The NVIDIA GeForce RTX 3090 has not even hit the shelves yet and there is already speculation as to potential revisions. As NVIDIA continues to push the envelope in packing as much as possible into the flagship cards, it is ever riding the edge of what suppliers can provide for them. With the inclusion of GDDR6X memory, one such limitation was encountered.
OC3D has raised an interesting question about what could be next, and why. You see, the new NVIDIA GeForce 3090 is packing a walloping 24 GB of VRAM. One complication in providing it is that Micron is currently only able to supply 1 GB GDDR6X modules. So that means in order to facilitate this amount of memory, NVIDIA actually had to use 24 x 1 GB modules. As you can imagine, there are some logistical issues present for such a solution. For starters, it had to place modules on both sides of the PCB. This is not completely unheard of, but pretty rare. It also helps explain the size of this card. Another issue with so many modules involves power efficiency being decreased as it’s spread across to them. Let us not forget the area size to cool is greatly increased as well.
Shifts in Product Stacking
Consumers have never had a GPU with this much memory offered to them before. The RTX 30 Series cards do continue a trend with NVIDIA’s product stack shifting. The previous Turing generation saw the RTX 2080 Ti edge into TITAN-class territory. With the RTX 3090, we see a card that, in some ways, sits between the previous TITAN RTX and Quadro RTX offerings. Previously, they were the only RTX cards to feature that much memory. As much as some balk at the idea of spending upwards of $1,500 for an RTX 3090, it still pales in comparison to the $4,000 a Quadro RTX 6000 will set you back. The TITAN RTX ran for $2,499 last time around, just a little cheaper. Both cards feature 24 GB GDDR6.
Incoming 2 GB GDDR6X Modules
So what’s next? OC3D reports that Micron is hoping to have 2 GB GDDR6X modules available in 2021. A revision to the RTX 3090, or any RTX 30 Series card, could easily incorporate these. It would allow them to halve the number of modules, improve power efficiency, and theoretically lower prices more. By decreasing the amount of power needed for memory, it would also potentially allow more to be delivered to other parts of the card, thus increasing performance even further. It seems that NVIDIA’s generational leaps with the RTX 30 Series may not be over yet. SUPER variants, or whatever it chooses to call them moving forward, seem inevitable, and Micron’s next wave of 2 GB GDDR6X modules could play a big factor in it. As AMD’s Big Navi gets closer to release, this also goes to show how NVIDIA will probably have more cards to put on the table.