Quick Summary (TL;DR):
For a four-drive TrueNAS or Unraid build, use the motherboard’s native SATA ports if they are sufficient, configured for AHCI/non-RAID, and each disk is visible individually with its identity and health data. A non-RAID PCIe SATA cardPriceAliExpress price is a reasonable way to add a few SATA HDDs when slots, power and cooling are limited; a documented six-port example is StarTech’s ASM1166-based 6P6G-PCIE-SATA-CARD. For six to eight disks, a proper SAS/SATA host bus adapter (HBA) is generally the stronger expansion choice, particularly if TrueNAS runs as a Proxmox VM and the entire controller can be passed through. Buy the exact controller, firmware, breakout cables and backplane combination—not just a card labelled “8 ports”. Do not put TrueNAS or Unraid data disks behind an opaque hardware-RAID volume, do not assume that “JBOD” means true disk passthrough, and do not use a SATA port multiplier as a substitute for independent drive connections. This is a specification- and documentation-based buying guide, not a hands-on test of the cards discussed.
Which Storage Controller Should You Choose?
| Your NASPriceAliExpress price build | Best starting point | Main reason | Important caveat |
|---|---|---|---|
| Up to four SATA HDDs; bare-metal NASAliExpress price | Motherboard SATA | No extra card, cables or PCIe slot | Check AHCI mode and M.2 port-sharing rules |
| Five or six SATA HDDs; one free PCIe slot | Quality non-RAID SATA card | Simple, individual SATA connections | Confirm chipset, negotiated PCIe bandwidth and OS support |
| Six to eight HDDs; long-term expandable NAS | SAS/SATA HBA | Server-oriented controller and SAS backplane options | Correct IT/HBA firmware, cables, airflow and slot width |
| TrueNAS as a Proxmox VM | Dedicated HBA passed through whole | Guest owns the controller and disks | Requires IOMMU isolation; host boot drive elsewhere |
| Existing SAS enterprise drives | SAS-capable HBA | SATA controllers cannot operate SAS disks | Match SAS backplane, connectors and power |
| Two to four SSDs doing sustained multi-GB/s I/O | Onboard ports or well-provisioned HBA | Avoid undersized shared PCIe links | Compare total PCIe bandwidth, not port count |
Our default recommendation is deliberately boring: do not add a controller you do not need. Every additional PCIe device introduces another driver, cable path, firmware version, heat source and potential failure point. If a good motherboard already exposes six usable SATA ports, the first question is why you would buy a card at all.
Why TrueNAS and Unraid Need Individual Disk Access
Both operating systems manage data protection above the physical drive layer. TrueNAS uses OpenZFS for pools, vdevs, checksums and recovery. Unraid uses its own array/parity model and can also use supported pools. In both cases, the operating system should identify the actual physical drives, read useful health information and control disk operations without a RAID controller silently remapping devices.
A suitable controller exposes disks independently. You should be able to see a model and a persistent serial identifier for each drive, query S.M.A.R.T. or equivalent health information where supported, and identify which physical bay needs attention. A RAID card that presents eight disks as one logical volume prevents the NAS software from making informed per-disk decisions. A RAID controller offering “JBOD” may or may not provide the same behaviour as a straightforward HBA; verify the specific controller and firmware.
The TrueNAS hardware guide recommends Broadcom/LSI SAS HBAs and explains why hardware RAID can obscure serial numbers, S.M.A.R.T. information and write-cache behaviour. Unraid’s installation checklist calls for AHCI on SATA and HBA rather than RAID mode. These are storage-software requirements, not marketing preferences.
Motherboard SATA Ports: Usually the Best First Choice
A modern desktop or NAS motherboardPrice often supplies four or six SATA sockets; specialist boards can provide more. Native chipset SATA is normally the simplest route for a small bare-metal NAS. There is no separate HBA to cool, no SAS breakout cable to select, and no additional card drawing power in a PCIe slot. It is also easier to trace a single failed cable or disk.
The catch is that a motherboard’s printed SATA count does not always equal the number available in your final build. On some boards, populating a particular M.2 slot disables one or two SATA ports because the lanes or chipset resources are shared. Some boards also combine chipset-native SATA with ports supplied by a secondary onboard controller. Read the exact board manual and identify which sockets belong to which controller.
- Set the SATA controller to AHCI or an explicitly documented non-RAID mode before creating the storage pool.
- Confirm every intended port remains available with your planned M.2 SSDs installed.
- Check whether the BIOS exposes a separate RAID/RST/VMD mode that changes how disks are presented.
- Use short, well-seated SATA data leads and independent, appropriately rated power connections.
- Leave at least one accessible replacement route; a cramped four-bay build can be harder to service than its specification suggests.
For choosing the underlying platform rather than the controller, see our DIY NAS versus prebuilt NAS cost comparison. A motherboard with the right ports can be more economical than buying a cheaper board and adding an HBA afterwards.
PCIe SATA Cards: Sensible Expansion, with Limits
A non-RAID PCIe SATA card adds ordinary SATA ports without requiring SAS breakout wiring. For a compact server with five or six SATA HDDs, this can be an elegant upgrade. The important words are non-RAID, documented chipset, adequate PCIe bandwidth and reliable driver support. Port count alone is not a meaningful quality metric.
