Skip to content
Theme
OnDevice Tools

GadgetsLaptops & PCs

Enterprise SSD testing gets a new Tom's Hardware lab

Tom's Hardware has detailed a workstation and automated benchmark suite for reviewing data centre SSDs. The setup aims to measure drives under sustained, real-world load rather than only peak specifications.

Illustration of an enterprise SSD test bench with server drives and cooling fans

Key takeaways

  • Tom's Hardware has built a dedicated bench for enterprise and data centre SSD reviews.
  • The system uses a Xeon 658X processor, PCIe 5.0 adapters and a custom Debian 13 image.
  • The benchmark suite tests drives under load, during write bursts and with other tasks active.
  • A full run can take much of a day on a 7.68TB PCIe 5.0 drive.
In this article
  1. New enterprise SSD test bench
  2. Workstation hardware
  3. Drive adapters and cooling
  4. Benchmark method
  5. Why it matters

New enterprise SSD test bench

Tom's Hardware has outlined a new enterprise SSD testing system for reviews of server and data centre storage. The publisher says the project follows a gap of several years in its in-depth coverage of these drives.

According to Tom's Hardware, standard PC SSDs and older hard disks are not suited to heavy AI tasks. They do not offer enough performance or endurance. That has increased demand for enterprise drives, which are built for higher capacity, stronger reliability and continuous service.

The company says formats such as U.2, U.3, E3.S and E1.S were once rare outside large systems. They are now appearing in high-end workstations and home labs, making independent reviews more useful for buyers.

Workstation hardware

Illustration: Corsair logo, privacy, storage, star and the labels Debian 13, Enterprise SSD, Xeon 658X

The test system is built around an Intel Xeon 658X processor. Tom's Hardware says the chip has a 24-core, 48-thread design and a 4.9 GHz boost clock. It can spread storage workloads across multiple test jobs.

Cooling is handled by a Noctua NH-U14S DX-4677 air cooler. The cooler supports LGA4677 and LGA4710 sockets and uses two NF-A15 140mm fans.

The motherboard is an Asus Pro WS W890E-Sage SE. It includes an MCIO x8 connection split into two x4 links for U.2 or U.3 drives. It also has seven PCIe 5.0 slots. Tom's Hardware says this lets the system use the Xeon 658X's 128 PCIe 5.0 lanes without hitting bandwidth limits.

Memory comes from a G.Skill G5 DDR5-6400 C32 64GB kit, made of four 16GB modules. It runs at the stock DDR5-5600 speed because Tom's Hardware says memory speed has no bearing on these SSD results. Corsair supplied an MP700 Pro XT 2TB drive for the bench's main storage. The PCIe 5.0 drive has a sequential read rating of 14,900 MB/s and a sequential write rating of 14,500 MB/s. Corsair also provided an HX1500i power supply rated at 1,500W. It is rated 80 Plus Platinum, with efficiency no lower than 89% under full load.

The project also reflects a difficult component market, with Corsair and G.Skill supplying parts despite a memory and storage shortage.

Drive adapters and cooling

Enterprise drives often use connectors that do not fit standard consumer slots. Icy Dock provided ExpressSlot Elite MB221E5P-B and MB303E5P-B adapters for one and two E1.S drives. Tom's Hardware says these adapters can be installed without tools and include 50mm fans.

For projects involving multiple E1.S drives, the list includes a 5.25-inch ToughArmor MB218E5P-B bay that supports up to eight drives. Tom's Hardware says it is intended for future RAID projects, which combine multiple drives. E3.S testing uses a JMT ADT-R27C adapter paired with a Noctua NF-F12 industrialPPC-24V-3000 IP67 PWM fan. The fan can run at 12V while still moving enough air to cool the drives.

The hardware sits on a TGDGamer E-ATX open-air test bench. Tom's Hardware says the open design makes installation easier and reduces heat build-up variables. It also supports motherboard sizes up to SSI-EEB.

Benchmark method

The new suite is designed to test drives as they behave in service, not just on a specification sheet. Tom's Hardware says the tests check how a drive performs once it has reached a stable operating point. They also measure behaviour while the drive is busy, during short intense write periods and when other tasks are active.

It also notes that a drive in real use is often nearly full. Background garbage collection is also at work. This is housekeeping that frees and organises storage.

Each drive goes through an identical automated sequence on the same hardware. Results are stored with the settings used, allowing the team to trace how each result was produced. On a 7.68TB PCIe 5.0 model, a complete pass can last for much of a day. It writes over the entire capacity repeatedly.

The benchmark can run unattended and can resume after a crash from the last completed stage. A shorter express mode is used for quick checks. Variant runs repeat selected tests or change one setting to investigate unusual results before contacting a vendor.

The bench runs a custom Debian 13 image built and maintained by Tom's Hardware. The image is installed on the bench's operating-system drive. The suite is copied into the image at build time. The kernel is pinned so each drive is tested with the same software stack.

Why it matters

The new setup gives Tom's Hardware a way to compare enterprise SSDs under conditions closer to real server use. That can help organisations decide whether a drive is suitable for AI, data centre or high-end workstation tasks.

The publisher says it still needs support for an RTX Pro graphics card. Such a card would help with real-world AI workload testing, but the current platform already covers the main performance measures.

For readers, the project shows how much hardware and automation is needed to review modern storage. It also highlights why enterprise SSD reviews require more than a quick consumer-style benchmark.

Sources

  1. How we test enterprise SSDsTom's Hardware

Tags

Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.