NVIDIA continues to work on the GeForce RTX 50, a new generation of graphics cards expected to launch later this year, which It could be based on the Blackwell architecture. I say could because it’s not actually confirmed yet and there’s always the possibility that what we saw at launch Hopper and Ada Lovelace.

Hopper is an architecture exclusive to NVIDIA’s professional solutions for servers, data centers, AI and other sectors where raw computing power and the power of artificial intelligence prevail. To give life to the GeForce RTX 40, the green giant created the Ada Lovelace architecture, so We cannot rule out that Blackwell remains exclusive to professional solutions, and that we see a different architecture in consumption.

We don’t have new details on this topic yet, but we do know that NVIDIA is already testing the new cooling systems it will use in the GeForce RTX 50 Founders Edition. These should retain the current line and design with little changes which would not significantly change its essence and will have enhancements that allow for dispersal more effective and efficient.

GeForce RTX 50: TGP between 120 and 600 watts

New information coming from Chinese sources ensures that NVIDIA has in its hands four different cooling solutions that have a high-end model power loss 600 wattsand the lowest model it can dissipate 250 watts without any problems. To put these numbers in context, the GeForce RTX 4070 SUPER has a TGP of 220 watts and the GeForce RTX 4090 has a TGP of 450 watts.

The source does not refer to a cooling solution for the lower power models, suggesting perhaps NVIDIA don’t release Founders Edition GeForce RTX 5060 and below, something that has already happened with the current generation. However, keep in mind that these are all rumors and nothing has been officially confirmed, so we can’t close any doors. In its information, it also states that AMD will not be able to launch the Radeon RX 8000 this year.

Based on all the information we’ve seen so far, I can share with you an estimate of how the GeForce RTX 50 series could be configured. in terms of consumption, bus and graphics memory:

• GeForce RTX 5090: 384-bit or 512-bit bus with GDDR7. 450W TGP.
• GeForce RTX 5080: 256-bit bus with GDDR7. TGP 320 or 350 watts.
• GeForce RTX 5070: 192-bit bus with GDDR7. TGP 220 or 225 watts.
• GeForce RTX 5060 Ti: 128-bit bus with GDDR6. TGP 150 or 160 watts.
• GeForce RTX 5060: 128-bit bus with GDDR6. TGP 120 or 125 watts.

In general, the TGP levels we can expect will be very similar to those of the GeForce RTX 40, but efficiency in terms of power per watt consumed will improve especially in the GeForce RTX 50, thanks to the improvements that NVIDIA will introduce at the architectural level, which will translate into higher raw power while maintaining the same power consumption, or otherwise increase it slightly.

AI generated cover image.