The GeForce RTX 5090 will be quite unusual in that it will get a board made up of three separate components |
|
 |
|
|
According to insider kopite7kimi, we're talking about the main part of the board, which will house the GPU, an I/O board and a separate PCIe board, although the author notes here that perhaps this part should not be considered as a fully-fledged separate board. How exactly all this will be realized is not very clear, but, apparently, two or all three parts will be literally separate pieces of textolite connected in some way. Why this particular design is needed is unclear. Perhaps it's easier to design the board this way, given the huge and complex GPU  GPU-accelerated computing is the use of a graphics processing unit (GPU) together with a CPU to accelerate deep learning, analytics, and engineering applications. and the new GDDR7 memory that the adapter will get. In general, we have already seen variants with separate boards from Nvidia. We can recall old dual-chip "sandwiches" like GeForce 9800GX2 or the current gas pedal Hopper H100, where there is a small separate board for some elements. |
