We reported back in March that Nvidia was preparing an RTX 5050 desktop GPU with a GB207-300 chip featuring 2,560 CUDA cores and 8 GB of VRAM. Later leaks alleged an April release date for the card. When that didn’t pan out, information attributed to AIBs claimed a July 2025 launch for the RTX 5050.
In the midst of this debate around the possible specs and release date of the RTX 5050, leaks around the type of 8 GB VRAM have provided conflicting information. For instance, kopite7kimi claimed in March that the RTX 5050 would feature 8 GB of GDDR6 VRAM, while a leak from earlier this month claimed GDDR7 modules like the rest of the RTX 50 series.
Reliable leaker MEGAsizeGPU now reports that the RTX 5050 will feature 8 GB of GDDR6 VRAM running at 20 Gbps. This will put the RTX 5050 right in line with the RX 9060 XT, as the Radeon card also relies on 20 Gbps GDDR6 VRAM modules.
Assuming the RTX 5050 does pack GDDR6, it will be the only card in the entire RTX 50 lineup with GDDR6, since the rest of the series, from the RTX 5060 to the RTX 5090, uses GDDR7. This will undoubtedly hurt the bandwidth and, hence, the performance of the card.
That said, the presence of GDDR6 is not a surprise. The RTX 5050 is an entry-level GPU that is likely to come in under $250, and might even reach $200. So, it was always logical that Nvidia would use the cheaper GDDR6 in place of GDDR7.
As for performance, nothing certain can be said at this point. We’ve conjectured in the past that, with a CUDA core count similar to the RTX 4050 mobile GPU but with a 15 W higher TDP and 33.3% wider bus, we can see a notable performance gain. However, the final performance is unlikely to be anything exciting, as the last-gen RTX 4060 might still even be faster.