Hopper-generation SXM accelerator with HBM3e, drop-in for HGX H100 baseboards.
NVIDIA's HBM3e refresh of the Hopper SXM module. 141 GB of GPU memory at 4.8 TB/s feeds the same Tensor Cores as H100, with up to 700 W TDP and NVLink. Pin-compatible with H100 HGX baseboards.
Specs
vram gb
141
compute
Hopper sm_90
tdp watts
700
bandwidth gbps
4800
const{jsx:e}=arguments[0];function _createMdxContent(n){const r={p:"p",...n.components};return e(r.p,{children:"Hopper memory refresh. Same compute, larger and faster GPU memory pool, drop-in for existing HGX servers."})}return{default:function(n={}){const{wrapper:r}=n.components||{};return r?e(r,{...n,children:e(_createMdxContent,{...n})}):_createMdxContent(n)}};