PulsCar analysis of NHTSA public data, September 2026 GM full-size V8 valvetrain and bottom-end noise: 424 owner complaints, 59 manufacturer bulletins, 2014-2026 model years Licensed CC BY 4.0. Queries run 18 September 2026. -------------------------------------------------------------------------- SOURCE FILES -------------------------------------------------------------------------- NHTSA ODI complaint flat files COMPLAINTS_RECEIVED_2015-2019, _2020-2024, _2025-2026 https://static.nhtsa.gov/odi/ffdd/cmpl/ (field dictionary: CMPL.txt) 777,040 distinct vehicle complaints (ODINO, product type V) received 1 January 2015 to 8 September 2026. NHTSA manufacturer communication (TSB) flat files TSBS_RECEIVED_2015-2019, _2020-2024, _2025-2026 https://static.nhtsa.gov/odi/ffdd/tsbs/ (field dictionary: TSBS.txt) Fields used: 1 NHTSA document ID, 4 bulletin number, 5 manufacturer communication date, 8 make, 9 model, 10 model year, 14 summary. NHTSA recall flat file FLAT_RCL.zip and the ODI investigations file (action number, make, model, year, opened, closed, subject). -------------------------------------------------------------------------- THE UNIVERSE (22,866) -------------------------------------------------------------------------- Every complaint in the file for the eight GM full-size V8 nameplates that share the 5.3L and 6.2L EcoTec3 / L87 family, model years 2014 to 2026: Chevrolet Silverado 1500 8,584 GMC Sierra 1500 5,644 Chevrolet Tahoe 2,400 GMC Yukon 2,116 GMC Yukon XL 1,537 Chevrolet Suburban 1,514 Cadillac Escalade 691 Cadillac Escalade ESV 381 Heavy-duty models (2500, 3500) are NOT in the universe: they do not carry the engines this page is about. -------------------------------------------------------------------------- WHICH COMPLAINTS COUNT (the noise set: 424) -------------------------------------------------------------------------- A complaint counts when its narrative contains BOTH a noise word: TICK | TICKS | TICKING | TICKED | TAP | TAPS | TAPPING | KNOCK | KNOCKS | KNOCKING | CLATTER | CLATTERING and a valvetrain or bottom-end part word: LIFTER(S) | CAMSHAFT | CAM SHAFT | PUSHROD(S) | PUSH ROD(S) | ROCKER ARM | VALVETRAIN | VALVE TRAIN | AFM | DFM | ACTIVE FUEL MANAGEMENT | DYNAMIC FUEL MANAGEMENT | ROD BEARING(S) | CONNECTING ROD | CRANKSHAFT | CRANK BEARING | MAIN BEARING | SPUN BEARING Matching is case-insensitive on the uppercased narrative, with word boundaries, over the WHOLE narrative rather than one sentence. WHY NOT A SENTENCE WINDOW. A stricter version requiring the noise word and the part word inside the same sentence returned 88 complaints. Reading them showed the loss is not noise but structure: NHTSA's call-center records are written as "there was a tapping sound coming from the engine compartment. The vehicle was taken to a dealer ... the lifters were inoperable" — symptom and diagnosis are always two sentences apart. The sentence window was dropped and the false-positive rate measured instead (below). SAMPLE READING (required before publishing any keyword count) Two rounds of 30 randomly drawn matches were read in full. Round 1 (seed 11): 1 of 30 was not a valvetrain-noise report -- a fuel economy complaint that mentioned an oil pump and a knock in passing. Round 2 (seed 29): 0 of 30 were off topic. Measured false-positive rate: 1 in 60, about 1.7%. One genuine artefact was found instead: the same 2026 Silverado narrative appears twice under two ODINOs (the owner filed the same text on the same day). Duplicate filings are not de-duplicated anywhere in this analysis. WHAT THE SET IS NOT. It is not "lifter failures". It is complaints whose author used a noise word and a valvetrain word. It includes trucks whose noise turned out to be something else, and it excludes engines that failed silently. Nothing in it is a failure rate: NHTSA publishes no denominator, so there is no way to divide by the number of trucks on the road. -------------------------------------------------------------------------- MILEAGE, TOW, CRASH AND FIRE -------------------------------------------------------------------------- Mileage: