 GPU is the PG132, which is said to have three variants (PG132-10, PG132-20, and PG132-30). These variants will presumably be replacements for the RTX 2080 Ti and RTX 2080 Super. The PG142-0 and PG142-10 boards are reportedly replacements for the RTX 2070 Super, and the PG190-10 a replacement for the RTX 2060 Super. These are likely considered replacements for the Super variants because the RTX 20-series Super cards all but replaced the non-Super variants in this generation.
The apparent specs provided by this leak are as follows:
| * | Replacing | VRAM | Bus Width | Launch Window |
| PG132-10 | RTX 2080 Ti | 24GB | 384-bit | Second half of Sept. |
| PG132-20 | RTX 2080 Super | 20GB | 320-bit | First half of Sept. |
| PG132-30 | RTX 2080 Super | 10GB | 320-bit | Mid-Sept. |
| PG142-0 | RTX 2070 Super | 16GB | Unknown | Unknown |
| PG142-10 | RTX 2070 Super | 8GB | 256-bit | Second half of Sept. |
| PG190-10 | RTX 2060 Super | 8GB | 256-bit | Unknown |
*The details in this not table are not confirmed and are only rumour at this point.
Release windows for these graphics cards are also given by Wccftech. The PG132-30 board is said to launch in mid-September, the 24GB PG132-10 in the second half of September along with the PG142-10 board, and the PG132-20 board in the first half of October. It is apparently not known when the PG190-10 and PG142-0 boards will launch.
If this rumour is accurate then we should see some of the best graphics cards at the high-end offering much more video memory than in the current Nvidia Turing generation. Currently, we only see Titan RTX graphics cards offering memory capacity of 20GB and over, with the RTX 2080 Ti offering 11GB GDDR6 memory, and the RTX 2080 Super offering 8GB. This rumour would have these numbers doubled in some board variants in the new generation.
With 4K gaming, RTX ray tracing, and hefty game texture files becoming ever more popular, it makes sense that such capacious memory configurations would make their way onto the graphics card market, especially at the top-end where memory is likely to be more important, and with ray tracing being pushed more heavily. We’ll have to wait to see whether these exact configurations hit the market, though. If they do, it might be a sign that ray tracing is about to become a whole lot more mainstream, since this graphics tech is incredibly memory-intensive.