Passer au contenu

Nous améliorons nos opérations pour mieux vous servir. Les commandes sont expédiées normalement depuis Laval, QC. Questions? Contactez-nous

Bitcoin accepté au paiement  |  Expédié depuis Laval, QC, Canada  |  Soutien expert depuis 2016

NO_COOLER_LOOSE Info

NerdOctaxe – Thermalright AXP90 Cooler Loose

One or both Thermalright AXP90-X53 coolers on the NerdOctaxe Gamma have lost seating pressure in shipping or handling. Spring screws backed off, thermal pad migrated, or the backplate shifted. Left uncorrected, powering on dumps ~208 W through 8 BM1370 chips into a loose heatsink and cooks one or more chips within minutes. Inspect-before-power-on is mandatory.

Informational — Monitor and address as needed

Affected Models: NerdOctaxe Gamma (dual Thermalright AXP90-X53 cooler revision)

Symptoms

  • Package arrived with visible impact marks, crushed corner foam, or the miner loose inside the outer carton
  • One or both `AXP90-X53` coolers visibly tilted, raised on one side, or not parallel to the PCB
  • Cooler rocks, shifts, or lifts perceptibly when you press or rock it with a finger
  • Spring screws proud of the PCB / backplate: threads visible above the spring washer instead of fully seated
  • Audible rattle or tick when the chassis is shaken gently (metal clip or screw, not the fan)
  • Gap visible between cooler cold-plate and chip surface under side lighting
  • Thermal pad extruded out the side of the cooler base, or pinched unevenly
  • After power-on: `ASIC_TEMP` spikes above 75 degrees C within 2 minutes, `Overheat Mode` banner in NerdOS
  • After power-on: chip count drops from 8 to 4 / 5 / 6 / 7 in the NerdOS UI
  • One cooler sits noticeably lower than the other (asymmetric bank height post-shipping)
  • Tower-fan clip appears displaced or off-center relative to the heatsink fins
  • Both coolers pass visual inspection but one rocks under finger pressure (silent failure case)

Step-by-Step Fix

1

Before connecting power, perform the finger-rock test. Open the box, find the NerdOctaxe, locate the two `AXP90-X53` towers. With a clean finger on the top of each fin-stack, gently rock each cooler left-right and front-back. You are looking for any perceptible movement. A correctly-seated cooler feels glued to the PCB. If either one shifts at all, do not power on. This 30-second test catches most shipping-mount failures before they become repair tickets.

2

Visually check cooler tilt from the side at eye level. Each cooler base plate should sit parallel to the PCB (imagine a ruler across the PCB — the base plate should match it). Any visible tilt means one or more screws backed out in shipping. Do not power on until you have retorqued.

3

Check for rattles with the power disconnected. Gently shake the chassis. Metallic ticks that move with motion indicate a cooler clip or spring screw is loose. The LILYGO display and the chassis itself are silent when everything is tight. Rattles mean open the chassis and inspect.

4

Inspect the eight spring screws (four per cooler) in good light, looking straight down. A correctly-seated screw has the spring washer fully compressed against the cooler base, with no visible thread between washer and base. If you see bare thread on any screw, that screw has backed out and needs retorquing.

5

Photograph the state before you touch anything. Take an overhead shot of each cooler plus a close-up of each spring screw. If you end up shipping back to D-Central, those photos cut diagnostic time in half. Also useful if you discover a factory-assembly issue and want to open a support ticket.

6

Gather tools. A quality Phillips #1 or JIS #1 screwdriver fits the `AXP90-X53` spring screws. A cheap worn-out #2 will cam out and round the head. An ESD wrist strap is recommended but not mandatory if you ground yourself on the chassis first.

7

Retorque in cross-pattern, top-down. Tighten the four spring screws on the first cooler in this order: top-left, bottom-right, top-right, bottom-left. Each pass is 1/4 turn. Do three passes per cooler (three cycles through the four screws). Stop when the spring washer is fully compressed and the cooler base is flush against the chip bank. Do not crank past that point — extra torque bends the PCB and cracks solder joints.

