Nvidia RTX 5070, 5080 and 5090: Leaks reveal the latest data and the start of the GPU generation

Update from October 16, 2024: The wccftech website provides initial information about the RTX 5070. As expected, Nvidia is stingy with the video memory, which, like the 4070, is 12 GB and is connected to a 192-bit narrow interface. The only difference is the change from GDDR6 to GDDR7, which increases the memory bandwidth by around a third to up to 672 GB/s. The TGP should increase slightly from 220 to 250 watts.

The exact chip configuration of the 5070 has not yet been determined, but a total of up to 50 streaming processors with up to 6400 shader units will be available on the GB205.

Based on the current state of information, a performance assessment of the 5070 is dubious, as neither the number of shader units nor the GPU clock speed is certain. The performance increase compared to a 4070 with 5888 shader units shouldn’t be too big. Update end

Original message from September 30, 2024:
The leaker Kopite7kimi, known for his accurate predictions, has revealed on The presentation of the 5000 series has now also been specifically scheduled; Nvidia will present the graphics cards at the CES on January 6, 2025, and the first graphics cards with the RTX 5080 and 5090 should be available in February.

The most important change compared to the old information about the new top model 5090 is the expanded memory interface for a whopping 32 gigabytes of video memory, which should now be 512 bits (instead of 448 bits) wide. Nvidia has also increased the number of shader units to an incredible 21,760.

In fact, when fully expanded, Nvidia can equip the GB202-300-A1 graphics processor with 24,576 shader units, but the full computing power of the “David Blackwell” chip is probably reserved for professional cards that Nvidia can sell at a significantly higher price.

However, it is also possible that Nvidia will add an RTX 5090 Ti or Titan towards the end of the 5000 generation product cycle – if the competitive situation requires it, which is not very likely with PC graphics cards.

In comparison: RTX 4090 vs. RTX 5090

Unfortunately, along with the enormous increase in computing power, the Total Graphics Power (TGP), i.e. the maximum graphics card power consumption at maximum load of the new top model RTX 5090, also increases from 450 to a crazy 600 watts.

Performance rating: RTX 5090

The technical data of the Nvidia RTX 5090 promise a massive increase in performance compared to the already quite fast RTX 4090. We therefore assume that the RTX 5090 will be an expensive graphics card. When the RTX 4090 FE launched, the RRP was 1,949 euros, but Nvidia has now “lowered” the recommended retail price to 1,799 euros. For the significantly better equipped and faster RTX 5090, we therefore expect an RRP of around 2000 euros.

Since Nvidia has de facto monopolized the graphics card market with a share of 88 percent and is technically far ahead of the last remaining competitor AMD, Nvidia can practically roll the dice on the price for the 5090. You better prepare yourself for a hefty price increase for what will once again be the “fastest graphics card in the world”.

In comparison: RTX 4080 vs. RTX 5080

Kopite7kimi has also announced the expected technical data of the RTX 5080: The jump in performance compared to the 4080 predecessor is quite small: The number of streaming processors increases by around 10 percent from 76 to 84. The memory connection also remains the same, However, the 5080 also gets the faster GDDR7 video memory.

Compared to the top model, the power consumption of the RTX 5080 only increases moderately by 50 watts. How much cards with the RTX 5080 will cost is still up in the air.

Performance rating: RTX 5080

How powerful the RTX 5080 is is much more difficult to estimate. Technically there are only 10 percent more shader units, but the new graphics chip architecture should also calculate more efficiently and the faster video memory should also provide a few percent more performance. The bottom line is that you can expect a performance increase of between 15 and 25 percent.

However, the difference to the 5090 is much more noticeable, which ideally could quickly offer twice as much computing power as the 5080. There has never been such a big difference between an 80 and 90 chip at Nvidia. In this respect, the leaker’s information should be viewed with caution.