How to Read Flash Die Markings and Lot Codes Before Approving a Production Run

TL;DR

Before you approve a flash production run, ask for three things and check them against each other: the die marking on the silicon, the lot/date code on the package or reel, and the identification data the finished product reports to a host (CID/CSD on SD, descriptors on USB, CID/CSD on eMMC). If those three do not line up with the supplier’s declaration, you do not have traceability — you have a shipment.

The core question buyers ask is: “Can I tell what flash is actually inside this card or drive before I accept the run?” The honest answer is partially, and only if you set the requirement before production. Die markings identify the maker, die family and lot; they do not by themselves certify grade or endurance. Lot codes let you group units by process history so that a failure in one group does not silently contaminate your whole acceptance decision. And the host-readable identifiers (CID, CSD, USB descriptors) let you verify after the fact that the finished product matches the declared bill of materials.

This article sets out a practical reading method, the fields that matter, and the points where buyers most often mistake a marking for a guarantee.

Why lot traceability matters more than a single “good sample”

A sample that passes your bench test tells you about the sample. It does not tell you about the run. Flash varies by die, by lot, by test bin and by the controller/firmware pairing. Two cards with identical external print can contain different die from different lots, and behave differently at temperature, at end-of-life, or after an unexpected power loss.

Traceability is the mechanism that lets you answer, six months later: which units in the field came from the same material as the failures I am seeing? Without lot-level records, a field failure is an anecdote. With them, it is a bounded problem.

This is also why traceability and grade are separate conversations. See original, white and black die grading for how the commercial grading system works, and bad block handling and what the controller does for what happens when the die is less than perfect.

The three layers of identification

Think of traceability as three layers that must agree.

Layer	Where it lives	What it identifies	Who controls it
Die marking	Laser/ink mark on the flash package or die	Silicon maker, die family, lot	Flash manufacturer
Package/reel lot & date code	Label on reel, tray, or card package	Batch and production date	Flash maker and/or assembler
Host-readable ID	CID/CSD (SD, eMMC), USB descriptors	Finished product identity and declared parameters	Card/drive assembler (controller + firmware)

A supplier who can only give you one of these layers is giving you a partial answer. A supplier who can give you all three, and whose numbers reconcile, is giving you a traceable run.

Layer 1 — Die markings

Die markings are typically a compact string on the flash package: a manufacturer logo or code, a part family code, a lot code and a date code. The exact format is proprietary to each flash maker and changes between generations. There is no public universal decoder.

What you can do without a decoder:

1. Record the marking verbatim from a sample of units, photograph it at magnification, and require the supplier to declare the same marking in writing.
2. Check consistency across the shipment. If the declared lot is one marking but you find three distinct markings in the sample, the shipment is mixed. That is not automatically wrong — but it must be declared and explained.
3. Check the date code against the promised material age. A “new” run built from material with an old date code is a different commercial proposition.
4. Do not attempt to infer grade from the marking. Grade is a test-and-sort outcome, not a printed field.

If you need a formal decode, the only reliable route is to ask the flash manufacturer or an authorised distributor to confirm the marking format for that part family. Do not rely on forum decoders for a purchasing decision.

Layer 2 — Lot codes and date codes

Lot codes group units by manufacturing batch. Date codes encode production timing. Practical rules:

• One lot is easier to qualify than many. A run built from a single declared lot has a single process history. A run assembled from leftover partial reels across many lots is harder to qualify and harder to investigate later.
• Ask for the lot list, not just “the lot”. A production run of any size may legitimately draw on several lots. The supplier should be able to list them.
• Date code is not shelf life. Flash does not expire on a fixed date, but long-stored material can have different solderability and different program/erase history. See how long flash memory holds data without power for the retention side of that question.
• Reel labels are part of the record. Photograph them at goods-in, before the reels are consumed.

Layer 3 — Host-readable identifiers

This is the layer most buyers can actually verify themselves, with a laptop and free tools.

microSD / SD cards. The card exposes a CID register containing manufacturer ID, OEM/application ID, product name, product serial number and manufacturing date, and a CSD register containing capacity and timing parameters. The SD Association publishes the physical and command-layer specifications that define these fields (see Sources). You can read them on Linux with the MMC subsystem, or on Windows via vendor tools and the documented storage driver interfaces.

USB flash drives. USB devices report descriptors — vendor ID, product ID, serial number, and for mass-storage devices the SCSI inquiry data. Microsoft documents how Windows exposes these (see Sources). The important point for traceability is that the serial number and inquiry strings are programmed by the assembler, so they are only as trustworthy as the assembler’s process discipline.

eMMC / managed NAND. JEDEC JESD84 defines the CID and CSD structure for eMMC, including manufacturer and OEM fields and a production date. If you are sourcing modules rather than cards, this is the layer to check — see microSD card vs embedded eMMC for when each applies.

Important caveat: host-readable identifiers are assertions by the assembler, not proof of the die inside. A CID that says one thing while the die marking says another is exactly the discrepancy traceability is designed to catch. Treat the two layers as cross-checks, not substitutes.

A pre-approval checklist

Use this before releasing a run. It is deliberately short enough to actually be done.

Before the order is placed

• Written declaration of the flash part family and grade the supplier intends to use.
• Written statement of how many lots may be drawn on, and whether substitution is permitted.
• Agreement on what traceability data will be delivered with the shipment (lot list, reel label photos, marking photos, CID/descriptor dump).
• Agreement on what happens if the delivered data does not match the declaration.

At goods-in

• Photograph every reel label and every carton label before opening.
• Sample units from each declared lot separately — not from the shipment as a whole.
• Record die markings verbatim at magnification.
• Read CID/CSD or USB descriptors from the same sampled units.
• Reconcile: declared lot ↔ reel label ↔ die marking ↔ host ID. Any mismatch is a hold, not a note.
• Run capacity and basic performance verification. See fake USB drive capacity and how to check and the TF card capacity estimator.

