NVIDIA RTX 50 Series flagship GPU Is Rumored To Feature 24 GB GDDR7 Memory
The NVIDIA RTX 50 series flagship GPU model is set to feature 24 GB GDDR7 memory, based on the latest rumor.
According to a well known NVIDIA leaker, kopite7kimi, the flagship model in the NVIDIA RTX 50 series lineup of GPUs will feature 24 GB GDDR7 memory. Additionally, they mentioned that the VRAM speed will be 28 Gbps with a 384-bit interface rather than 512-bit.
When compared to the current flagship GPU model from the NVIDIA RTX 40 series, the RTX 50 flagship model will offer a jump in speed from 21 Gbps to 28 Gbps with the implementation of GDDR7 in the new Blackwell architecture.
Based on the leaker’s information, the RTX 50 series GPU lineup should be configured as follows:
- GB202 – 512-bit interface, 24 GB memory, 1536 GB/s bandwidth
- GB203 – 384-bit interface, 16 GB memory, 1024 GB/s bandwidth
- GB204 – 256-bit interface, 12 GB memory, 768 GB/s bandwidth
- GB206 – 128-bit interface, 8 GB memory, 512 GB bandwidth
- GB207 – 128-bit interface, 8 GB memory, 512 GB/s bandwidth
Previously, the same leaker had shared that the Nvidia RTX 5000 series GPUs will include Display Port 2.1 support. It was also stated by them that these Blackwell GB20x architecture based graphics cards will use TSMC 3 nm process technology. However, it was not revealed whether it will be a custom 3N node or one of the available ones, such as N3E, N3P, and N3X.
Nvidia has yet to announce a release date for its Blackwell lineup of GPUs, though the company expects to begin shipping Blackwell HPC GPUs sometime late in 2024, which should be followed by the gaming series shortly after. The aforementioned Blackwell HPC series was already rumored to be using a 3 nm node, and, if true, the new rumor from kopite7kimi regarding the gaming series also using the same foundry would make sense.
The existing lineup of gaming GPUs from Nvidia, the GeForce 4000 series, succeeded the GeForce 3000 series. The RTX 4090 and RTX 4080 were released on October 12 and November 16, 2022, respectively, following the series’ announcement on September 20, 2022, at the GPU Technology Conference (GTC) 2022 event.
