- Joined
- Aug 19, 2017
- Messages
- 2,896 (1.04/day)
Kioxia's LC9 Series 122.88 TB NVMe SSD was photographed on the show floor at NVIDIA's GTC 2025 by Tom's Hardware, revealing the first production-ready implementation of the company's BiCS FLASH generation 8 3D QLC technology with 2 Tb die density. The 2.5-inch form factor drive delivers immense storage density in the enterprise SSD market on a PCIe 5.0 connection. The LC9 Series features a dual-port design providing 32 GT/s throughput via PCIe 5.0 1x4 or 2x2 configurations, with NVMe 2.0 and NVMe-MI compliance. Built specifically to address density requirements for AI workloads, the drive leverages CMOS directly Bonded to Array (CBA) technology in its flash architecture. The enterprise-class unit is rated at 0.3 drive writes per day (DWPD) and 67,000 TBW endurance, positioning it for read-intensive applications rather than write-heavy workloads. The reported read speeds are around 15 GB/s, as demonstrated at the booth.
The LC9 supports vector database operations through Kioxia's AiSAQ technology implementation, which transfers vector database operations from DRAM to SSD storage. This approach significantly reduces system-level costs while maintaining acceptable latency profiles for AI RAG operations. Kioxia demonstrated the SSD at booth #1811 at the San Jose McEnery Convention Center through March 21, with live performance demonstrations highlighting its capabilities in AI dataset retrieval operations. The drive represents a critical advancement for enterprise AI infrastructure where data density and retrieval speeds are becoming bottlenecks in scaled LLM deployments. A massive capacity will allow entire enterprises to operate AI LLMs on top of company data with RAG and at low latency and high endurance.
View at TechPowerUp Main Site | Source
The LC9 supports vector database operations through Kioxia's AiSAQ technology implementation, which transfers vector database operations from DRAM to SSD storage. This approach significantly reduces system-level costs while maintaining acceptable latency profiles for AI RAG operations. Kioxia demonstrated the SSD at booth #1811 at the San Jose McEnery Convention Center through March 21, with live performance demonstrations highlighting its capabilities in AI dataset retrieval operations. The drive represents a critical advancement for enterprise AI infrastructure where data density and retrieval speeds are becoming bottlenecks in scaled LLM deployments. A massive capacity will allow entire enterprises to operate AI LLMs on top of company data with RAG and at low latency and high endurance.


View at TechPowerUp Main Site | Source