KIOXIA LC9 debuts with 122TB of QLC NAND, PCIe Gen5, and dual-port support, built for AI, LLMs, and cloud storage.
The high-capacity enterprise SSD market just got a new contender with the KIOXIA LC9 Series NVMe SSDs being announced today. The drive features capacities up to 122.88TB and is built on KIOXIA’s generation 8 3D flash memory technology QLC 2 terabit (Tb) die. As with all other enterprise SSDs, the LC9 is designed to address the growing storage needs of AI, large language models (LLMs), and vector databases. As these workloads demand more efficient storage solutions, the KIOXIA LC9 SSD aims to optimize performance while reducing power consumption per terabyte.
Breaking Down the KIOXIA LC9 SSD
The KIOXIA LC9 SSD uses a 2.5-inch U.2 form factor, which is critical when stuffing so much NAND into a single SSD. Vendors selling 122.88TB SSDs today require the real estate that twin PCBs can offer in the 15mm z-height SSD body. The SSD supports PCIe Gen5 (1×4 or 2×2) and NVMe 2.0, bringing a potential performance uplift over Gen4 drives, at least regarding read speed; sustained writes for a QLC drive are unlikely to surpass the Gen4 barrier.
The LC9 also features dual-port capability, making it viable for high-availability enterprise storage solutions. KIOXIA quotes an endurance rating of 0.3 DWPD, which should be sufficient for the read-heavy workloads these drives will likely be tasked with. At the time of this release, KIOXIA doesn’t have a published datasheet. We’ll be interested in seeing the performance profile when these drives get closer to general availability.
The Role of the KIOXIA LC9 in AI Workloads
Storage requirements in AI, particularly for training and inference, continue to push the boundaries of available technology. Large-scale models require extensive datasets, and high-density SSDs like the KIOXIA LC9 are becoming critical in managing these workloads efficiently. The ability to store vast training data on a single drive reduces complexity and improves access speeds, particularly for AI systems leveraging retrieval-augmented generation (RAG) architectures.
KIOXIA also highlights its AiSAQ technology, which optimizes vector database performance by storing elements on SSDs instead of DRAM. KIOXIA says this approach improves efficiency and lowers the cost of high-speed AI inference systems.
KIOXIA LC9 vs. Solidigm P5336
Solidigm was the first to introduce a 122TB SSD with its P5336 in November last year. With KIOXIA now offering a comparable product in the LC9, there are a few key areas to examine:
- Interface & Performance: The KIOXIA LC9 supports PCIe Gen5, whereas the P5336 is a PCIe Gen4 drive. Will the updated interface give KIOXIA a performance advantage, particularly in bursty read scenarios?
- Dual Ported: Enterprise storage arrays have historically preferred a dual-ported drive for high-availability reasons. Will this give the LC9 an advantage, or have enough storage array vendors adapted to single-port SSDs for their arrays?
- Endurance & Use Cases: Both drives feature 0.3 DWPD, which is best suited for read-heavy workloads such as AI dataset storage, cloud archival, and large-scale object storage.
- Form Factors: Both drives have a U.2 form factor, but the P5336 also has an E1.L form factor, giving the Solidigm drive a little more flexibility for system designers.
- Availability: None of the above matters if customers can’t buy the drives—the real winner in this 122.88TB SSD showdown is going to be whoever can actually produce them and get them into customers’ hands.
For a deeper look at the Solidigm P5336, check out our extensive review here: Solidigm D5-P5336 Review.
Final Thoughts
The KIOXIA LC9 Series SSD marks a significant milestone in enterprise storage, leveraging QLC NAND and PCIe Gen5 for improved performance and density. As AI, cloud storage, and hyperscale data centers demand ever-larger and more efficient storage solutions, the battle for dominance in the 100TB+ SSD space is just getting started.
Engage with StorageReview
Newsletter | YouTube | Podcast iTunes/Spotify | Instagram | Twitter | TikTok | RSS Feed
