Quick answer

Used and factory-recertified Seagate Exos or WD Ultrastar SATA drives cut NAS storage cost per terabyte sharply when you verify SATA rather than SAS interfaces, handle SATA Pin 3 power-disable resets, compare SMART hours against Seagate FARM logs, run a full-surface burn-in before pool creation, and deploy RAIDZ2 with one cold spare.

Contents

At a glance

Class
Homelab

Provenance

  • Nov 14, 2026Published

Hyperscalers retire helium enterprise drives by the pallet while those platters still have years of mechanical life left. That surplus changed homelab storage math. Instead of stretching a budget across four small retail NAS drives, builders routinely fill wider arrays with 16TB to 24TB Seagate Exos, Western Digital Ultrastar (formerly HGST), and Toshiba MG drives at a steep discount—a spread you can watch move in our HDD price index.

Every used drive I put through intake starts on probation. Buying enterprise pulls works brilliantly when you treat each arriving disk as untrusted hardware until it survives burn-in. Skip three physical compatibility checks, though, and you end up staring at drives that either cannot plug into your motherboard or refuse to spin their motors at all.

Manufacturer-recertified vs seller-refurbished

Listings throw “Renewed,” “Refurbished,” and “Factory Recertified” around as if they describe one grade of hardware. They do not.

  • Manufacturer-recertified: Returned or decommissioned units processed on the drive maker’s own line. Seagate’s Factory Recertified page describes sanitizing data, running factory-level diagnostics, and re-labeling the drive (green-bordered or laser-marked recertified label) with standard SMART counters reset to zero. Read the fine print on warranty: unless you bought directly from the manufacturer’s own storefront, warranty replacement runs entirely through the reseller—ranging from 90 days on marketplace storefronts to two or three years at dedicated enterprise-drive specialists.
  • Seller-refurbished / datacenter pulls: Wiped and SMART-screened by a third-party IT asset disposition (ITAD) house or reseller. Honest pulls leave original SMART Power_On_Hours untouched (15,000–35,000 hours is normal 2–4 year fleet duty) and cost less per terabyte, usually carrying a 90-day to 2-year seller RMA policy.

Neither label substitutes for arrival testing. Neither replaces parity. For acoustic expectations before putting 7,200 RPM helium drives near a living room, read our Exos vs Ultrastar vs Toshiba MG noise comparison.

Trap 1: SAS vs SATA (read the exact model suffix!)

Every week someone orders six cheap 18TB Exos drives off eBay, opens the anti-static bags, and discovers none of their SATA cables fit. They bought 12Gb/s SAS drives for a motherboard with 6Gb/s SATA ports.

As Seagate’s SAS installation documentation states plainly, SAS drives will not operate on a SATA controller. Physically, an SFF-8482 SAS drive plug bridges the notch between the 15-pin power blade and 7-pin signal blade (carrying secondary port pins underneath) specifically to stop you plugging a SATA cable onto it. Electrically, a motherboard PCH SATA port cannot speak Serial Attached SCSI. The reverse works fine, since an LSI SAS HBA speaks SATA Tunneling Protocol (STP) and drives both SAS and SATA disks, so only buy SAS drives when your NAS uses a true SAS HBA and SFF-8482 SAS breakout cables or a SAS backplane.

Check the exact datasheet part number before clicking buy:

LineSATA 6Gb/s exampleSAS 12Gb/s exampleQuick tell
Seagate Exos X20 / X24ST20000NM007D / ST24000NM002HST20000NM002D / ST24000NM000HNever guess from ST...NM; verify SATA 6Gb/s on Seagate’s spec table
WD Ultrastar DC HC550 / HC580WUH721818ALE6L4WUH721818AL5204ALE6 = SATA 6Gb/s (512e); AL52 / AL42 = SAS 12Gb/s

Also verify sector formatting on used SAS pulls. Enterprise storage appliances like NetApp or EMC often format SAS drives with 520-byte or 528-byte sectors instead of 512e or 4Kn, requiring sg_format --format --size=512 (or 4096) before Linux or ZFS can partition them.

Trap 2: SATA Pin 3 (3.3V PWDIS) “dead on arrival” drives

You unwrap a SATA Ultrastar or Exos, plug it into a desktop ATX power supply, boot up, and feel zero spindle vibration. The drive feels dead. Hold off on the RMA. Check SATA Revision 3.3 Power Disable (PWDIS) first.

