Nvidia will reportedly present the GeForce RTX 5090, 5080 and 5070 in January
Nvidia will reportedly present the GeForce RTX 5090, 5080 and 5070 in January. Only the top model is expected to make a big leap in performance.
Various manufacturer cards of the GeForce RTX 4070 Ti.
(Image: c't)
Nvidia apparently wants to use the CES 2025 tech show in early January to announce the first three desktop graphics cards from the Blackwell generation: the GeForce RTX 5090, GeForce RTX 5080 and GeForce RTX 5070. A suitable opportunity would be the opening keynote by Nvidia CEO Jensen Huang on the night of January 6-7 at 3:30 am.
Meanwhile, the website Wccftech and some X accounts, which are considered to be well-informed, are busy sharing possible specifications. According to them, only the top model, the GeForce RTX 5090, is expected to see a big leap in performance. The graphics card is to be released with 21,760 shader cores and 32 GB of GDDR7 memory on a 512-bit interface. Compared to the GeForce RTX 4090, this would mean 33 percent more shaders and RAM.
GeForce RTX 5080 beaten
The current status of the smaller models is sobering. The GeForce RTX 5080 is to be released with 10,752 shaders and initially 16 GB of GDDR7 RAM on 256 data paths. That would be just 10.5 percent more processing units than the GeForce RTX 4080 and the same amount of memory.
There would hardly be any potential for more computing power, as the GeForce RTX 5080 is already supposed to use the full GB203 GPU. More shader cores would only be possible with a significantly slimmed-down version of the GB202 chip. As Nvidia continues to rely on mature 4-nanometer production technology from TSMC, this seems unrealistic.
The GeForce RTX 5090 is already supposed to use a partially deactivated variant – the full GPU would have 24,576 shaders, i.e. 13 percent more than the GeForce RTX 5090. The chip yield should be so high that there is no need for graphics cards with even more slimmed-down GPUs.
| Angebliche Spezifikationen der RTX 5000, basierend auf GerĂĽchten | |||
| Grafikkarte | GPU | Speicher | TBP |
| GeForce RTX 5090 | GB202, 21.760 Shader | 32 GByte GDDR7, 512 Bit, 1,8 TByte/s | 600W |
| GeForce RTX 5080 | GB203, 10.752 Shader | 16 GByte GDDR7, 256 Bit, 1 TByte/s | 400W |
| GeForce RTX 5070 | GB205, max. 6400 Shader | 12 GByte GDDR7, 192 Bit, 672 GByte/s | 250W |
Smallest plus for the GeForce RTX 5070
Nvidia is supposedly launching the GeForce RTX 5070 with the GB205. The graphics chip is said to house 6400 shaders. If the GeForce RTX 5070 uses the full configuration, this would be an increase of 8.7 percent compared to the GeForce RTX 4070. The memory capacity of 12 GB is not expected to change.
In addition, optimizations to the graphics architecture should improve performance in games. Higher GPU clock frequencies may also be possible. The GDDR7 memory chips with 28 to 32 gigabits per second and pin also significantly increase the transfer rate. The shader cores can thus load data from the RAM faster.
In the case of the GeForce RTX 5080, rumors have already circulated that Nvidia is planning a version with 24 GB of RAM without adjusting other specifications. This would be possible on the 256-bit interface with "crooked" 3 GB modules. Theoretically, a GeForce RTX 5070 with 16 GB would also be feasible.
As Nvidia is not switching to a new chip production generation, the company has to pay for the additional performance with higher electrical power consumption. TBP values of 600, 400 and 250 watts are being discussed for the GeForce RTX 5090, 5080 and 5070. TBP stands for Total Board Power and means the possible maximum. The real consumption could be lower.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmmung wird hier ein externer Preisvergleich (heise Preisvergleich) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (heise Preisvergleich) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
(mma)
