NVIDIA GeForce RTX 50 “Not Much Different” To RTX 40’s Memory Configuration: Leaker
NVIDIA GeForce RTX 50 “Not Much Different” To RTX 40’s Memory Configuration: Leaker
A while ago, word on the street went that NVIDIA’s GeForce RTX 50 series GPUs may finally get a big spec bump in the memory department, with a 512-bit memory bus coming into play for the next generation of GPUs. However, as we’re getting closer to the launch (2025 at the earliest), new claims from @kopite7kimi have pulled the brakes on such a possibility.
The leaker says all the gaming GPUs for the next-gen, which is expected to use the GB202 chip at its most powerful configuration, will largely stick to the existing hierarchy established on the RTX 40 series. Meaning, that the flagship RTX 50 GPU will still be 384-bit wide. “Although I still have fantasies about 512-bit, the memory interface configuration of GB20x is not much different from that of AD10x,” the leaker said on X (Twitter).
Although I still have fantasies about 512 bit, the memory interface configuration of GB20x is not much different from that of AD10x.
— kopite7kimi (@kopite7kimi) March 9, 2024
So, there goes the possibility for more VRAM and we’re once again locked into the state of low memory capacities for the price? Hold your horses. The leaker is inclined to believe that Team Green is waiting for higher-density 24Gb memory packages to become available – meaning, the 128-bit memory bus should pack 12GB of VRAM.
While not explicitly mentioned, GDDR7 memory is a possibility as it will be the first VRAM generation to support 24Gb packages via PAM3 signaling, compared to the traditional NRZ method of data signaling. With the upgrade in memory generation, we can still see increased memory bandwidth overall via the new modules, which can go as high as 32Gbps per chip.
Source: Videocardz
Pokdepinion: Pair 384-bit with 24Gb memory packages – would NVIDIA actually put 36GB on VRAM on their next flagship? We shall see.