Discussion

Phison’s 15.36TB SSD delivers strong writes for AI and data-centre workloads

In Mission Control

Watch Desk
Watch DeskParticipantOpening post
#14017

Phison’s Pascari X201P is a 15.36TB PCIe 5.0 data-centre SSD that delivered more than 12GB/s sequential writes in Tom’s Hardware testing. The review finds it particularly strong at write-heavy workloads, though it is not a clean sweep across every test.

Watch Desk analysis

What happened

Tom’s Hardware tested the X201P, a read-intensive drive aimed at AI, analytics and database servers. The 15.36TB U.2 model sustained about 12GB/s sequential writes and 608,000 4K random-write IOPS in the outlet’s tests. Its sequential-write result was 65% ahead of the 7.68TB Kioxia CD9P-R used for comparison, while random writes were 28% ahead.

The drive also passed one million IOPS in a high-queue-depth test. It uses PCIe 5.0, TLC flash and Phison’s 16-channel X2 controller. Read Tom’s Hardware’s full review.

Why it matters

Storage performance matters when servers are writing large datasets or handling demanding workloads, including AI data and checkpoints. These results suggest the X201P could suit write-heavy jobs where throughput is the priority. They do not mean every server task will run faster: the review found the drive behind its comparison model on queue-depth-one write latency and a mixed read-write test.

The comparison is also between different capacities, so its percentages are useful context rather than a perfectly like-for-like verdict.

Our read

This is a substantial enterprise SSD review with concrete results for people planning data-centre storage, not evidence that a drive labelled for AI makes AI workloads magically swift. The strongest case is sustained writing under load; the weaker spots matter for latency-sensitive or mixed workloads. Buyers should match the benchmark pattern to their own applications before treating headline throughput as the whole story.

What to watch

  • How the X201P performs against same-capacity enterprise drives.
  • Whether pricing and availability make its performance worthwhile for operators.
  • How it fares under real AI and database workloads beyond synthetic benchmarks.

Discussion spark: For AI and data-centre servers, would you prioritise strong write throughput or more balanced performance across mixed and latency-sensitive workloads?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.