{"product_id":"antminer-s19k-pro-hashboard-repair-service","title":"Antminer S19k Pro Hashboard Repair Service","description":"\u003cp\u003e\u003cstrong\u003eWhen one Antminer S19k Pro hashboard comes up short, stock firmware does not run on the other two: it powers the machine down, so a single failed board reads as a dead miner.\u003c\/strong\u003e The log names the board and the count. A healthy board reports 77 chips, and any other number is a fault. That board is repairable.\u003c\/p\u003e\n\n\u003cp\u003eAntminer S19k Pro hashboard repair is the whole job on one board: diagnosis, component-level repair, and validation under real mining load before it goes back. Bitmain built this model in two mechanically different versions and we take one of them, which is worth a look before you ship anything. Beyond that, working out what failed is not your problem — that is what intake is for, in our USA-based lab.\u003c\/p\u003e\n\n\u003ch2\u003eTechnical Diagnostics and Common Issues with Antminer S19k Pro\u003c\/h2\u003e\n\n\u003ch3\u003eQuick Symptom Checklist\u003c\/h3\u003e\n\u003cp\u003eThese are the lines S19k Pro owners send us, taken from real stock-firmware logs. The chain numbers and counts differ from machine to machine — match the shape of the line, not the digits. The machine numbers its boards from zero, so \u003ccode\u003eChain[2]\u003c\/code\u003e is the one you would call the third.\u003c\/p\u003e\n\u003cul\u003e\n\u003cli\u003e\n\u003ccode\u003eChain[2]: find 6 asic, times 0\u003c\/code\u003e repeating as \u003ccode\u003etimes 1\u003c\/code\u003e and \u003ccode\u003etimes 2\u003c\/code\u003e — the firmware tries three times before it gives up, so a single line is not yet a verdict\u003c\/li\u003e\n\u003cli\u003e\n\u003ccode\u003eChain 2 only find 6 asic, will power off hash board 2\u003c\/code\u003e, followed by \u003ccode\u003eERROR_SOC_INIT: soc init failed!\u003c\/code\u003e and \u003ccode\u003e****power off hashboard****\u003c\/code\u003e\n\u003c\/li\u003e\n\u003cli\u003e\n\u003ccode\u003eEEPROM error: CRC 3RD REGION\u003c\/code\u003e on one chain, or on all three\u003c\/li\u003e\n\u003cli\u003e\n\u003ccode\u003eERROR_UNBALANCE: Unbalance happened, waiting!\u003c\/code\u003e after \u003ccode\u003eSweep error string = N:3.\u003c\/code\u003e, on a miner that reached full speed first\u003c\/li\u003e\n\u003cli\u003eThe dashboard reporting a fan problem while every chain still shows its full 77 chips\u003c\/li\u003e\n\u003cli\u003eThe miner initialising, hashing for a few seconds, then shutting its own boards down\u003c\/li\u003e\n\u003c\/ul\u003e\n\n\u003ch3\u003eModel-Specific Patterns We See on Antminer S19k Pro\u003c\/h3\u003e\n\u003cp\u003e\u003cstrong\u003eThe EEPROM error is real and it is not the one that names your board.\u003c\/strong\u003e On a machine with one dead chain, \u003ccode\u003eEEPROM error: CRC 3RD REGION\u003c\/code\u003e appears on all three chains — including the two that then go on to find 77 chips each and work perfectly. Owners read three errors and conclude three dead boards, or that the boards need rewriting. The count lines are what identify the failure; the CRC line is noise on top of it.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eA fan error can be a board, and a board error can be a fan.\u003c\/strong\u003e When a fan drops out, and when the machine protects itself on temperature, the firmware cuts power to every hashboard — which is also what happens when a chain comes up short. All three end in the same place: zero hashrate and boards powered down. From the interface they are indistinguishable, which is why owners replace a healthy fan on a board fault, and ship a healthy board on a fan fault.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eHalf the alarming lines in the log are on healthy machines too.\u003c\/strong\u003e \u003ccode\u003eOpen miner sn file \/config\/sn error\u003c\/code\u003e, \u003ccode\u003eget test_loop_securely_find_asic_num failed\u003c\/code\u003e and \u003ccode\u003ebad chain id = 3\u003c\/code\u003e all contain a word like failed or bad, all appear during a completely normal boot, and the last one is simply the firmware looking for a fourth chain on a three-chain machine. So is the wait: from cold this model prints \u003ccode\u003eWait for hashboard cool, please wait, this may take up to 2 minutes...