Liquid-cooled servers find a new contender: Cold-Plate SSD
In the rapidly evolving world of Artificial Intelligence (AI), the demand for high-performance storage solutions has never been greater. This is where cold-plate SSDs come into play, offering a significant boost in AI performance within liquid-cooled environments.
Cold plates provide direct-to-chip liquid cooling for SSDs, a more efficient cooling method than air cooling. This efficiency is crucial in systems critical for AI, which are often burdened with intensive data processing and heavy computational loads.
One such cold-plate SSD is the E1.S D7-PS1010 SSD by Solidigm. With a sequential read bandwidth of 14,500 MB/s and a write bandwidth of 4,100 MB/s, this SSD is designed to handle the intense thermal demands of AI workloads. It is available in a 15-mm U.2 form factor and employs a x4 PCI Express (PCIe) Gen 5 interface with capacities exceeding 15 TB.
The implementation of cold plates in SSDs improves thermal management and reliability. In AI data centres where SSDs handle massive data loads continuously, overheating can throttle performance or cause failures. Cold plates help mitigate this risk by reducing SSD temperatures, ensuring sustained high-speed operation with better power efficiency.
Moreover, liquid-cooled servers using direct water cooling, which include cold plates, show measurable benefits. For instance, Lenovo's Neptune™ water cooling tech, integrated with cold plates for microelectronics including SSDs, achieves up to 10% higher system performance and reduces energy consumption by up to 40%.
The shift to liquid cooling (and cold plate solutions) in AI data centres is accelerating. Major cloud and AI providers are adopting these solutions to manage the heat from high-end AI chips and storage components, ultimately supporting scalability and thermal efficiency in AI infrastructure.
In summary, cold plates in SSDs facilitate more effective liquid cooling, allowing AI systems to operate at higher performance levels with better energy efficiency and reliability in liquid-cooled environments. This addresses the intense thermal demands of AI workloads, making cold-plate SSDs a valuable addition to the AI landscape.
Key points: - Cold plates enable higher sustained throughput for SSDs in AI workloads. - Improved reliability and reduced thermal throttling of storage in AI workloads. - Increased system efficiency (up to 10% performance boost, 40% energy reduction shown in liquid-cooled servers). - Critical for modern AI data centres adopting liquid cooling as standard infrastructure. - The E1.S D7-PS1010 SSD by Solidigm is a hot-swappable SSD designed for liquid-cooled environments. - The D7-PS1010 SSD maintains a read latency of 60 μs and a write latency of 7 to 8 μs. - The mean time between failures (MTBF) for the D7-PS1010 SSD is 2.5 million hours. - The maximum lifetime program/erase cycles for the 15.36-TB variant of the D7-PS1010 SSD over a five-year period are on the order of 28 petabytes written (PBW). - The SE models of the D7-PS1010 SSD can deliver 1.0 drive writes per day (DWPD). - The D7-PS1010 SSD uses 176-layer, TLC 3D NAND.
[1] Solidigm Press Release [2] Lenovo Press Release [3] TechTarget Article [4] Lenovo Whitepaper [5] Solidigm Whitepaper
Embedded systems employing cold-plate SSDs, such as Solidigm's E1.S D7-PS1010, can benefit from improved thermal management and reliability, consequently reducing the risk of overheating and performance degradation in data-and-cloud-computing applications like Artificial Intelligence. Thus, the integration of technology like cold plate SSDs and liquid cooling in AI infrastructure is a significant stride towards enhancing system efficiency and accommodating the intense thermal demands associated with AI workloads.
