Seagate joins the HDD price hike party, blames AI for spike in demand Seagate has joined Western Digital in increasing the prices of hard drives, with rising demand due to the huge data requirements of AI taking the blame. AI is also behind a rapid growth in orders for Enterprise solid state drives (SSDs). One of the big three makers of traditional rotating hard disk drives (HDDs), Seagate …
"According to TrendForce, the increasing use of QLC SSDs for AI applications is because inference servers primarily perform read operations, and SSDs have a speed advantage here. Enterprise SSDs also consume less power and are generally more compact than hard drives."
Bzzt. Wrong. On a per TB basis, Spinning rust actually uses less power until a tipping point around 65 TB flash media.
I know - I was surprised too! To verify, pick any server vendor that can publish power numbers for a specific config (Dell, Supermicro, and AIC are the ones I have modeled for a large (>10PB) storage platform), and compare an all-flash/NVME config with similar capacity in HDD. For bonus points, use the same number of drives. HDD configs are still lower power draw than SSD, uniformly, and it's actually a significant difference. Tipping point is the 60 TB media coming on line but that has some * interesting * internal architectures that mean it really, really should be thoroughly qualified if the file system addressing it was built for previous generations of flash media. And, sorry to say, QLC ain't all that much faster than a spinning drive if the read patterns tend toward the sequential...
"To verify, pick any server vendor that can publish power numbers"
More to the point, verify using your own draw measurements.
Vendors specify worst case scenarios - for liability reasons(*) - and whilst it's unusual to exceed them in real world ops, it's not so unusual to find long term averages are way lower than the specs
(*) Overloading cooling plant is a really bad idea
The rising demand comes from AI training requiring huge volumes of data: OpenAI's GPT-3 model is said to have been trained using 45TB of data, which may have been surpassed for newer models. And while flash-based SSDs boast high-speed and low-latency, storing everything in flash would still be costly. Seagate launched a 30TB hard drive line last year.
If you can afford the GPUs for building models then you can afford a NAS/whatever with 100TB or so of SSD
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
