Counterfeit, Remarked & Recycled ASIC Chip Detection — A Bench Authentication Guide

Counterfeit, remarked, and recycled mining ASICs are real: chips harvested from dead hashboards get cleaned, repainted, re-laser-marked, and resold as new. This is a detection-and-QA guide for buyers and repair benches — how to authenticate BM-class chips with a microscope, a solvent rub, date-code logic, dimensional checks, and post-install electrical binning. Authentication only, never replication.

If you have ever bought loose ASIC chips for a repair, or a “new” miner at a price that felt too good, you have been a target. The consensus from a decade of semiconductor counterfeit research is blunt: more than 80% of all counterfeit components in the supply chain are recycled or remarked parts pulled from scrapped boards and dressed up to look fresh (Tehranipoor et al., Counterfeit Integrated Circuits: A Rising Threat, Proceedings of the IEEE). Mining hardware is a soft target: dead hashboards exist by the tonne, the chips are small and visually similar across generations, and buyers are often time-pressed.

This is written for the hardware-fluent sovereign Bitcoiner and the small repair shop, and it is strictly an authentication and QA procedure — every technique below exists to catch counterfeits, never to produce them. Chip facts (packages, board architecture, signal behaviour) come from D-Central’s internal mining hardware reference.

The seven counterfeit categories, in a mining context

Counterfeit-component researchers group fakes into seven well-established types. Not all are common in Bitcoin mining, but knowing the taxonomy tells you what you are actually defending against when someone hands you a tray of BM-series chips.

For practical bench work, treat recycled and remarked as your primary threats, with defective riding along inside the same lots. Everything that follows is tuned to those three.

What a genuine BM-class chip looks like first

You cannot spot an anomaly until you know the baseline. Bitmain’s BM-series hashing ASICs (BM1387 through BM1370) are small leadless QFN-class packages — the BM1397 of the S17+/T17+ generation is documented as a QFN-34, for example. They are flip-chip-style parts soldered flat to the hashboard, with the top face carrying the laser marking and the underside carrying the solder pads and a central thermal/ground pad. They are tiny: a single hashboard such as an S21 carries 108 of them across 12 voltage domains, and an S19 Pro carries 114. Other manufacturers’ mining ASICs use entirely different package styles and footprints — so package style and size are themselves a first-order authenticity cue for a given model.

A few baseline behaviours matter for the electrical stage later. BM-series chips are daisy-chained on a shared UART bus; at power-up a broadcast SET_ADDR command assigns each chip a sequential address, so a healthy chain reports the full expected ASIC count in order. Each die also carries on-die temperature diodes. A recycled or damaged chip betrays itself precisely here — a gap in the address chain, an erratic nonce return, or a temperature reading that disagrees with its neighbours — and that is the cross-check that catches fakes which pass visual inspection.

Build your reference from known-good parts: pull two or three confirmed-genuine chips of the exact model — ideally from a sealed, traceable source — and photograph them under your microscope. Those images are your “golden sample,” and most steps below are comparisons against it rather than absolute judgments.

The bench authentication workflow

Work through these in order. The cheap, non-destructive checks come first; the destructive or instrument-heavy ones are last and only for high-value lots or disputes. None of this requires a lab to start — a decent stereo microscope, a jeweller’s scale, calipers, and isopropyl alcohol get you most of the way.

Step 1 — Marking and laser-etch consistency under the microscope

Use a stereo microscope at meaningful magnification — roughly 30x or more is a sensible practical minimum for marking inspection, and more is better. You are comparing the suspect chip’s top mark to your golden sample, looking for:

• Etch depth and character formation. Genuine laser marking is crisp, uniform-depth, and consistent across every character. Re-marked parts often show shallow, fuzzy, or doubly-struck characters, inconsistent stroke width, or “burn holes” where an imprecise laser over-fired.
• Ghost marks. Look for faint remnants of an old marking beneath or beside the new one. Ghosting is a near-certain sign the surface was resurfaced and re-marked.
• Font and logo geometry. Spacing, kerning, logo proportions, and the dot/period style differ subtly between a factory mark and a copied one. Overlay your golden-sample photo if your scope software allows it.
• Mark placement vs. pin-1 indicator. The orientation dot/notch and its relationship to the text should be identical across a genuine lot. Drift here is a red flag.

Step 2 — The solvent (acetone) rub test for blacktopping

“Blacktopping” is resurfacing the package top with a thin coating so it can be re-marked. The classic field test is a solvent rub. Dampen a cotton swab and rub a non-marked area of the package top:

• The formal benchmark is MIL-STD-883 Method 2015 (Resistance to Solvents), which specifies a mixture such as one part by volume of isopropyl alcohol to three parts by volume of mineral spirits. Genuine factory marking survives this unchanged.
• Many shops use acetone for a faster, harsher check: if the swab picks up black/grey residue, the surface texture changes, or the marking smears, the package was coated — i.e., remarked.
• Caveat (this is why you do not stop here): sophisticated counterfeiters now use acetone-resistant coatings specifically to beat this test. A clean rub is reassuring but not proof. Use it to confirm suspicion, not to grant a pass.

