Rig Sees 5 of 6 GPUs: A Checklist for Finding the Missing One
The card sits in the rig, its fans spin, the RGB glows — and the system reports five GPUs out of six. Bad news: something in the "card → riser → slot → driver" chain is broken. Good news: nine times out of ten it is not the card itself. Here is a methodical checklist that finds the culprit in one evening.
TL;DR (for those skipping the read)
- "Card not detected" almost never means "dead card". The usual suspects are the riser, its power feed, or slot contact.
- Work from simple to complex: power → riser → slot → driver → swap test. Do not disassemble half the rig before checking a cable.
- The key diagnostic move is swapping: exchange the suspect card with a working one. If the problem travels with the card — the card is guilty; if it stays on the slot — the wiring is.
- Change one element at a time. Otherwise you will not know what fixed it, and the problem will return.
- An intermittently vanishing card is worse than a dead one: it is usually a degrading contact that will become permanent over time.
Step 1: power to the card and riser
Start with what your hands can check without a single tool. On the missing card, verify every power connector: on the card itself and on the riser. A plug seated "almost all the way" looks connected but makes no contact — and the fans may still spin from slot power, creating the illusion of a live card.
Inspect the cable itself: kinks near the connector, scorch marks, melted plastic. Power lines feeding two cards, or a card plus a riser, are the prime suspects — we covered load distribution in the wiring article and the PSU budget article.
A quick test: move the missing card's power feed to a line known to work. If it comes alive, the line or cable was guilty.
Step 2: the riser — prime suspect
Risers are the weakest link of a modern rig (we have a dedicated article on them). The USB cable, the riser board, the board connector — every element dies more often than a GPU. Symptoms of a riser problem: the card vanishes out of nowhere, vanishes and reappears after reboots, and it is always the same card even though the card itself is fine.
Replacement order: the USB cable first (cheapest and most worn), then the whole riser board. Keep one or two known-good risers in stock — they are the most-consumed spare part on any farm.
While you are there, look at the motherboard slot itself: dust and oxidation on contacts are cured by blowing it out and reseating the riser a couple of times — replugging cleans the contact by itself.
Step 3: driver and operating system
If the hardware side is clean, look at software. The card may be visible to the system but fail driver initialization — then it shows up in the device list with an error while the miner ignores it. Reinstalling the driver with a full cleanup of the previous one usually helps.
Classic software causes of a missing GPU: an over-the-air driver update mid-operation, a conflict after adding a new card to the rig, a card-count limit in an older OS, or PCIe settings reset in BIOS (Above 4G mode, link speed). If the disappearance coincided with an update — roll back and check.
Step 4: the swap test as the verdict
When indirect checks are exhausted, move to swapping — it gives an unambiguous answer. Take the missing card together with its riser and swap places with a working card-plus-riser pair:
- The problem moved to the new slot with the card — the card or its riser is guilty. Split them: working card on this riser, this card on a working riser. Now you know exactly which of the two it is.
- The problem stayed on the old slot — the slot, cable, power line or motherboard is guilty. Test them one by one.
- Everything started working after the swap — it was a bad contact. Do not celebrate early: watch it for a couple of days, degrading contacts love to come back.
Write down which card sat where and what you changed. Without notes, day three of debugging becomes "did I already swap this or not?".
Where automation fits
A missing card is minus one sixth of the rig's income for as long as it takes you to notice. OCBot sees a card drop immediately and sends a notification instead of waiting for you to check the dashboard. For intermittent cards this is especially valuable: the drop log shows the card vanishing every night at three — which is nearly a diagnosis by itself (power or contact, not driver).
This article is general information summarizing common practice; actual causes depend on your hardware. OCBot is an independent project, not affiliated with NVIDIA or HiveOS/Hiveon. Use at your own risk: disclaimer.