H100 NVL
Data center GPU platform for large language models and accelerated computing.
- Memory
- 94GB HBM3 per GPU
- Form factor
- PCIe dual-slot
- Interconnect
- NVLink
← Product directoryNV
NVIDIA
AI & accelerated compute. Browse source-aware reference records and add compatible models to one structured inquiry.
3 reference products · Availability on requestNVIDIAModels for
Models for
your shortlist.
NVIDIA and the NVIDIA logo are trademarks of NVIDIA Corporation. Public records remain attached to their source and review status.
H200 SXM
Hopper GPU for AI training, inference and high-performance computing.
- Memory
- HBM3e
- Form factor
- SXM
- Platform
- HGX H200 / certified systems
HGX B300
Eight-GPU Blackwell Ultra platform for AI factory and data center workloads.
- GPU configuration
- 8 × Blackwell Ultra SXM
- Total memory
- 2.1TB
- NVLink
- Fifth generation