IMD SolutionComputing LibraryShop main store

Selection guide

Qualifying an industrial PC after an MLC to TLC SSD change

Industrial and panel PCs Published 6 min read

Short answer

Accept the TLC configuration when the delivered SSD's specified TBW divided by your measured daily host writes exceeds the service life the machine needs. MLC stores 2 bits per cell and TLC stores 3, moving manufacturer per-cell cycle ratings from 4000-plus down to 900-1200, but capacity and spare area set the drive's TBW. Decide on the delivered drive's numbers, not the NAND label.

On this page
  1. What the cell change alters
  2. The three numbers the decision needs
  3. The calculation, worked through
  4. Decision table for the transition
  5. Acceptance on the delivered unit
  6. What to send with the order
  7. Questions asked

A NAND label change neither qualifies nor disqualifies an industrial PC. MLC stores 2 bits per cell and TLC stores 3, and manufacturer per-cell cycle ratings drop from 4000-plus to 900-1200 across that change — but service life follows from the delivered SSD's specified TBW divided by the writes your application issues per day. The transition now moving through industrial PC supply — one vendor's July notice stopped accepting new orders for 512 GB configurations during the switch and projected the TLC version back around September 2026 — is an engineering change you accept or reject with arithmetic. Measure the writes, read the TBW at the delivered capacity, and let the quotient decide.

What the cell change alters #

MLC in common usage means 2 bits per cell, TLC means 3 and QLC means 4. More bits per cell packs more voltage levels into the same cell, narrows the read margins and wears the cell out in fewer program-erase cycles: manufacturer ratings cluster at 4000-plus cycles for MLC and 900 to 1200 for TLC.

That per-cell figure is not the drive's endurance. Terabytes written (TBW) is the vendor's figure for how much data the drive absorbs before it no longer meets specification, and every capacity carries its own rating, so endurance scales with capacity and spare area rather than with the label. As an idealized ceiling that ignores write amplification: 512 GB of TLC at 1,000 cycles is 512 GB × 1,000 = 512,000 GB ≈ 512 TB of NAND writes, while 256 GB of MLC at 4,000 cycles is 256 GB × 4,000 = 1,024,000 GB ≈ 1,024 TB. The larger TLC drive holds more total writes than the smaller MLC drive despite the lower cycle rating.

The rating itself has a defined basis: JEDEC JESD218 sets how an SSD's endurance rating is measured and the conditions it assumes, and JESD219 defines the write workload behind a client rating. In plain words, a TBW figure is proven for drives used about the way the test workload uses them; a plant-floor box writing far more per day is judged on its own measured writes, under the operating conditions — temperature included — that the drive's specification states. A real-world endurance test that wrote 31 PB against a drive's rated TBW shows why the rating, not the label, is the reference worth arguing from.

The three numbers the decision needs #

  • Daily host writes, in GB/day: read the drive's total host-writes counter on a machine already running the application, at the start and end of a representative week — one that includes batch reports or archive cycles — and divide the difference by the days elapsed.
  • Specified TBW at the delivered capacity: quoted from the SSD's documentation for the exact drive being supplied, since the rating differs by capacity.
  • Required service life, in days: taken from the machine's obsolescence plan. Five years is 365 × 5 = 1,825 days.

One integrity check belongs beside these numbers: power-loss behavior is a firmware property documented for the specific drive, not a property of the NAND label, and on a logging computer that rides through outages you read it in the delivered specification before acceptance.

The calculation, worked through #

Take example inputs and carry the units. Suppose the delivered SSD is specified at 250 TBW — an illustrative rating, not a product claim — and the service target is 5 years, or 1,825 days. 250 TBW is 250,000 GB.

HMI with light logging, 40 GB/day measured:

250,000 GB ÷ 40 GB/day = 6,250 days, and 6,250 ÷ 365 ≈ 17.1 years. Against a 5-year target that is a pass with margin: accept the TLC configuration for this duty.

Continuous history and vibration logging, 600 GB/day measured:

250,000 GB ÷ 600 GB/day ≈ 417 days ≈ 1.1 years. That fails the target. The drive this duty needs is 600 GB/day × 1,825 days = 1,095,000 GB ≈ 1,095 TB of specified endurance — specify 1,100 TBW or better, or cut the write rate, because every halving of daily writes doubles the service years. High-rate logging boxes carry a parallel sizing question in sizing edge storage for high-rate vibration data, where retention and rate are set together.

A TBW pass does not prove sustained write speed. Check the drive's sustained write specification against the peak logging rate before acceptance: a drive can hold a comfortable endurance quotient and still stall when buffered history flushes.