8

Repeat the cross-pattern retorque for the second cooler. Confirm the finger-rock test passes on both coolers before closing up the chassis. A correctly-retorqued cooler should feel completely solid, with no perceptible movement under light finger pressure.

9

Re-seat the tower fan clips. Each `AXP90-X53` has a 92 mm PWM fan held to the fin-stack by two wire clips. Pop them off, visually inspect fan blades for shipping damage, press the fan back onto the fin-stack, reattach both clips firmly. The fan should not rotate freely in your fingers when clipped.

10

Dry power-on smoke test. Connect the 12 V / 18-20 A XT60 PSU (minimum 216 W — do not use an undersized Bitaxe or NerdQAxe PSU; the NerdOctaxe will boot on 120 W but sag catastrophically at full hashrate). Power on and watch the LILYGO display boot, join WiFi, and start hashing. Within 5 minutes all 8 chips should read 52-55 degrees C in the NerdOS dashboard. Any chip 15 degrees C hotter than bank-mates means that corner has lost contact — power off and back to Step 7.

11

If a chip runs hot post-reseat, remove the failing cooler. Power off, unplug PSU. Remove the four spring screws on the problem cooler. Lift the cooler straight up, do not twist. Examine the thermal pad on the underside of the cold-plate: it should show four clean rectangular imprints from the four chip tops, uniform depth, silicone still slightly tacky. A bad imprint means bad pad.

12

Replace the thermal pad. Peel the old pad off the cooler base. Clean both the cooler base and the four chip tops with IPA 99% and lint-free wipes until completely clean. Cut a fresh 1.5 mm silicone thermal pad (Thermalright Odyssey 12.8 W/mK, Arctic TP-3, or equivalent high-conductivity pad) to match the cooler footprint. Place the pad on the cooler base. Lower the cooler straight down onto the four-chip bank — do not slide or twist. Retorque per Step 7.

13

Alternative: liquid thermal paste on paste-capable revisions. If your NerdOctaxe revision shipped with thermal paste rather than pads (early batches / some service replacements), use Arctic MX-6 or Thermal Grizzly Kryonaut, one small rice-grain bead per chip top (four beads per cooler). Mount cooler straight down. Paste handles die-height tolerance worse than a pad, so only use paste if that is what your unit shipped with or D-Central confirmed the swap.

14

Inspect backplate standoffs for stripped thread. If a screw spins freely without compressing the spring in Step 7, the brass standoff in the backplate has stripped. Loosen all four screws, remove the cooler, flip the chassis, pull the backplate, and inspect. A stripped standoff is Tier 4 territory — not field-repairable without specific replacement hardware.

15

Log everything for support and future reference. Photograph the pad imprint pattern before disposal, photograph the reassembled cooler, and record the per-chip temperatures post-fix. If the issue recurs on your unit, D-Central support uses that data to determine whether it is a one-off shipping event or a batch-level assembly concern worth flagging upstream.

16

Stop DIY and book a D-Central Open-Source Miner Repair slot when a spring screw strips its backplate thread, the PCB shows visible flex or cracking near a chip bank, a chip count drops permanently (8 to 6 or fewer, even after re-seat), or a post-power-on event killed a chip before you caught the loose cooler. You are past reseat territory at that point. Service intake at https://d-central.tech/services/asic-repair/.

17

D-Central bench process: full diagnostic with thermal imaging under load to confirm which chips have contact, continuity test to confirm no chip is electrically dead, optional BGA reflow if a chip position is compromised, BM1370 chip replacement from graded inventory if silicon is dead, full backplate or standoff replacement if stripped, and post-repair 24-hour burn-in at user-specified overclock with logged temps. Laval workshop — the same bench that built the unit.

18

Ship safely for repair. Anti-static bag, double-box with at least 5 cm of foam on every side, and add extra foam directly above each cooler fin-stack to stop cooler travel during a second round of shipping. Include a note with: date received, what you observed (which cooler rocked, which chip ran hot), whether you powered on before catching the loose mount, and a phone number. That metadata cuts 30-45 minutes off diagnostic time, which keeps the invoice low.

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.