Nvidia's RTX-5070 is rumoured to be revealed at CES 2025, but I think people should calm down about its unconfirmed specs.

As we near a possible launch date, rumours and unconfirmed information about Nvidia's new-generation graphics cards are abounding. But the latest scuttlebutt has caused quite a stir. Nvidia is expected to unveil its next-generation consumer GPUs in January 2025 at CES. It won't be just the RTX5080 and RTX5090.

Wccftech reports that the RTX 5070 is set to debut at next year's show alongside its larger sibling. This would be a bit of a surprise as Nvidia usually shows off its high-end cards first and then a midrange card a few months (or even several months) later.

Wccftech has also reported some specs for the card that are interesting to read if they do come true. The main news is that the card will have 12 GB of VRAM, which has been confirmed by kopite7kimi. It also boasts a 192-bit bus and a memory bandwidth of 28 Gbps. This is a 33% improvement over the RTX 4700 and RTX 4700 Super GPUs.

This has caused a lot of online grumbling. Some have speculated that it's an attempt to push gamers to the higher-end cards (which are much more expensive), due to their higher VRAM loads. Some of the most demanding games use more than 12GB VRAM on 4K high settings. However, there are a few reasons why this may not be as big a deal as some people think.

Nvidia explained (in a tone that was distinctly weary) why it considers cache to be more important than the total VRAM of the 40-series graphics cards. This is because GPUs first access the L1 and the L2 caches before reaching the VRAM. Therefore, it has increased the L2 caches to get the best performance instead of simply cramming in as much VRAM on the board.

It's also worth mentioning that some games are programmed to use as much VRAM possible simply because it is an option. It's not that VRAM is not important, but that Nvidia doesn't think it's a priority for best gaming performance.

The rest of the specs indicate a maximum power draw of 250 W, which is a 50 W improvement over the RTX 4700. It seems to be right, since we expected the new cards would have higher power requirements than the previous generation. For reference, the RTX 5090 has been rumoured as having a peak rated draw of 600 W. The RTX 5080, on the other hand, has reportedly been bumped to 400 W from the 320 W rating the RTX 4080.

Is Nvidia worried about the backlash that could result from its perceived VRAM sparsity? This seems unlikely. AMD provided its current generation cards, like the RX7800XT, with large amounts VRAM for their respective market segments and they sold poorly. Nvidia released a 16 GB RTX 4060T version after a major VRAM controversy erupted in relation to the RTX 4006 Ti. It was more to show that the 16 GB RTX-4060 Ti did not improve gaming performance, despite its higher price.

Another thought is that the higher VRAM models could be held back for future Super and Ti versions. It's possible, but the price is what really matters.

The RTX 4070 was arguably the best mid-range card of this generation due to its $599 launch price, which was much more affordable than the higher-end 40-series cards. If Nvidia can match this pricing again, I'd bet on the RTX-5070 doing the same trick even if its VRAM specs are lower than expected.

This is dependent on a decent increase in performance over the previous card at the same price. Time will tell. CES 2025 is an odd event to announce a new Nvidia GPU, but rumours suggest AMD will unveil its next-gen RDNA4 GPUs at the time.

Could it be Nvidia trying to cover for AMD? It's unlikely, but it is possible, especially since AMD seems to be ignoring the high-end market this generation. However, there is always the possibility that AMD has a midrange stunner that Nvidia would like to counter with a midrange card. If these reports are true, we could be flooded with new graphics cards in January.