[Korean Media: Samsung Developing 8-Layer HBM4E to Meet NVIDIA's Requirements] BlockBeats News, August 28, according to the *Seoul Economic Daily*, Samsung Electronics is developing 8-layer HBM4E products as per NVIDIA's requirements, intended for NVIDIA's custom high-bandwidth memory NVHBM. This plan reduces the number of stacked layers compared to Samsung's originally prepared 12-layer and 16-layer products. NVIDIA's target speed is 17 to 18Gbps, approximately 20% higher than Samsung's earlier HBM4E samples at 14.4Gbps. The 8-layer design can lower the difficulty of wafer thinning, precision stacking, and backend packaging, alleviating yield pressure and improving supply capacity. NVHBM may be applied starting with the Rubin Ultra AI GPU, expected to launch next year. Rubin Ultra's GPU interconnect scale will expand from the current 72 units to up to 576 units, using more high-speed GPUs to compensate for the reduced storage capacity of individual GPUs. Samsung possesses design and production capabilities for DRAM, logic chips, and foundational bare dies, enabling it to provide one-stop services for custom HBM. Industry insiders believe Samsung has already demonstrated high transmission speeds with HBM4, and as NVIDIA shifts its competitive focus from high stacking to high-speed transmission, Samsung's related capabilities may garner more attention. [Original Link]