Decision table for the transition #

Option Gate it must pass Decision
Product-qualified TLC configuration TBW ÷ daily host writes covers the required service life, and the application baseline passes on the delivered unit Accept it; this is the vendor-supported path through the change
Existing MLC configuration Still orderable, and configuration control needs the established baseline Keep it as the compatibility baseline; obtain written order acceptance for each additional unit
Restricted 512 GB configuration Vendor accepts the order and confirms a delivery date Treat the projected September 2026 resumption as planning input, never as a commitment
Independent SSD substitution by your shop You run the full endurance, performance and application qualification yourself A separate engineering change you own, not equivalent to the vendor-qualified transition

If no option closes the numbers on your timeline, treat the machine as a re-selection: for cabinet duty, specify a fanless industrial mini PC with the drive's endurance documented at the capacity you need, and apply the same three-number check to it.

Acceptance on the delivered unit #

Change one thing at a time. Restore the operating-system image, application project, communication settings and recipes from the MLC baseline, confirm the delivered SSD's identity and capacity against the order, and only then test.

  • Compare a live controller value with the screen. A stale value points at the tag binding, the driver or the controller connection — it is not a storage symptom.
  • Write a record to the logging destination and read it back, with free space checked at the file system.
  • Run one full shift of the real logging load, then read the host-writes counter again and confirm the GB/day still matches the number the calculation used.
  • If the unit runs slow under load, prove an IPC slowdown is hardware, not software before touching the storage configuration.

What to send with the order #

The supply notice asks for model name, quantity, required delivery date, and certification or approval requirements before an order is placed. Add two engineering items: the accepted configuration revision, and the specified TBW at the delivered capacity. With both in the order record, the delivered unit is checked as a controlled replacement instead of a surprise, and a failed check has a paper trail to act on.

Questions asked

How many bits per cell do MLC and TLC actually store?

In common usage MLC stores 2 bits per cell, TLC stores 3 and QLC stores 4. More bits per cell narrows the voltage margins, so manufacturer cycle ratings run 4000-plus for MLC against 900 to 1200 for TLC. Because TBW scales with capacity and spare area, a larger TLC drive can outlast a smaller MLC drive, which is why the delivered rating, not the label, settles the question.

Does the SSD change affect PLC tags or HMI drivers?

No. Swapping storage changes where the operating system and the application persist data; it does not touch tag bindings, communication drivers or controller addressing. A stale process value on screen points at the tag or driver path, while missing history or failed file writes point at the storage path. Compare a live controller value against the display before suspecting the drive.

The vendor quotes endurance in DWPD, not TBW. How do I compare?

Convert first: DWPD multiplied by capacity and by the rating period gives TBW. Example: 1 DWPD on a 512 GB drive over 5 years is 1 × 512 GB × 365 × 5 = 934,400 GB, about 934 TB. A 250 TBW drive at the same capacity and period is therefore about 0.27 DWPD, and both figures drop into the same service-life division.

The calculation fails on my logging machine. What closes the gap?

The quotient is linear, so any halving of daily writes doubles the service years. In order of preference: specify a drive whose rating meets service life times daily writes, lengthen the sampling or aggregation interval so fewer bytes reach the disk, or buffer at the edge and commit in batches. Do not accept the configuration on the strength of the NAND label.

Is the projected supply resumption date a delivery commitment?

No. The notice projects the TLC version of the restricted 512 GB configuration to return around September 2026, and that figure belongs in planning scenarios only. Order acceptance and a confirmed delivery date are separate facts to obtain in writing before committing a machine schedule; until then, run the MLC baseline and the TLC qualification in parallel.

Keep reading

  • Checklist

    Running a compact industrial PC as an always-on server

    Yes, a compact fanless industrial PC can host small IoT, VPN and SCADA services for years, but its rugged build guarantees survival, not capacity. Qualify the exact model against measured peak load: CPU below 70%, memory below 75%, SSD write endurance with margin, thermal margin at the site, and a restore test that proves recovery.

    Industrial and panel PCs

  • Troubleshooting

    How to prove an IPC slowdown is hardware not software

    A slow HMI on a booting IPC is not fault evidence. Capture per-process CPU, memory, disk and network under the production workload first: saturation with a clean event log is a sizing or software problem, and only an error that repeats under a separate one-component stress load — twice, with a matching Windows event — convicts the hardware.

    Industrial and panel PCs

  • How-to

    Using a KVM extender to run an HMI from a panel IPC

    Yes — mount the IPC in the control panel and put a purpose-built KVM extender pair between it and the operator station, so one Cat6 run carries video and USB to where grit and heat destroy exposed computers. It works only if the model supports your video output, the display's resolution and refresh rate, and every USB device, and the link stays point-to-point. Size the run against the model's stated length, not against the Cat6 label.

    Industrial and panel PCs