Before you scale

• Keep the goods-in record for the life of the product, keyed to the lot codes.
• If field failures appear, pull the record and identify which lots are implicated before making any claim about the supplier.

The sample acceptance checklist tool covers the bench-test side of this; the checklist above covers the paperwork and identity side.

Worked example: reconciling a shipment

(Illustrative only — not WINTEL data)

Suppose a buyer orders 20,000 microSD cards and the supplier declares “single lot, grade A, date code 2612.” At goods-in the buyer samples 30 units across three cartons and finds:

Check	Declared	Observed	Verdict
Lot codes on reel labels	1 lot	2 distinct lot codes	Mismatch — query
Die marking	One marking	Two markings, one dominant	Consistent with two lots
CID manufacturing date	2612	2612 on 27 units, 2540 on 3 units	Mismatch on 3 units
Capacity	32 GB class	All report 32 GB	Pass

The correct response is not to reject the shipment outright. It is to hold it, ask the supplier to explain the second lot and the older CID dates, and decide whether the mixed material is acceptable for the application. For a low-value promotional drive the answer might be yes with a price adjustment. For a dashcam or industrial card it is almost certainly no — see how to choose a TF card for a dashcam for why mixed material is a poor fit there.

The point of the exercise is that the buyer found out at goods-in, not eighteen months later from a customer complaint.

Where traceability commonly breaks down

1. Substitution after approval. The supplier qualifies one part, then builds with another when supply tightens. This is the single most common traceability failure, and it is why the written declaration matters more than the sample. The controller shortage and semiconductor cycle article covers how this pressure develops.

2. Assembler-programmed IDs treated as proof. A CID or USB serial number is only as good as the assembler’s discipline. If the same serial range is reused across lots, the identifier stops being a traceability tool.

3. Reel-level records lost at the line. The reel label is photographed at goods-in and then thrown away with the reel. If the finished-goods record does not carry the lot forward, the trace ends at the factory door.

4. Grade conflated with marking. Buyers sometimes assume a particular marking means “original grade.” It does not. Grade is a commercial classification backed by test data, not a printed field.

5. Traceability promised verbally. If it is not in the order documentation, it will not survive a change of production planner.

How WINTEL handles this in an RFQ

WINTEL is a manufacturer of custom TF/microSD cards, USB flash drives, UDP modules and semi-finished boards, working with OEM/ODM buyers. Traceability requirements — what lot data is declared, what marking records are delivered, what host-ID dump accompanies a shipment — are agreed per order as part of the RFQ, not assumed.

What this article does not claim: it does not state WINTEL’s speed classes, endurance ratings, temperature ranges, MOQ, lead times, test results or certifications beyond those published on the company profile, and it does not describe a standard traceability report format. If lot-level reporting is a requirement for your programme, raise it explicitly when you request a quote so it can be confirmed in writing.

To start that conversation, use the RFQ form and state: the product, the application, whether single-lot material is required, and what traceability data you need delivered with each shipment.

Related reading on the manufacturing side: inside a flash memory factory, flash package formats explained, and flash brand and controller pairing.

FAQ

Can a die marking tell me whether the flash is “original” or “downgrade” grade? Not by itself. Die markings identify the manufacturer, the die/part family and the production lot. Grade (original, white, black, downgrade) is a commercial classification of tested-and-sorted die, and the marking does not reliably state it. Grade has to be verified through the supplier’s declaration, incoming test data, and consistency across lots — not from a laser mark alone.

What is the difference between a lot code and a date code? A date code encodes when the die or package was produced (often week/year). A lot code identifies a specific manufacturing batch, which may span several days. Both matter: the date code tells you the age of the material, the lot code tells you which units share a common process history and therefore should be treated as one quality group.

Is the CID register on a microSD card the same as the die marking? No. The CID is programmed by the card assembler/controller and contains manufacturer ID, OEM/application ID, product name, serial number and manufacturing date. It is a card-level identifier, not a die-level one. For traceability you need both: the CID ties the finished card to a production record, while the die marking ties the silicon to the flash maker.

How many lots should I sample to judge lot-to-lot consistency? There is no universal number. A practical approach is to require the supplier to declare every lot used in the run, then sample from each declared lot separately rather than from the shipment as a whole. If a run contains many small lots, that itself is a signal to ask why.

Does WINTEL publish die markings or lot data for its cards? WINTEL can discuss traceability requirements as part of an RFQ, but this article does not claim specific marking, lot-reporting or data-logging services beyond what is agreed in writing for a given order. Confirm the exact deliverables with the sales contact before approving a run.

Sources

• SD Association — SD Specifications (Physical Layer, Part 1; command layer). Defines CID and CSD register structures used for card identification. https://www.sdcard.org/downloads/pls/ — accessed 2026-10-20.
• USB Implementers Forum — USB documents. Defines device descriptors and mass-storage class behaviour referenced when reading USB drive identification data. https://www.usb.org/documents — accessed 2026-10-20.
• JEDEC — Standards and documents (including JESD84 for eMMC). Defines eMMC device identification and CID/CSD structure. https://www.jedec.org/standards-documents — accessed 2026-10-20.
• Microsoft Learn — Windows hardware driver documentation. Describes how Windows exposes USB and storage device identification to host software. https://learn.microsoft.com/en-us/windows-hardware/drivers/ — accessed 2026-10-20.
• Linux kernel documentation. Documents the MMC/SD and USB subsystems and how CID/CSD registers and descriptors are read. https://docs.kernel.org/ — accessed 2026-10-20.