Nvidia and AMD will (probably) present these graphics cards at CES 2025

The current graphics card generations from Nvidia and AMD have been in circulation for longer than usual. The two manufacturers used to present and release new cards every other autumn. Now it's a few months later, but at least Nvidia is set to present four new models.

Traditionally, Nvidia presents the top models first - such as the upcoming RTX 5090 and RTX 5080. This time, however, the RTX 5070 Ti and RTX 5070 will also be presented. The reason for this is that AMD is not offering any graphics cards in the next generation that are comparable to the RTX 5080 or even RTX 5090. Nvidia probably wants to offer alternatives right from the start

RTX 5070 Ti and RTX 5070

It is unclear exactly when the RTX 5070 Ti and RTX 5070 will be released. However, it is assumed that the faster RTX 5070 Ti model will be launched first - more details are known about this card and the specifications are said to be finalised, which is not yet the case with the RTX 5070.

The 5070 Ti is said to be based on the GB203-300 GPU, which has 8960 Cuda cores. The memory is said to be 256-bit GDDR7 with a capacity of 16 gigabytes. The leaks contradict each other when it comes to power consumption: there is talk of 285 to 300 watts. However, these different figures could also be due to the fact that either Total Graphics Power or Total Board Power is meant. Both values could therefore be correct.

The presumed specifications compared to the predecessor:

The specifications suggest a whopping double-digit increase in performance. I think it's 20 per cent faster than its predecessor, which would put it in the RTX 4080 range.

The 5070, in turn, is said to be based on the GB205-300 GPU and have 6144 Cuda cores. The memory is said to be 192-bit GDDR7 with 12 gigabytes of capacity. The power consumption is assumed to be 250 watts.

The presumed specifications compared to the predecessor:

In contrast to the 5070 Ti, the performance increase of the 5070 looks smaller on paper. However, due to the new production and architecture, a double-digit increase in performance can still be expected. Perhaps it won't be 20 per cent more, but rather twelve to 15.

RTX 5080

As with the smaller models, the exact release date of the RTX 5080 is not yet known. However, the card is set to be the first in the series to be launched in mid-January.

There are already several leaks regarding specifications. For example, the RTX 5080 is said to have the fastest memory in the series, namely a whopping 32 GBit/s. The GB203-400 GPU is also said to have 10752 Cuda computing cores and the card is said to have a maximum power consumption of 400 watts.

The presumed specifications compared to the predecessor:

The RTX 5080 can therefore be expected to deliver a performance increase of around 20 per cent. Perhaps even more. Whether the card will even beat the RTX 4090 is unclear. According to some leaks, it is enough to outperform the card, but not according to others. What is clear is that in most games with the RTX 5080, more than 60 frames per second at 2160p resolution should be possible.

RTX 5090

The RTX 5090 should follow at the end of January or beginning of February as the second card in the series. It will offer by far the best performance - probably also at a high price. In return, it will offer over twice the number of Cuda cores of the RTX 5080: 21,760 in total. The GPU with the designation GB202-300 is said to measure 744 square millimetres. Compared to its predecessor, this would be an increase in surface area of 22 per cent.

The memory is said to be 512-bit GDDR7 with a capacity of 32 gigabytes and the maximum power consumption is said to be an insane 600 watts. The presumed specifications compared to the predecessor:

The RTX 5090 should clearly be more powerful than its predecessor. The more than 30 per cent more Cuda computing cores alone speak a clear language. But just that: The price is likely to be a bit steep.

RX 9070 XT and RX 9070

The most obvious change at AMD: The nomenclature has changed. The Radeon RX 8000 series is actually next in line. However, AMD is skipping this according to rumours and jumping straight to the RX-9000 series. Not only that, but the manufacturer is also said to be taking a new approach to naming the models. For example, they are to be called 9070 (XT) instead of 8800 (XT). This would bring AMD closer to Nvidia's designations.

Apart from the name change, AMD is doing little in terms of specifications for the time being. The RX 9070 XT is said to have 4096 computing units and 256-bit GDDR6 memory with a capacity of 16 gigabytes. It is based on the navi 48 XTX GPU. The maximum power consumption should not increase compared to its predecessor. The increase on paper is a modest one, as the comparison with the predecessor shows:

But even the RX 9070 based on the Navi 48 XT doesn't seem much more powerful on paper than its predecessor:

If the increase is modest on paper, it is according to leaks huge: The 9070 XT is said to be within five per cent of the RTX 4080. That would be a huge leap in performance of around 40 per cent. AMD is also said to have significantly improved the ray tracing performance of the RDNA4 architecture and the RX 9070 XT is said to be on a par with the current flagship RX 7900 XTX.

All information in this article should be taken with a grain of salt, as it has not been officially confirmed. The specifications will be presented by the manufacturers at CES and the performance data must first be confirmed by independent tests.