Western Digital’s Power Disable tech brief explains the mechanism clearly. Hyperscale enclosures repurpose Pin 3 of the 15-pin SATA power connector so a backplane can hard-reset a hung drive remotely by asserting +3.3V high. Legacy and many consumer ATX SATA power cables wire continuous +3.3V down that orange fifth wire, holding a PWDIS drive in perpetual reset so the motor never receives permission to spin.

Three clean fixes wake the drive immediately (without modifying drive firmware):

  1. Backplanes fed purely via 4-pin Molex: Classic 4-pin Molex carries only +12V (yellow), +5V (red), and ground—zero +3.3V exists on the cable, so PWDIS stays low and every drive spins normally.
  2. Quality crimped Molex-to-SATA power adapter: Use a reputable crimped (push-in IDC style, never cheap injection-molded PVC blocks) adapter cable that omits the +3.3V wire.
  3. Kapton polyimide tape over Pin 3: Slice a ~2 mm strip of thin electrical Kapton tape under a magnifier and mask Pin 3 (or Pins 1–3, all formerly +3.3V unused by 3.5-inch HDDs) on the drive’s power blade.

Inspecting SMART vs Seagate FARM logs on arrival

Before starting a multi-day surface test, inspect what the drive’s firmware claims. Cross-check it immediately. In early 2025, heise online exposed used Seagate Exos and IronWolf drives sold as brand-new whose standard SMART Power_On_Hours read near zero while Seagate’s internal Field Access Reliability Metrics (FARM) log held 10,000 to 50,000 spindle hours.

Install smartmontools 7.4 or newer (release 7.4 added -l farm) and capture both reports before writing a byte:

sudo smartctl -x /dev/sdX > arrival-smart-sdX.txt
sudo smartctl -l farm /dev/sdX > arrival-farm-sdX.txt
grep -E "Power_On_Hours|Power on Hours|Spindle Power on Hours|Head Flight Hours" arrival-*-sdX.txt

Read two caveats honestly. First, on legitimate Seagate Factory Recertified drives, SMART Power_On_Hours is reset during recertification while FARM frequently preserves prior fleet hours, so a mismatch on a clearly labeled recertified listing is expected whereas a mismatch on a drive sold as brand new proves tampering. Second, Tom’s Hardware subsequently reported that counterfeiters found tools to reset FARM logs too, and WD/Toshiba drives do not expose Seagate’s FARM page. You can paste your sanitized smartctl output into our open Drive Arrival Census to compare seller cohorts.

Check Reallocated_Sector_Ct (ID 5), Reported_Uncorrect (ID 187), Current_Pending_Sector (ID 197), and Offline_Uncorrectable (ID 198) using our mechanical hard drive health guide. On a recertified arrival, I require every single one of those four raw counters to sit at 0 before burn-in starts, and still sit at 0 when burn-in finishes.

The 4-step arrival burn-in protocol

Shipping shocks kill marginal heads inside the first 72 hours of full-stroke seeking. Force that failure early. Break weak drives inside your return window rather than six months later with family photos on the pool.

  1. Confirm the exact device node and serial number first (lsblk -d -o NAME,SIZE,MODEL,SERIAL)—badblocks -w overwrites every byte on the target device with zero undo!
  2. Short SMART self-test (~2 minutes): From smartctl(8), run sudo smartctl -t short /dev/sdX to catch immediate electronics or servo faults.
  3. Full-surface destructive write/read burn-in (badblocks inside tmux): As documented in badblocks(8), -wsv writes and verifies four bit patterns (0xaa, 0x55, 0xff, 0x00) across the entire platter surface from outer to inner diameter. Use -b 4096 (essential on >16TB drives so the 32-bit block counter does not overflow at 2^32 × 512B = 2 TiB!):
    # DESTRUCTIVE: triple-check /dev/disk/by-id/ serial before running!
    sudo badblocks -b 4096 -c 65536 -wsv /dev/disk/by-id/ata-ST20000NM007D_SERIAL
    
    (If 60–90 hours for four patterns is too long on a 24TB drive, a single full-zero write pass via dd or badblocks -b 4096 -c 65536 -wsv -t random followed by an extended SMART self-test still sweeps every physical sector twice.)
  4. Extended SMART self-test + post-burn diff: Run sudo smartctl -t long /dev/sdX (roughly 24–32 hours on 20–24TB), keep drive temps under 45°C with a desk fan if testing on an open bench, and re-run smartctl -x afterward. Any badblocks error or non-zero climb in ID 5, 187, 197, or 198 goes straight back for seller RMA.

