Nerd Family – Thermal Paste Degradation After 6-12 Months
Informational — Monitor and address as needed
Symptoms
- Miner has been running 24/7 continuously for 6+ months without a heatsink removal or paste refresh
- Chip temperature in dashboard reads 5+ C higher than at commissioning, at the same OC profile, ambient, and fan duty
- Hashrate is 3-12% below the value measured during the first month of operation, with no other change to OC, Vcore, pool, or environment
- Fan duty has crept upward over months - the miner is audibly louder than it used to be without conscious tuning
- Heatsink fins feel only mildly warm to the touch while dashboard reports a die temperature that should make them hot - die-vs-heatsink TIM gradient
- Ambient at the intake (IR thermometer) is the same as commissioning - the room hasn't gotten warmer; the miner has gotten worse at moving heat
- On NerdQAxe / Hex / Octaxe / GT (multi-chip variants): per-chip telemetry shows one or two chip positions creeping 3-8 C hotter than the others - localized pump-out
- Miner occasionally trips firmware self-protect latch (Bitaxe AxeOS at 75 C, NerdQAxe-OS at variant-specific threshold) under summer ambient where it never tripped last year
- You re-pasted recently with stock or unknown silicone paste; temperatures came down at first then crept back up within 90 days - silicone pump-out signature
- Visible inspection on heatsink removal shows: cracked, flaky, or chalky paste; bald spot or thinning over the chip's hot centre; paste squeezed entirely out from under the contact patch
- Power draw at the wall is unchanged (chip still drawing rated wattage) but reported hashrate is lower - heat generated but not effectively rejected, chip is throttling silently
- No hard fault yet - no Device Overheat banner sustained, no chip-not-detected error, no shutdown - but the trend lines are pointing the wrong way
Step-by-Step Fix
Pull historical baseline. Open AxeOS / NerdQAxe-OS / NerdAxe / NerdMiner dashboard. Note today's chip temperature, OC frequency, Vcore, fan duty, and reported hashrate at steady state (20+ minutes from cold boot). Compare against commissioning numbers. If chip temp is 5+ C higher at the same OC + ambient + fan duty, AND hashrate is 3+% lower, TIM degradation is the leading hypothesis. If you have no baseline, take baseline now, set a calendar reminder for 90 days, and revisit.
Confirm the die-vs-heatsink gradient. With the miner at steady-state hashing for 20+ minutes, read the dashboard chip temperature, then carefully touch a heatsink fin. Hold a finger on the fin for 5 seconds: pain in < 2 s = > 65 C heatsink; comfortable for 5+ s = < 60 C heatsink. Dashboard > 75 C AND heatsink fin < 60 C = TIM bottleneck confirmed. Dashboard hot AND heatsink hot = airflow problem, not TIM.
Verify ambient and fan haven't changed. IR thermometer at the intake - target <= 25 C for Bitaxe-class, <= 30 C for NerdMiner-class. Confirm fan RPM matches nameplate. Visually inspect heatsink fins for dust - a 2 mm dust mat reduces heat rejection 30-40% and looks identical to TIM degradation in dashboard data. Compressed-air the fins from the exhaust side. Re-test for 48 hours after dusting before tearing into hardware.
Check per-chip spread on multi-chip variants (Hex / GT / NerdQAxe / Octaxe). Spread < 3 C between chips: uniform TIM degradation across all positions, whole-board re-paste. Spread 5-10 C on one or two positions: localized pump-out, whole-board re-paste with extra attention to even mount pressure on those positions. Spread > 12 C on one position: localized pump-out very likely, but rule out chip-level fault if re-paste does not close the spread.
Estimate paste age and decide. Stock paste < 6 months: TIM probably not the issue. Stock paste 6-12 months: dead-centre of typical degradation window, re-paste justified preventively. Stock paste > 12 months: well past service life, re-paste with PTM7950 / MX-6 / Kryonaut not back to stock. MX-6 / Kryonaut > 36 months: refresh window. PTM7950 > 60 months: optional refresh, usually still healthy.
Tier 1 - gather supplies. Isopropyl alcohol 99% (NOT 70% rubbing alcohol), lint-free wipes (microfibre or paper, not cotton balls), Phillips #1 + Torx T8 driver depending on variant, plastic spudger for adhesive-bonded heatsinks, your chosen TIM (Arctic MX-6 first re-paste, Kryonaut premium, PTM7950 set-and-forget). Total cost $15-45 CAD if starting from zero, $5-15 per subsequent re-paste once you own supplies.
Photograph the assembly from multiple angles before disassembly - top-down, side, every screw orientation. Reference for reassembly. Power off, disconnect power, wait 60 seconds to discharge residual capacitance. Move to a clean, well-lit, dust-free workspace. Paste contamination from a dirty workspace is a real failure mode.
Remove heatsink mechanism. Most Bitaxe + NerdAxe variants use 2-4 push-pins or screws-with-springs from the underside. On thermal-adhesive bonded variants (some early NerdQAxe, some clone Bitaxe heatsinks), warm the heatsink to ~60 C with a hairdryer for 30 seconds to soften adhesive, then gently pry with a plastic spudger. Never metal, never force - if it does not come off easily, stop and ship to D-Central. A broken die or lifted BGA pad costs 10x what a heatsink swap costs. Multi-chip variants: note which chip is which position before separation.
