Immersion Cooling Silicone Thermal Pad Oil Bleed

Read the dielectric vendor's published TIM compatibility list before any miner enters the tank. Engineered Fluids EC-100, BitCool BC-888, Castrol Tection HE Mining, Submer SmartCoolant, and Shell SDF all explicitly exclude silicone-based thermal interface materials. The only fluids that tolerate silicone TIMs are select two-phase fluorocarbons (3M Novec 7100 / 7200), which are a different cooling architecture entirely. If you skipped this at fill, this is your reset moment — every silicone pad in the tank is bleeding right now.

Pull one representative miner. Open the chassis. Identify thermal pads on the hashboards: pink, grey, pale-blue gum-textured pads = silicone = guaranteed bleed in any single-phase tank. Black graphite sheets, white phase-change waxy film, or factory-applied paste only = different conversation, different fix. The visual check costs five minutes and confirms the failure family.

Sample-jar `100 ml` of fluid from the lower third of the tank and compare against a sealed `1 L` reference of the original-fill fluid kept since day one. Hazy, blue-grey-cast, sheen-on-pour, or a visible cloud = contamination in progress. Document with a photo for your repair log; you will reference it again at the 30-day post-repair re-sample.

Quarantine the tank from new-miner additions until the repad-or-replace decision is made. Adding more silicone-padded miners doubles the rate of contamination — every additional pad area is more bleed surface. Run any pending miners on air cooling until they have been bench-prepped.

Email or web-form the dielectric vendor with your sample photos and lab analysis (if available). Engineered Fluids and Submer both run technical support that will quote a tolerance threshold and a remediation path specific to your fluid lot — and may quote a take-back / re-refine on the contaminated fluid that beats virgin replacement on cost.

Strip every silicone pad off every hashboard going back into the tank. Plastic spudger or wooden chopstick edge — never a metal blade. Silicone tears; if you nick PCB solder mask, you create a future short. Wipe down with `99%` IPA on a lint-free wipe; for stubborn residue, use a small splash of naphtha or xylene on a wipe (off the board), then dry-IPA-wipe to clear. PMICs, DDR memory packages, voltage-regulator ICs, controller IC, and any heat-spreader / shroud contact pad all need stripping.

Replace silicone pads with Honeywell PCM45F phase-change pads at primary IC-to-heatsink interfaces. PCM45F: `1.4 W/m·K` thermal conductivity, melts and reflows once at first thermal cycle to fill micro-gaps, paraffin-wax carrier matrix (no oil bleed in dielectric). Cut to component footprint, place dry, install — first power-on does the wetting. Engineered Fluids' published recommended TIM for in-fluid use on Antminer-class hardware.

Use Panasonic EYG-S graphite pads or Laird Tflex HD80000 series for secondary thermal paths. Graphite: `400+ W/m·K` in-plane, `8–18 W/m·K` through-plane, zero bleed. Cut with sharp scissors to component shape; install with no compression worry — graphite is conformable but doesn't squish-out. EYG-S0508 (`0.5 mm`) or EYG-S0612 (`0.6 mm`) covers the original silicone pad thickness range on Antminer-class boards.

Refresh ASIC die paste with an immersion-rated paste while you're already in there. Arctic MX-6, Thermal Grizzly Kryonaut, or Cooler Master MasterGel Maker Nano: all metal-oxide based, no silicone bleed component, rated for sustained `100 °C+` operation, single-phase compatible. Apply uniform thin layer over the ASIC die, reseat heatsink, torque to spec.

Document the repad on every miner that goes back into service: sticker on the chassis, log entry in the workshop spreadsheet, photo of the open hashboard. Six months from now when fluid analysis says it's clean, the documentation is what proves the fix held — and which units (if any) still need attention.

Drain the contaminated fluid into a sealed vendor-clean `200 L` or `1000 L` IBC. Label `CONTAMINATED — SILICONE PDMS — FOR RE-REFINE OR DISPOSAL` in marker on the tote. Engineered Fluids and Submer run fluid take-back / re-refine programs at lower cost than virgin replacement; for sub-`100 ppm` contamination they may quote multi-pass polish on their own equipment and return the fluid to spec. Do not pour contaminated dielectric down a drain — environmental compliance varies by jurisdiction.

Deep-clean the heat exchanger plate pack. Open gasket-clamped plate-and-frame exchangers, wipe each plate with naphtha-soaked lint-free until the slick film is fully gone (bare-metal-visible under shop light). Replace gaskets if they show silicone exposure damage — Buna-N and EPDM both swell in PDMS-loaded hydrocarbon, and re-using a swollen gasket invites a leak the next month. Brazed-plate exchangers can't open: backwash with hot naphtha through the dielectric loop, drain, repeat until effluent runs clean.

Repad and re-paste every hashboard in the system before refill — even units that haven't shown thermal drift yet. The silicone is bleeding from every pad in the tank, fast or slow doesn't matter. Run the repad workflow from steps 6–9 across the whole fleet. Skipping a single miner restarts the contamination clock the day refill is complete.

Refill with fresh dielectric from a verified-clean drum, same vendor and same SKU as the original fill (mixing PAO grades or PAO with GTL is a separate problem). Order against vendor SKU you started on, document the lot number and seal-photograph before broaching. Fill via filtered transfer (5 µm in-line) — virgin fluid still has trace particulates from drumming and shipping.

Burn-in for `24 hours` at nameplate, monitor thermal response, and re-sample fluid at `30` and `90 days`. Pull `50 ml` samples at both intervals, ship for FTIR (`$60–150 CAD` per sample). Two consecutive clean reports = repair held. Any returning siloxane signal = a miner went back into the tank with silicone pads still in place (audit your repad log) or a previously untouched system component (gasket, pump impeller, hose) is leaching residual PDMS — investigate that path.

Stop DIY when post-repad thermal numbers don't return to fleet baseline within `48 hours` of redeployment, lab analysis shows `>500 ppm` PDMS without storage capacity for take-back, gaskets are swollen / weeping, multi-tank thermal drift is farm-wide, or any hashboard shows capacitor bulging / burnt-component smell. At that point you need ultrasonic clean of heatsinks, IR reflow check on BGA arrays, and voltage-domain validation under programmable load — bench territory. Book a slot at https://d-central.tech/services/asic-repair/.

Multi-tank farms with confirmed `>500 ppm` PDMS load: schedule full fluid replacement across all tanks simultaneously, not tank-by-tank. Cross-contamination via shared pumps, transfer hoses, or service tools will recontaminate a freshly-filled tank within weeks if the surrounding farm infrastructure is still silicone-loaded. Strip the entire farm's plumbing in one campaign — pumps, hoses, gaskets, pump impeller seals all get audited together.

Ship affected boards to D-Central safely if remote bench review is needed: drain and IPA-wipe each board to dry, anti-static bag, double-box with `≥ 5 cm` foam every side. Include: dielectric brand and lot number, contamination level (if labbed), repad date and material, fleet baseline thermal numbers, current thermal numbers post-repad. Diagnostic time saved at the bench equals bench cost saved equals your repair bill saved.