Roger Corell
Sr. Director Leadership Marketing, Solidigm
Dave Sierra
Data Centre Solutions Marketing, Solidigm
Solid-state drives help AI perform to its full potential — it’s time for data centres to modernise.
Artificial intelligence (AI) is growing and scaling up at an exhilarating rate. Without improvements to crucial digital infrastructure, that pace is at risk.
How SSDs revolutionise AI data storage
Today, approximately 90% of data centres use hard disk drives (HDDs) to store the reams of information that make all types of AI work, from machine learning to large-language models. However, HDDs, which were invented in the 1950s, are not well suited to incredibly data-intensive AI.
Enter solid-state drives (SSDs), which are smaller, higher-density, enable lower cost, last longer and require much less power to run. Quad-level cell (QLC) technology, the most advanced SSD as of today, allows for four bits of data to be stored in each memory cell.
“Whether AI is running image recognition in a doctor’s office or doing quality control checks on a smart factory floor, it needs a massive amount of data, and it needs to be able to access that data at speed,” says Roger Corell, senior director of leadership marketing at California-headquartered Solidigm, which has the second-largest share of the global enterprise SSD market. “Compared to HDDs, SSDs can enable a lower total cost of ownership, are orders of magnitude faster and, with their massive capacities can deliver huge infrastructure efficiency improvements,” he adds.
Inefficient HDDs hinder AI
Corell argues that the advent of AI requires a complete hardware overhaul. “The number one challenge to scaling AI data centres is power consumption. Rightfully so, everyone talks about GPUs (graphics processing units) — but storage is a big factor as well because you need more HDDs to meet performance requirements. They are always on and consume a massive amount of power,” he says.
A single rack of SSDs can store the same amount of data as around three racks of HDDs.
HDDs are slow, so you have to combine many of them to get to the performance of a single SSD. “They can’t store nearly as much data as an SSD. Those are big problems, particularly with all the news about power constraints holding back AI data centre buildouts. For some AI usages, HDDs can account for up to 35% of that energy consumption,” explains Corell.
SSDs improve infrastructure efficiency
A single rack of SSDs can store the same amount of data as around three racks of HDDs. “In that situation, you’ve reduced your data centre footprint, reduced your power consumption by about 80%, reduced your cooling cost, and got a further saving when it comes to end-of-life because you’ve got fewer products to dispose of,” adds Corell.
While AI presents the most urgent storage issue for data centres, it’s not the only technology putting pressure on them. “There are other very data-intensive workloads such as big data analytics and high-performance computing that need the density and performance of high-cap SSDs,” says Corell.
SSD long-term efficiencies outweigh cost
According to the International Energy Agency, data centres accounted for about 1% of global electricity demand in 2020. This is only expected to grow as AI and cloud computing advance dramatically. This means that as tech companies reduce their carbon emissions in other areas, for many, their overall emissions are going up due to the data centre boom.
SSDs come with a higher upfront cost than HDDs, but due to their operational and performance efficiencies and longer lifecycles, they’re more cost-effective over time. “Once you look at total cost analysis, there’s a crossover point — particularly as our capacities increase — where SSDs are more economical in the long term,” says Dave Sierra, data centre solutions marketing at Solidigm.
“That’s because you need fewer drives, and you get more efficient power savings too. We’re focused on ensuring our customers evaluate the full solution, not just the initial upfront cost.”
High-density SSDs reduce energy consumption
As AI continues to evolve, its energy consumption cannot be ignored. High-density storage solutions, such as those pioneered by Solidigm, offer a promising path toward more sustainable data centre operations. By deploying hyper-dense QLC storage, data centres can significantly reduce their physical and energy footprints.