Always test a non-marking area first, and never soak the part — you are wetting a swab, not bathing the chip.

Step 3 — Date-code and lot-code cross-checks

Every legitimate chip’s markings encode a manufacturing date/lot. This is a logic test, not just a visual one:

• Internal consistency within a lot. A genuine reel or tray generally shares a tight date-code range. A “lot” with wildly scattered date codes — or, suspiciously, every single chip bearing the identical code — points to mixed salvage or mass remarking.
• Date code vs. silicon generation. A date code that predates the chip’s known introduction, or postdates a part that is end-of-life, is incoherent. Cross-reference the model/chip pairing against a trustworthy table (see our chip reference) before you accept the code.
• Date code vs. physical age. A “fresh” date code on a package showing oxidation, flux staining, or pad wear is the single most common recycled-chip contradiction. The mark says new; the metal says used.

Step 4 — Surface texture, edge, and pad inspection (the recycled tells)

This is where recycled chips fail even when the marking is convincing. Under the scope, walk the whole package:

• Sanding / micro-abrasion marks on the top surface — fine parallel scratches from resurfacing.
• Texture mismatch. Genuine mold compound has a consistent, slightly grainy factory texture. Recoated tops often look too smooth, too matte, or unevenly glossy. Image-processing researchers have built entire recycled/remarked classifiers around exactly this package-texture and indent signature.
• Edge and corner damage from desoldering tools, plus chipped mold compound.
• Underside pad / ball condition (the big one). A truly new QFN/BGA chip has clean, uniform, untouched solder pads. A recycled part shows oxidized pads, dull or unevenly sized reballed bumps, leftover solder, or flux residue in the pad field. For reballed parts, ball size and pitch should match the model — mining repair reballing typically uses 0.4 mm solder balls, and sloppy or mixed ball sizes are an immediate tell.
• Heat/oxidation tint. A faint straw or brown discoloration around the package edges or pads indicates the part has already been through one or more reflow cycles.

Step 5 — Weight and dimension checks

Counterfeits — especially recoated and repackaged parts — frequently drift outside the genuine tolerance band:

• Calipers. Measure length, width, and thickness against your golden sample. Resurfacing adds material; thickness is the most sensitive axis. Confirm the package family is even correct for the model — a BGA where the genuine part is a small QFN means you are not looking at the right chip at all.
• Mass. A jeweller’s scale resolving to 1 mg can flag outliers. Weigh ten known-good chips, take the mean and spread, then weigh suspects. A chip sitting well outside that band (coating adds mass; a hollowed/blank package removes it) warrants escalation.
• Coplanarity / flatness. Recycled chips that were yanked hot can be warped, which both fails this check and predicts bad reflow yield later.

Step 6 — Internal inspection: bond wires, die, and voids

If a lot is high-value or the visual stage is ambiguous, go inside. These methods need real instruments and are usually outsourced:

• X-ray. A 2D/CT X-ray reveals missing, broken, or misplaced bond wires, the die outline, and lead-frame geometry. The famous industry example is a counterfeit microcontroller whose X-ray showed an entirely different die inside — a blatant fake that looked perfect externally. X-ray a golden sample alongside the suspect and the internal layout should match.
• Scanning Acoustic Microscopy (SAM/CSAM). Acoustic imaging finds delamination and internal voids from thermal abuse, and can reveal original laser etching hiding under a blacktop coat — strong direct evidence of remarking.
• Decapsulation. Chemically or mechanically opening one sacrificial sample from the lot exposes the die for marking, geometry, and revision inspection. Destructive, so it is sample-based, but definitive.

Step 7 — Electrical binning after install

The final and, for mining, most decisive gate is putting the chips to work. Because BM-series chips self-enumerate on the chain, a test fixture or a populated hashboard will tell you the truth that paint cannot hide:

• Chain enumeration. After SET_ADDR, every chip should report at its expected address. A short count, or a chain that breaks at a specific position, isolates dead or counterfeit chips to a physical location.
• Per-chip nonce / hashrate. Read the chain via the miner API (for example a devs query to the cgminer/bmminer socket). A recycled or defective die produces low or zero accepted nonces and elevated hardware-error rates while its neighbours behave.
• Temperature-diode sanity. A chip whose on-die diode reads implausibly (cold while hashing, or runaway hot) is degraded or misrepresented. A thermal camera over a running board makes outliers obvious as cold or hot spots.
• Parametric drift. Counterfeit-detection literature repeatedly shows fakes failing AC/parametric limits — a part that hashes but only at abnormally high voltage or with poor efficiency is functionally suspect even if it “works.”

