Qualcomm Cloud AI 100 Ultra
$8,910
The Qualcomm Cloud AI 100 Ultra is a high-performance AI inference card designed for generative AI and large language models. The Cloud AI 100 Ultra is built around Qualcomm’s seventh-generation AI core, which uses a separation-of-concerns compute architecture with dedicated units for tensor, vector, and scalar operations.
Description
| Model | Cloud AI 100 Ultra |
| Purpose | AI Inference accelerator for Generative AI, LLMs, Computer Vision, NLP |
| AI Cores per SoC | 16 seventh-generation AI cores |
| AI Cores per Card | 64 AI cores |
| On-Chip SRAM | 576 MB total (144 MB per SoC) |
| On-Card DRAM | 128 GB LPDDR4x with ECC |
| DRAM Bandwidth | 548 GB/s |
| Peak INT8 Performance | Up to 870 TOPS |
| Peak FP16 Performance | Up to 290 TFLOPS |
| Supported Data Types | FP16, INT16, INT8, FP32 |
| Host Interface | PCIe Gen4 x16 |
| Form Factor | PCIe FH3/4L (Full Height, 3/4 Length) |
| TDP | 150 W |
| Process Node | 7 nm |
| LLM Support | 100B params on 1 card; 175B params on 2 cards |
Additional information
| Weight | 1.1 kg |
|---|---|
| Condition | NEW |
