First information from RTX 5090 shared
Kopite7kimi today shared new details about the Blackwell series, the code name of NVIDIA’s next generation GPU series. However, it is now expected that the Blackwell architecture will cover both the data center and gaming series. At this point, Nvidia will use different naming schemes for the two GPU series; GB1XX (HPC) for high-performance computing and GB2XX for gaming.
GeForce RTX 5090 / RTX 4090
- 50% scale increase (probably core)
- 52% higher memory bandwidth increase
- 78% higher cache (probably L2)
- 15% higher clock speed
- 1.7 times higher performance
At this point, considering that the RTX 4090 has 21 Gbps memory, the 5090 will most likely come up with 32Gbps speeds. Therefore, we can say that this claim refers to GDDR7 memories. There are also rumors that the new card has a 512-bit bus. However, this configuration can be left to higher versions (such as Titan/RTX or RTX 5090 Ti). Niketim may not be able to switch to GDDR7 memory from the first day. Therefore, RTX 5090 configurations such as 512 bit/24 Gbps or 448 bit/28 Gbps can also be considered.
Considering that other claims are also based on the RTX 4090, the 15 percent increase in frequency indicates clock speeds of 3GHz or higher. In the cache, 128MB L2 sizes will be reached. We are more than a year away from the potential launch of the next-generation GeForce series. At this point, it should not be forgotten that the information shared is a leak. However, we think that more information will be shared in the process ahead.