Electrical binning is what catches the well-disguised recycled chip that sailed through visual inspection. Budget for it — it is non-negotiable for any lot you intend to ship inside a customer’s miner.

A practical buyer’s and repair-shop checklist

One principle ties this together: industry standards such as SAE AS6081 exist precisely because no single test is sufficient. They mandate a layered regime — source control plus visual, mechanical, and electrical inspection — chosen against cost and risk. A home repairer cannot run the whole stack, but the cheap layers (microscope, solvent, calipers, scale) plus a final electrical bin already eliminate the overwhelming majority of recycled and remarked fakes.

Why this matters for refurbished and salvaged supply

There is nothing wrong with a recycled chip honestly sold as recycled and verified to still hash within spec — salvage and refurbishment are how the mining ecosystem keeps hardware out of landfill and keeps older fleets alive. The problem is misrepresentation: a degraded salvage chip sold as factory-new, dropped into a customer’s board, and failing weeks later. The whole point of authentication is to let buyers tell the difference and price accordingly.

Provenance and bench discipline are exactly the work a serious repair operation does on your behalf — incoming inspection, electrical binning, and honest grading. D-Central, like the broader open-repair and counterfeit-detection community whose methods this guide draws on, treats verification as table stakes rather than a selling point. If you are sourcing spares or a refurbished machine, you can have a hashboard or chip lot inspected, start a repair intake, or read up on buying used ASIC miners first. For the silicon facts behind every check above, the ASIC chip reference and the model hubs for the Antminer S19 family and Antminer S21 family confirm what a genuine part should be — the same bench-verified spirit as our public mining facts reference.

Frequently asked questions

Can I authenticate an ASIC chip without any lab equipment?

You can do a great deal with a stereo microscope at 30x or more, isopropyl alcohol or acetone, calipers, and a milligram scale — plus a confirmed-genuine “golden sample” to compare against. Those layers catch most recycled and remarked fakes. What you cannot do without instruments is internal inspection (X-ray, acoustic microscopy, decapsulation), so reserve those for high-value lots or disputes, and always finish with electrical binning on a test fixture or board.

Is a recycled chip always bad?

No. A salvaged chip that still hashes within spec and is sold honestly as recycled is legitimate and often a sensible, lower-cost choice. The danger is a degraded or defective salvage chip misrepresented as factory-new. Authentication is about correct disclosure and verified function, not about rejecting all reclaimed silicon.

The acetone rub came back clean — does that prove the chip is genuine?

It is reassuring but not proof. Counterfeiters increasingly use acetone-resistant coatings specifically to defeat the rub test. Treat a clean rub as one passed layer among several. Always combine it with marking inspection, surface and pad examination, date-code logic, and a final electrical test before trusting a part.

How do counterfeit chips most commonly enter the mining supply chain?

Overwhelmingly as recycled and remarked parts. Dead and burnt hashboards are abundant; chips are desoldered, cleaned, often reballed, sometimes repainted and re-laser-marked with a fresh date code, then sold loose or built into “new” boards. More than 80% of counterfeit components industry-wide are this category, and mining hardware fits the pattern exactly.

Why does electrical binning catch fakes that visual inspection misses?

Because paint and re-marking cannot fake function. BM-series chips self-assign sequential addresses on a shared bus at power-up, so a populated board or test fixture reports the true chip count, per-chip nonce rate, and on-die temperatures. A degraded or counterfeit die shows up as a missing address, low or zero accepted nonces, high hardware-error rates, or an implausible temperature — regardless of how perfect the package looks.

What standards should a serious buyer or shop reference?

For the solvent test, MIL-STD-883 Method 2015 (Resistance to Solvents) defines the benchmark mixture genuine marking must survive. For an overall program, SAE AS6081 codifies counterfeit avoidance for independent distribution — source control plus layered visual, mechanical, and electrical inspection. The academic literature (notably Tehranipoor and colleagues) supplies the taxonomy and defect-coverage framework behind this workflow.

Bitaxe Heatsink — High-Performance Aluminum Cooler for Bitaxe & Nerdaxe Gamma / Supra / Ultra — Silent Operation & Stable Overclocking" width="80" height="80" loading="lazy" style="width:80px;height:80px;object-fit:contain;border-radius:6px;background:#1A1A1A;flex-shrink:0;">

Premium Bitaxe Heatsink — High-Performance Aluminum Cooler for Bitaxe & Nerdaxe Gamma / Supra / Ultra — Silent Operation & Stable Overclocking $14.90 CAD

Shop Heatsinks →