\u003c\/code\u003e and then warms the boards to temperature before it hashes, which takes ten minutes or more. Owners in a cold room decide it has hung and power-cycle it, which starts the wait again.\u003c\/p\u003e\n\n\u003ch3\u003eHardware Notes\u003c\/h3\u003e\n\u003ctable style=\"width:100%;border-collapse:collapse;table-layout:fixed;font-family:sans-serif;font-size:14px;margin:16px 0 24px;\"\u003e\n\u003ccolgroup\u003e\n\u003ccol style=\"width:32%\"\u003e\n\u003ccol style=\"width:68%\"\u003e\n\u003c\/colgroup\u003e\n\u003cthead\u003e\n\u003ctr style=\"background-color:rgb(248,248,248);border-bottom:2px solid rgb(238,238,238);\"\u003e\n\u003cth style=\"padding:12px;text-align:left;vertical-align:top;\"\u003eSpecification\u003c\/th\u003e\n\u003cth style=\"padding:12px;text-align:left;vertical-align:top;\"\u003eDetails\u003c\/th\u003e\n\u003c\/tr\u003e\n\u003c\/thead\u003e\n\u003ctbody\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eAlgorithm and cooling\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eSHA-256. Air-cooled, four fans. There is no hydro version of this model.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eMiner hashrate\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003e115 or 120 TH\/s as sold, both around 23 J\/TH. The 136 TH\/s figure that circulates is a high-power figure, not the rate a stock machine runs at.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eHashboards per miner\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003e3, at roughly 37 to 40 TH\/s each. On stock firmware one failed board does not cost a third of the rate — it costs all of it, because the machine stops.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eChips per board\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003e77, and 231 in the machine. A healthy board reports 77 in the log; any other number is a fault.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eHow the chips are grouped\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003e11 groups of seven, fed together. Your screen will not use that word: the miner calls one board a \u003ccode style=\"word-break:break-all;\"\u003echain\u003c\/code\u003e, calls each chip an \u003ccode style=\"word-break:break-all;\"\u003easic\u003c\/code\u003e, and reports a count.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eChip family\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003e\n\u003ccode style=\"word-break:break-all;\"\u003eBM1366\u003c\/code\u003e, marked \u003ccode style=\"word-break:break-all;\"\u003eBM1366BS\u003c\/code\u003e or \u003ccode style=\"word-break:break-all;\"\u003eBM1366BP\u003c\/code\u003e. Both appear on this board, nobody publishes what the letters mean, and the rule that matters is to match the marking already on your own board.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eHashboard part number\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003e\n\u003ccode style=\"word-break:break-all;\"\u003eBHB56902\u003c\/code\u003e is the common one; \u003ccode style=\"word-break:break-all;\"\u003eBHB56903\u003c\/code\u003e and \u003ccode style=\"word-break:break-all;\"\u003eBHB56907\u003c\/code\u003e also exist. A firmware build that does not know your board number says so in as many words: \u003ccode style=\"word-break:break-all;\"\u003eboard name doesn't match\u003c\/code\u003e.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eTwo board builds\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eOne carries five long aluminium heatsinks held down by screws — two on the front, three on the back. The other carries 154 small heatsinks soldered on, one over every chip on both faces. See Gotchas below: we repair the first and not the second.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eReleased\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eAnnounced April 2023, shipping from August 2023. No S19k Pro in the field is still inside a Bitmain warranty.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\n\u003ch3\u003eWhat the Log Is Telling You\u003c\/h3\u003e\n\u003cp\u003eNone of this needs action from you. It is here so you know what state your machine is in before deciding to ship it.