Pics Reveal GDDR7 Modules on the GeForce RTX 5090 are From Samsung and Not Micron
Unlike previously thought, the GeForce RTX 5090 from the NVIDIA Blackwell lineup isn’t using memory from Micron but Samsung. Two reports have confirmed this and we have the naked PCB that proves this. Quasar Zone mentioned in its RTX 50 series key facts that the GDDR7 memory on the RTX 50 series is not from Micron but from Samsung.
This also indicates that other RTX 50 GPUs including the RTX 5090 are using GDDR7 memory from Samsung as well but this is just a speculation. The RTX 5090 is confirmed to use Samsung GDDR7 memory modules as confirmed by a local NVIDIA senior official, as per the report.
Another proof is the pic by @Geeky_Vaidy, who showed a clear PCB shot, revealing the RTX 5090 package and the memory surrounding it. The user said that RTX 5090 uses 16x 2GB memory modules from Samsung and the GPU went into production on 17th September 2024. For comparison, the user also posted a PCB of NVIDIA Drive AGX Thor, which also uses the Blackwell GPU but uses Micron memory.
The GeForce RTX 5090 brings 32GB total VRAM, the first to feature such a huge VRAM capacity with 16x 2GB memory modules. On the other hand, the RTX 5090 laptop GPU will feature 3GB GDDR7 memory modules to reach 24GB total VRAM. As far as NVIDIA’s switch from Micron to Samsung is concerned, this comes as a surprise because NVIDIA partnered with Micron previously to get GDDR6X modules for higher-end RTX 40 series GPUs.
This means, Samsung was able to win the contract due to various reasons. NVIDIA could have outsourced Samsung GDDR7 memory due to cost-effectiveness and availability. This also means fierce competition in the market and increasing pressure on Micron by losing the contract from a giant like NVIDIA.