Why RAIDZ2 + 1 cold spare beats buying fewer new NAS drives

Run the budget arithmetic in our NAS & ZFS storage calculator alongside our 6-bay NAS rebuild-time analysis and best NAS hard drives guide.

Resilvering a 20TB or 24TB enterprise drive reads tens of terabytes across every surviving member for a full day or longer. Single-parity RAIDZ1 leaves zero margin during that rebuild—one latent read error or a second used drive dropping out takes the volume down.

When factory-recertified enterprise drives cost 35%–50% less per terabyte than shrink-wrapped retail NAS boxes, spend part of that savings on architecture instead of shrink-wrap:

  • Build RAIDZ2 (two parity drives, surviving any two concurrent disk failures during a long 20TB+ resilver).
  • Buy one extra burned-in drive from the same order (N + 1), pass it through the exact same badblocks + smartctl -t long gauntlet while your return window is open, and store it powered-off in an anti-static bag on the shelf beside the server.

When a drive eventually throws pending sectors on a Sunday night three years from now, you slide a pre-tested cold spare into the bay in five minutes instead of waiting four days on shipping while a degraded array sweats. Keep an independent 3-2-1 off-box backup regardless, because RAID parity buys uptime rather than backup, and recertified enterprise drives deliver datacenter mechanics at homelab prices.

Frequently asked questions

What is the difference between manufacturer-recertified and seller-refurbished hard drives?
Manufacturer-recertified drives go back through the drive maker’s own factory line—Seagate documents its Factory Recertified process as data sanitization, factory diagnostics, firmware updates, and laser-etched Recertified labeling—though when bought through third-party resellers the warranty comes from the seller rather than Seagate’s retail 5-year policy. Seller-refurbished drives are wiped and graded by a liquidator or reseller, often retaining their original datacenter SMART Power-On Hours unless cleared.
Will a SAS enterprise hard drive work on motherboard SATA ports?
Never. Although a SAS SFF-8482 drive connector looks almost identical at a glance, a continuous plastic bridge joins the power and data blades so a standard SATA data cable cannot plug in, and motherboard SATA controllers cannot speak the Serial Attached SCSI protocol. A SAS drive requires a dedicated SAS Host Bus Adapter (such as an LSI 9300/9400/9500 in IT mode), whereas a SAS HBA can run both SAS and SATA drives.
Why won’t my refurbished WD Ultrastar or Exos drive spin up in a desktop PC?
Enterprise SATA 3.3 drives often implement the Power Disable (PWDIS) feature on SATA power Pin 3. Western Digital’s Power Disable tech brief explains that if an older or consumer power supply sends continuous +3.3V down Pin 3, the drive stays held in hard reset and never spins its motor. Feeding the drive through a good Molex-to-SATA adapter (which omits +3.3V) or masking Pin 3 with Kapton tape lets it spin immediately.
How do I check if SMART hours were wiped on a Seagate Exos drive?
Run smartctl -l farm /dev/sdX using smartmontools 7.4 or newer to read Seagate’s Field Access Reliability Metrics (FARM) log alongside smartctl -a. Factory recertification and third-party tools routinely zero standard SMART Attribute 9 (Power_On_Hours), while FARM often retains cumulative spindle and head-flight hours—though 2025 fraud investigations showed scammers can alter FARM too, so always burn-in every drive regardless of reported hours.
How long does badblocks burn-in take on a 20TB or 24TB hard drive?
Expect roughly 55 to 90+ hours. Running badblocks -b 4096 -wsv writes and reads four distinct bit patterns across every sector on the platters—eight full surface passes in total. At ~190–220 MB/s average across outer and inner tracks, a single 20TB pass takes ~26 hours, which is why running all drives in parallel inside tmux before building your ZFS pool saves weeks.

Sources and corrections

Last updated
Methodology
See our methodology for research and review standards. It draws on 8 cited sources, listed below, each checked against the original page on the date above. Interface, pinout, and diagnostic mechanics on this page come from Seagate’s Factory Recertified program documentation and SAS installation KB, Western Digital’s SATA 3.3 Power Disable Pin engineering brief, smartmontools 7.4+ (smartctl -l farm and smartctl(8)), badblocks(8) from e2fsprogs, and the January–March 2025 heise online / Tom’s Hardware investigations into reset Seagate SMART and FARM counters.
Corrections
Spotted an error? Email a correction with the affected line and what you found. The draft includes this page’s address. Verified changes are recorded here. Contact details.

About the author

Written by Lowell K. Wood IV, who builds and runs TechFuelHQ from St. Louis, Missouri.