With the rise of AI digital workloads and with nearly 150 countries adopting data sovereignty requirements, requiring data to stay in country, the demand for data storage to support these efforts is growing. SSDs will see growth in sales for higher performance storage for AI but hard disk drives will see growth as well as a less expensive option for storing less frequently access data. In addition, new interconnect technologies will be needed to efficiently support changing AI workflows.
Let’s look at an announcement by Seagate on its Mozaic HDDs for data center and enterprise applications and by Panmensia on its new composable AI infrastructure.
Seagate says that it has shipped over 1 million Mozaic HAMR HDDs. The company announced that it is launching 30TB Exos M and IronWolf Pro HDDs built on the company’s Mozaic 3+ platform. The Exos M and IronWolf Pro 30TB drives are prices at $599.99. The Exos M is meant for high-capacity energy-efficient storage that the company says empower organizations to scale storage, optimize data placement and support real-time edge analytics. The IronWorf Pro HDDs are meant for on-premise NAS systems.
These are supporting intensive workloads such as video analytics, image recognition, retrieval-augmented generation, RAG and inferencing at the edge. NAS companies Qnap and Ugreen are offering the IronWolf Pro in their NAS offerings. The image below shows an IronWolf Pro drive next to a QNAP NAS.
Seagate says that these drives are built with more renewable energy and recycled materials than any previous Seagate product.
Panmnesia is a fabless startup developing interconnect technologies. The company recently announced a line of products that it calls its Link Solution for next generation AI infrastructure. The image below gives a view of these products.
This full-stack product lineup includes hardware, silicon IP, software and customized design solutions. These products are meant to enable composable architectures for easy addition, removal and reconfiguration of devices such as GPUs, AI accelerators and memory modules to meet the needs of AI workflows. They will accelerated AI workloads by minimizing inter-device communication overhead.
Panmnesia is participating in various industry consortia such as the CXL consortium, UALink consortium, the Open Compute Project and PCI-SIG. The hardware products include switch chips in an SoC package and retimer chips.
The switch chip expands infrastructure by enabling large-scale device connectivity and managing data transmission among connected components. The retimer chip addresses physical distance challenges by restoring signal strength lost during long-distance communication.
The company also offers silicon IPs (semiconductor IPs) targeting a wde range of system devices that make up AI infrastructure — including CPUs, GPUs, NPUs, memory, and storage. Device manufacturers can use these IPs as design building blocks to enable their system devices to support specific interconnect protocols.
The company’s custom design solutions include PanEP, custom silicon services for AI accelerator or memory chips tailored to client needs. It also includes PanFabric, network architecture design services to minimize communication overhead for AI applications.
Seagate announces 30TB Mozaic HDDs infrastructure for AI. Panmnesia introduces hardware, silicon IP, software and customized design solutions for AI infrastructure.
Read the full article here