NHTSA's MILES field, rows with 0 < miles < 400,000 only. Only 75 of the 424 (17.7%) carry a usable mileage, against 392,935 of 777,040 (50.6%) across the whole file. The gap is a property of the form: long free-typed narratives, which is what this set is full of, often leave the structured field empty and put the odometer reading in the text. Of the 75: median 86,000; first quartile 48,650; third quartile 110,000; 22.7% at or under 36,000 miles; 32.0% at or under 60,000. All-complaint baseline: median 70,000. SEVENTY-FIVE IS A SMALL SAMPLE. The per-model-year medians in lifter-knock-by-model-year.csv rest on 0 to 18 readings each and are published as an indication of direction, not as a statistic. Structured flags (TOWED / CRASH / FIRE, share answering Y): noise set: towed 7.08%, crash 0.00%, fire 0.24%, 0 injuries, 0 deaths baseline (777,040): towed 5.11%, crash 4.37%, fire 1.82% A separate count of narratives that use a towing word (TOW, TOWED, TOWING, FLATBED, WRECKER) reaches 27.1% -- the structured flag understates towing badly, and both figures are published together for that reason. -------------------------------------------------------------------------- PHRASE SUBSETS (share of the 424 whose narrative matches) -------------------------------------------------------------------------- lifter / valve lifter oil manifold / VLOM 314 74.1% camshaft or cam lobe 129 30.4% crankshaft or connecting rod 68 16.0% rod / main / crank / spun bearing 48 11.3% bent or broken pushrod 44 10.4% rocker arm 9 2.1% AFM or DFM named as the cause 58 13.7% engine replaced or quoted for replacement 96 22.6% lifters replaced, engine kept 19 4.5% metal debris in the oil 23 5.4% misfire or a P030x code 61 14.4% P0016 / cam-crank correlation 3 0.7% cold start or first start of the day 16 3.8% oil consumption 10 2.4% stalled, seized, stranded or lost power 148 34.9% no warning light 66 15.6% check engine light mentioned 111 26.2% mentions the word "recall" 120 28.3% names 25V-274 or an N252494 bulletin 15 3.5% says the truck is not in the recall / not covered 28 6.6% shop could not duplicate it or called it normal 4 0.9% names the 6.2 or L87 96 22.6% names the 5.3, L84 or L83 35 8.3% Categories overlap; they do not sum to 100%. "Names the 6.2" counts the words the owner typed, not the engine in the truck -- most narratives name no engine at all, so these two rows are a floor, not a split. -------------------------------------------------------------------------- OWNER-REPORTED REPAIR AMOUNTS (62 complaints) -------------------------------------------------------------------------- A dollar figure counts when: it is between $500 and $40,000; a repair word (REPAIR, REPLAC, QUOT, COST, ESTIMATE, BILL, PAID, CHARGE, ENGINE, FIX, OUT OF POCKET) appears within 90 characters; and a purchase word (PURCHAS, BOUGHT, BUY, MSRP, BRAND NEW, TRADE, LOAN, FINANC, WORTH, VALUE, PRICE OF THE) does NOT appear within 90 characters unless a repair word does too. Where one narrative carries several qualifying figures the largest is used. n = 62. Median $11,184. Quartiles $7,832 and $15,000. Range $1,200 to $33,000. 66.1% are $10,000 or more. The purchase-word filter was added after a first pass returned $60,000, $37,000 and $35,000 figures that reading showed were the price of the truck, not of the repair. THIS IS A TAIL, NOT A PRICE. Sixty-two owners out of 424 chose to type a dollar amount into a federal safety complaint, and people do that over a five-figure engine, not over a $200 oil change. The distribution in lifter-knock-owner-reported-amounts.csv is the shape of the expensive end of this failure, and the article says so wherever it quotes it. -------------------------------------------------------------------------- MANUFACTURER BULLETINS (59 documents, 16 families) -------------------------------------------------------------------------- Same eight nameplates plus SUBURBAN 1500 (an alternate spelling in the file), model years 2014-2026, where the bulletin summary names a noise word (TICK, TICKING, TAP, TAPPING, KNOCK, KNOCKING, CLATTER, NOISE) and a