Clean both surfaces thoroughly. IPA 99% on a lint-free wipe, every trace of old paste off die top and heatsink baseplate. Repeat with fresh wipes until they come away clean. Inspect under good light - both surfaces should look mirror-clean with no visible paste streak. Old paste residue under new paste creates an air gap that defeats the entire job - this step is non-negotiable.
Apply new paste using the rice-grain method. For MX-6 / Kryonaut: small rice-grain (~2 mm^2) at the centre of each die. Do NOT pre-spread - heatsink mounting pressure spreads it uniformly into the engineering target 50-100 micron bondline. For multi-chip variants (Hex, NerdQAxe, Octaxe, GT): one rice-grain per chip, NOT one big blob across all positions.
Reseat heatsink with even pressure. Lower straight down - no twisting or sliding (which moves paste off-centre and creates dry spots). Engage push-pins or screws in a cross / star pattern, tightening each a partial turn at a time so pressure builds evenly across the contact patch. Final torque: finger-tight + a quarter turn for screws-with-springs, full click for push-pins. Do NOT over-tighten - bending the PCB cracks BGA solder joints, which is far worse than warm paste.
Tier 2 - PTM7950 phase-change pad install (the long-term fix). Cut a square of PTM7950 slightly smaller than the die top with a clean craft knife on a plastic cutting mat - ~1 mm margin all around the chip is fine, do not let it overhang BGA edges. Multi-chip variants: cut one square per chip, do not try to use one large pad across multiple positions. Peel one liner, position pad-side-down on cleaned die, press gently to seat, peel second liner. Avoid bubbles and skin/finger contact - oils contaminate the bondline.
Verify the result. Reconnect, boot, observe 30 minutes of steady-state hashing. Record the new chip temperature. Expected drop on a miner with degraded TIM: 8-18 C. First boot with PTM7950 will show momentarily elevated temps (75-80 C for ~2 minutes) as the pad melts and forms its bondline - normal phase-change behaviour. After phase change, expect rock-stable temps for 5+ years with no pump-out. If improvement is < 5 C, your mount pressure or paste application is wrong - redo from Step 9.
Document and schedule. Photograph the new dashboard reading, note the date in your records. Set a calendar reminder for 36 months (MX-6 / Kryonaut) or 60 months (PTM7950) from today. Set a personal monitoring threshold at 72 C (Bitaxe-class) or 5 C below firmware self-protect (NerdQAxe / Octaxe) - alerts via Grafana, Home Assistant, or a Python cron on the dashboard JSON API.
Tier 3 - audit your heatsink for flatness. A stamped aluminium clone heatsink often has a warped or non-flat baseplate - point contact rather than full-surface contact, no paste fixes that. Test: place baseplate face-down on a known-flat surface (granite countertop or plate glass). If you can rock it, it is warped. Upgrade to a D-Central machined Bitaxe / Hex heatsink at d-central.tech/product-category/bitaxe/ - we pioneered the originals for both standard Bitaxe and Bitaxe Hex, flat-lapped baseplates spec'd for BM1366/1368/1370 contact patches.
Tier 3 - install the upgraded heatsink with PTM7950 + improve airflow ducting. D-Central machined heatsink + PTM7950 is the gold standard for 24/7 home-miner duty. Same Tier 1 cleaning + Tier 2 pad install workflow. Add a 3D-printed or sheet-metal shroud channelling fan air directly across the heatsink fins (worth 3-8 C). The original D-Central Bitaxe Mesh Stand - which we manufactured first, before any clones - was designed with airflow geometry in mind. Pair with a known-good fan (Noctua-class).
Tier 3 - tune OC/UV around the new thermal headroom + multi-chip per-position rework. Re-paste + flat heatsink + good airflow earns 10-20 C of margin. Use 5 MHz of that for higher frequency, OR use it for lower Vcore (run cooler, run quieter, run longer). Take half the margin as performance, leave the other half as longevity insurance. Underclock-and-undervolt operators see 2-3x chip lifetime. On Hex / NerdQAxe / Octaxe, if one position still lags by > 5 C after Steps 12-16, redo that one chip with extra care: clean, fresh PTM7950, deliberate even pressure on that corner.
Tier 4 - stop DIY and ship to D-Central if any of: die shows visible heat damage (scorching, discoloration, lifted edges, BGA-pad damage); heatsink is thermal-adhesive bonded so aggressively it cannot be removed without pulling solder off the PCB; after a clean Tier 3 the chip still reports > 80 C at stock OC/Vcore (chip degraded, not TIM); multi-chip variant shows > 15 C per-chip spread on one position after clean Tier 3 (cracked thermal vias suspected); capacitor bulging, MLCC cracking, or burnt-component smell. Ship in anti-static bag, double-boxed with >= 5 cm foam, include AxeOS / NerdQAxe-OS screenshots, what TIM you used + when, description of symptoms. Book ASIC Repair at d-central.tech/services/asic-repair/. Turnaround 5-10 business days, Canada / US / international.
When to Seek Professional Repair
If the steps above do not resolve the issue, or if you are not comfortable performing these repairs yourself, professional service is recommended. Attempting advanced repairs without proper equipment can cause further damage.
Related Error Codes
Still Having Issues?
Our team of Bitcoin Mining Hackers has been repairing ASIC miners since 2016. We have seen it all and fixed it all. Get a professional diagnosis.