\u003c\/p\u003e\n\u003ctable style=\"width:100%;border-collapse:collapse;table-layout:fixed;font-family:sans-serif;font-size:14px;margin:16px 0 24px;\"\u003e\n\u003ccolgroup\u003e\n\u003ccol style=\"width:34%\"\u003e\n\u003ccol style=\"width:33%\"\u003e\n\u003ccol style=\"width:33%\"\u003e\n\u003c\/colgroup\u003e\n\u003cthead\u003e\n\u003ctr style=\"background-color:rgb(248,248,248);border-bottom:2px solid rgb(238,238,238);\"\u003e\n\u003cth style=\"padding:12px;text-align:left;vertical-align:top;\"\u003eWhat you see\u003c\/th\u003e\n\u003cth style=\"padding:12px;text-align:left;vertical-align:top;\"\u003eWhat it usually means\u003c\/th\u003e\n\u003cth style=\"padding:12px;text-align:left;vertical-align:top;\"\u003eWhat we do with it\u003c\/th\u003e\n\u003c\/tr\u003e\n\u003c\/thead\u003e\n\u003ctbody\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003e\u003ccode style=\"word-break:break-all;\"\u003eChain N only find M asic, will power off hash board N\u003c\/code\u003e\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eThat board enumerated fewer than 77 chips three times running, and the firmware has stopped the machine over it.\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eWe take that board, find where the chain stops answering, and repair it at component level.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eA count of 76 rather than 6\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eOne chip has dropped out. It is the same class of fault as a count of 6, not a lesser one.\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eWe test the chips separately, because a chip that answers is not necessarily a chip that works.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003e\u003ccode style=\"word-break:break-all;\"\u003eEEPROM error: CRC 3RD REGION\u003c\/code\u003e\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eAbnormal, but it appears on healthy chains too and does not tell you which board failed.\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eWe read it as context and let the chip counts name the board.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003e\u003ccode style=\"word-break:break-all;\"\u003eERROR_UNBALANCE: Unbalance happened, waiting!\u003c\/code\u003e\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eThe machine hashed and then stopped itself. Often heat or power delivery rather than the pool or the firmware version it followed.\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eWe reproduce it under load before treating it as a board fault, and say so if it is not one.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eA fan error with all three chains at 77\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eCould be a fan, could be a board — both end with every board powered down.\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eWe separate the two on the bench rather than guessing from the dashboard.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003ctr style=\"border-bottom:1px solid rgb(238,238,238);\"\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003e\n\u003ccode style=\"word-break:break-all;\"\u003ebad chain id = 3\u003c\/code\u003e and other startup noise\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eNothing. These lines print on machines that hash perfectly.\u003c\/td\u003e\n\u003ctd style=\"padding:12px;vertical-align:top;word-break:break-word;\"\u003eNothing — we tell you so and save you the shipping.\u003c\/td\u003e\n\u003c\/tr\u003e\n\u003c\/tbody\u003e\n\u003c\/table\u003e\n\n\u003ch3\u003eDiagnostics Focus\u003c\/h3\u003e\n\u003cp\u003e\u003cstrong\u003eSeven chips to a group is a coarse split, and it changes what a missing chip means.\u003c\/strong\u003e On this board the chips sit in eleven groups of seven rather than in pairs, so a chip that stops answering may be the chip or may be the group it shares its supply with. Separating those two is the first job, and on a seven-chip group getting it wrong costs six working chips.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eNothing is judged cold.