Most of the lithography that is dedicated to RAM is being done for HBM modules, which are not consumer grade. So more likely it will end up in landfills.
Chip designs take years, so if there’s a sudden glut of HBM, there’s no good way to put it to use outside of existing designs.
That being said, a lot of LPDDRX is being produced for Nvidia servers and a few other systems. That would be useful. Doubly so if we packaged as LPCAMM.
HBM are thinned die stacks which are assembled at the GPU periphery using silicon interposers. My AMD GPU has HBM.
In case of HBM overproduction post-bubble we might see resurgence of GPUs with HBM rather than GDDRx.
Most of the lithography that is dedicated to RAM is being done for HBM modules, which are not consumer grade. So more likely it will end up in landfills.
Or new consumer HBM modules
Chip designs take years, so if there’s a sudden glut of HBM, there’s no good way to put it to use outside of existing designs.
That being said, a lot of LPDDRX is being produced for Nvidia servers and a few other systems. That would be useful. Doubly so if we packaged as LPCAMM.
My AMD GPU has HBM.
These are giant dedicated HBM chips that are on the motherboard, they won’t be going into any GPUs.
HBM are thinned die stacks which are assembled at the GPU periphery using silicon interposers. My AMD GPU has HBM. In case of HBM overproduction post-bubble we might see resurgence of GPUs with HBM rather than GDDRx.