- Huawei shifts focus from HBM toward SSD capacity to control AI loads
- Oceandic LC 560 can boast of 245TB, the largest SSD ever made
- Huawei promotes diskbooster -software as a memory pooling solution across HBM and SSD
Huawei has revealed (originally in Chinese) a set of AI SSDs designed to facilitate the addiction of high bandwidth memory, an area where Chinese companies are facing the restrictions of supply.
At its data storage AI SSD launch event, the company presented Oceanic ex, Oceanic SP and Oceanic LC 560 drive.
The company described the launches as a response to the “memory wall” and “capacity wall” problems that are currently limiting workloads to artificial intelligence.
Huawei Oceandic Ex, SP and LC 560 Drive
“The increasingly difficult ‘memory wall’ and ‘capacity wall’ have become key bottlenecks for AI training efficiency and user experience. This creates challenges for the performance and cost of the IT infrastructure, which affects the positive AI conjunction,” said Zhou Yuefeng, vice president and leader of Huawei’s data warehouse product.
Oceandic EX 560 is described as an “extreme performance drive” with writing speeds reaching 1,500K IOPS, latency below 7µs and endurance of 60 DWPD.
Huawei says it can increase the number of fine-tuning model parameters on a single machine that is six times, potentially making it useful for LLM fine adjustment.
The SP 560 takes a more cost -focused approach with 600K IOPS and lower shelf life of 1 DWPD.
It is targeted at inference scenarios with claims to reduce first-token latency by 75%while doubling flow.
The LC 560 represents the largest SSD in the series with a specified single-drive capacity of 245 TB and read bandwidth of 14.7 GB/s.
Huawei promotes this model, suitable for handling massive multimodal data sets in cluster training, although practical adoption in scale will depend on integration with existing systems.
In addition to hardware, Huawei also introduced Diskbooster, a driver said to expand the overall memory capacity twenty -fold by coordinating AI SSDs with both HBM and GDR.
The company also highlighted multi-stream technology aimed at reducing writing reinforcement, which could expand the driving force of life.
These software-driven optimizations can provide more flexibility to AI workloads, but actual gains depend on how good applications adopt and support them.
Huawei’s latest SSD is an attempt to mitigate the impact of the United States’ continued tightening of controls on advanced HBM chips that have left Chinese companies with limited access.
The company aims to reduce China’s dependence on imported HBM chips, emphasize domestic NAND -FLASH and move the value of SSD technology.
Whether these new drives can really equalize HBM limits in LLM training is still uncertain.
Nevertheless, Huawei’s solution does not replace HBM; It is only a partial alternative, but the company claims that high -capacity storage with a fixed state with specialized software can reduce the demand for large amounts of costly memory.
This claim still needs to be tested fully, but the procedure suggests a shift against “system supplement of individual points”, where different layers of storage balancing restrictions in HBM.