valvetrain or bottom-end part word. One bulletin = one NHTSA document ID; the flat file carries one row per make/model/year, so counting rows would multiply one document by its vehicle list. 59 distinct documents. Grouping by bulletin number with the revision letter stripped gives 16 families -- see lifter-knock-gm-bulletins.csv. The largest is 15-06-01-002, the AFM lifter and valve lifter oil manifold bulletin: 13 documents from revision C (10 August 2016) to revision O (30 March 2023). Summaries are NHTSA's record of the manufacturer's document. Where the article quotes a bulletin word for word it says whether the words come from the summary record or from the PDF on static.nhtsa.gov. -------------------------------------------------------------------------- COMPLAINTS BY QUARTER, AND WHY THE SHAPE IS NOT A FAILURE CURVE -------------------------------------------------------------------------- lifter-knock-by-quarter-received.csv counts the 424 by the quarter NHTSA received them. Ten complaints arrived in the whole of 2024; 52 arrived in 2025 Q1, the quarter NHTSA opened PE25-001 (16 January 2025), and 75 in 2025 Q2, the quarter GM filed recall 25V-274 (24 April 2025). 2026 has not returned to the old level: 144 in three quarters. A recall is published, owners read about it, and owners who already had the noise file a report -- often about a truck that failed years earlier. The curve therefore measures ATTENTION as much as failure, and the CSV carries the recall-mention column beside the count so a reader can see the two move together (0 of 10 complaints in 2024 used the word "recall"; 21 of 75 did in 2025 Q2). -------------------------------------------------------------------------- LIMITATIONS -- READ BEFORE QUOTING ANY NUMBER -------------------------------------------------------------------------- 1. Complaints are self-reported, unverified and voluntary. They are not failure rates and cannot be turned into one. 2. Counts follow sales volume and press coverage. Nothing here says any truck is more or less reliable than another. 3. The keyword method carries about one mis-sort in sixty on this set (measured, see SAMPLE READING). Small counts -- the 2025 and 2026 model year rows, the Escalade ESV row -- are a handful of owners, not a rate. 4. Mileage, towed, crash and fire are the owner's own entry. Towed and crash shares of any complaint set are inflated relative to the road, because a frightening event is what prompts a filing. 5. Duplicate filings of the same narrative exist and are not removed. 6. Engine displacement is not a field in the complaint file. Every statement here about the 6.2 or the 5.3 counts the words an owner typed. The authoritative engine-by-engine boundary is the recall's own build window, not this data. -------------------------------------------------------------------------- FILES -------------------------------------------------------------------------- lifter-knock-by-model-year.csv model_year, complaints, pct_of_set, miles_reported, median_miles, pct_narrative_says_towed, pct_engine_replaced_or_quoted, pct_names_6_2_or_L87. lifter-knock-by-model.csv make, model, complaints, pct_of_set, all_complaints_2014_2026_same_models, pct_of_that_nameplates_complaints, median_miles, miles_reported, top_model_year, complaints_in_top_year. pct_of_that_nameplates_complaints is the ONE normalization available here: the share of everything ever filed about that nameplate in these model years that is a valvetrain-noise report. It compares vocabularies, not vehicles, and it is still not a failure rate. lifter-knock-by-quarter-received.csv quarter_received, complaints, cumulative, mentions_recall, names_25V274_or_N252494. lifter-knock-what-owners-name.csv what_the_narrative_names, complaints, pct_of_424. lifter-knock-owner-reported-amounts.csv band, complaints, pct_of_62. lifter-knock-gm-bulletins.csv bulletin_family, documents_in_family, first_date, last_date, example_bulletin_no. No personal data is published: no VIN, no city, no dealer name, no owner free text beyond short quoted snippets in the article itself.