\u003c\/strong\u003e This model takes ten minutes or more to come up to temperature before it hashes at all, and a bench test cut short on it produces failures that are not there. Every board gets the full warm-up before any verdict is written down.\u003c\/p\u003e\n\n\u003ch2\u003eOur Professional Repair Process\u003c\/h2\u003e\n\n\u003ch3\u003eGotchas\u003c\/h3\u003e\n\u003cp\u003e\u003cstrong\u003eWe repair one of the two S19k Pro boards, and it is worth ten seconds to check which you have.\u003c\/strong\u003e Look at the board. Five long metal bars held down by screws, two on the front and three on the back, is the one we take. A small individual heatsink sitting on top of every single chip, soldered on with no screws anywhere, is the one we do not — reaching a chip on that board means desoldering metal off both faces first, and it is also the build Bitmain itself will not accept as a bare board. Check before you pay for shipping.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eChips are counted after the board is powered, not before.\u003c\/strong\u003e How many need replacing is not knowable from a log line, so the repair price covers the work and the chips are settled once the board is on the bench and the number is known. You hear the number before it is done, not after.\u003c\/p\u003e\n\n\u003ch3\u003eTypical Service Scenario\u003c\/h3\u003e\n\u003cp\u003e\u003cstrong\u003eA machine bought in 2023 or 2024, failing just outside its year.\u003c\/strong\u003e This model shipped from August 2023 with a one-year term, so the boards that fail now fail as the owner's problem rather than Bitmain's, and that is most of what arrives here.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eA miner in an enclosure or a warm corner.\u003c\/strong\u003e Unbalance shutdowns and temperature stops get blamed on the firmware update that started catching them. The update is rarely the cause; it is usually the first thing strict enough to notice.\u003c\/p\u003e\n\u003cp\u003e\u003cstrong\u003eA machine that was fine until somebody changed the firmware.\u003c\/strong\u003e Flashing between stock and third-party builds on a board number a build does not recognise leaves a miner that will not initialise at all, which looks exactly like dead hardware and is not.\u003c\/p\u003e\n\n\u003ch3\u003eRunning the Miner While the Board Is With Us\u003c\/h3\u003e\n\u003cp\u003eStock firmware will not run this machine a board short: it powers the whole miner down, so one failed board costs all of the rate rather than a third. On \u003ca href=\"https:\/\/antminer-repair.com\/pages\/firmware\"\u003ethe firmware that lets it keep hashing\u003c\/a\u003e, that board is switched off and the other two carry on, leaving two thirds of a three-board S19k Pro. The empty slot has to be blocked; otherwise the remaining boards run hotter than they did with a full set. A dead chip cannot be bypassed — only a whole board — and it still has to come in.\u003c\/p\u003e\n\n\u003ch3\u003eDiagnostics and Validation Equipment\u003c\/h3\u003e\n\u003cp\u003eBench measurement, STASIC and ASIC REPAIR fixtures for chip count and work verification, a per-chip test that scores all 77 separately, and final validation inside a working S19k Pro under real load for at least an hour. Extended burn-in beyond that hour is a separate add-on. The fixture proves the chain; the per-chip test finds the chip that answers but does not work; only the miner proves the board.\u003c\/p\u003e\n\n\u003cp\u003eAll repairs are performed personally by Alex, who has been servicing ASIC miners since 2019. \u003ca href=\"https:\/\/antminer-repair.com\/pages\/aleksandr-muranov\"\u003eSee credentials →\u003c\/a\u003e\u003c\/p\u003e\n\n\u003cp\u003e\u003ca href=\"https:\/\/antminer-repair.com\/pages\/contact\"\u003eContact our repair team\u003c\/a\u003e today and get your miner back to full power.\u003c\/p\u003e","brand":"BITMAIN","offers":[{"title":"Default Title","offer_id":48416875708689,"sku":"HB-S19KPRO","price":150.0,"currency_code":"USD","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/0788\/4917\/9921\/files\/S19KPronewtopangle.jpg?v=1712027316","url":"https:\/\/antminer-repair.com\/products\/antminer-s19k-pro-hashboard-repair-service","provider":"Antminer Repair","version":"1.0","type":"link"}