One concrete, documented example is the StarTech 6P6G-PCIE-SATA-CARD. StarTech specifies six SATA 6Gb/s connectors, an ASMedia ASM1166 controller, a PCIe 3.0 x4 physical interface, non-RAID operation and up to 16Gb/s combined throughput. SATA data cables are sold separately. The x4 physical connector must not be mistaken for a guarantee that all six ports can deliver their individual 6Gb/s headline speeds simultaneously.
That aggregate bandwidth is adequate for many HDD arrays and typical 1GbE/2.5GbE NAS workloads. It is less compelling for a set of SATA SSDs performing simultaneous large transfers. Also, an “ASM1166” chip name printed on a low-cost listing does not prove the seller implemented the same power circuitry, firmware or connector arrangement as a reputable branded card.
What to avoid on cheap SATA expansion cards
- Listings that promise 10–16 independent SATA 6Gb/s ports from an unexplained tiny PCIe x1 connection.
- Undocumented port multipliers that make several drives share one SATA link.
- Hardware RAID cards where individual physical disks cannot be presented transparently.
- Cards whose actual chipset, firmware and Linux support cannot be confirmed.
- Unbranded boards with no warranty route, poor connector support or reports of disappearing disks under sustained load.
A small, low-power SATA expansion card can be the right purchase. It simply should not be confused with a dedicated enterprise HBA, and a card marketed as “RAID compatible” is not proof that it is suitable for a ZFS data pool.
SAS/SATA HBAs: The Better Expansion Platform for Many 8-Drive Builds
An HBA is a storage host controller that connects disks to the operating system without requiring the controller to manage a hardware RAID volume. Common Broadcom/LSI families provide internal SAS ports, which can also connect compatible SATA disks using the correct cables or backplane. This is why an eight-drive NAS often uses an HBA even when all eight drives are SATA HDDs.
The appeal is not that a SAS HBA magically makes a SATA HDD faster. It is the combination of well-established storage drivers, individual disk visibility, scalable cabling, and support for SAS backplanes and expanders when the build grows. It is particularly useful when a hypervisor must pass one entire PCIe controller to a storage VM.
An HBA is not automatically the best answer for a fanless N100 box. Enterprise cards may require deliberate airflow and a physically larger PCIe slot. Older used cards also bring provenance and firmware questions. The correct HBA is the one whose chipset, firmware, connector type, cooling and operating-system support are all known.
Concrete Controllers Worth Comparing
| Controller / option | Disk connections | Host connection | What to verify | Most sensible role |
|---|---|---|---|---|
| Motherboard chipset SATA | Board-specific, commonly 4–6 SATA sockets | Integrated chipset | M.2 sharing; AHCI; IOMMU grouping | Small bare-metal NAS |
| StarTech 6P6G-PCIE-SATA-CARD (ASM1166) | 6 × SATA 6Gb/s | PCIe 3.0 x4 physical; up to 16Gb/s combined (manufacturer) | OS driver; port visibility; bracket clearance | Simple 5–6 SATA-drive expansion |
| Broadcom/LSI 9207-8i (legacy/used) | 8 internal SAS/SATA lanes; 2 × SFF-8087 | PCIe 3.0 x8 | Genuine card, firmware, driver, cable type | Budget 6Gb/s SAS/SATA HBA where supported |
| Broadcom/LSI 9300-8i (legacy/used) | 8 internal SAS/SATA lanes; 2 × SFF-8643 | PCIe 3.0 x8 | IT/HBA firmware, airflow, used-card history | Established 8-drive NAS HBA |
| Broadcom HBA 9500-8i | 8 internal ports; 1 × x8 SFF-8654 SlimSAS | PCIe 4.0 x8 | Correct SlimSAS cable/backplane; price; host compatibility | Modern higher-end SAS/SATA expansion |
The 9207-8i and 9300-8i are older generations often encountered on the second-hand market; their presence in a comparison is not a claim that new retail stock or manufacturer support is guaranteed in 2026. The modern Broadcom HBA 9500-8i product page documents eight internal ports, PCIe 4.0 x8, one x8 SFF-8654 connector and 12Gb/s SAS generation. Its tri-mode capability should not be taken as a promise that an arbitrary SATA backplane can accept NVMe drives.
For a straightforward eight-SATA-HDD home NAS, the practical benefit of a 9500-class card over a suitable older HBA may be limited. Spend money on verified compatibility, cooling and a separate backup before paying for enterprise bandwidth you cannot use.
PCIe Slot Width and Lane Bandwidth Matter More Than Port Labels
The card’s PCIe connector shape, the number of electrical lanes actually wired on the motherboard, the negotiated PCIe generation and the controller’s own bandwidth limit are different things. A card may physically fit an x16-length slot while that slot is electrically x4. An x8 HBA may negotiate fewer lanes if the board does not supply them. Read both manuals and inspect the negotiated link after installation.
| Link example | Approximate one-direction raw payload ceiling | Practical implication |
|---|---|---|
| PCIe 2.0 ×1 | ~0.50 GB/s | A poor match for several fast drives |
| PCIe 3.0 ×1 | ~0.985 GB/s | Can bottleneck multiple SATA SSDs |
| PCIe 3.0 ×2 | ~1.97 GB/s | Often enough for several HDDs; shared by all attached disks |
| PCIe 3.0 ×4 | ~3.94 GB/s | Substantial headroom for an HDD array |
| PCIe 3.0 ×8 | ~7.88 GB/s | More than enough host bandwidth for eight ordinary SATA HDDs |
These are theoretical PCIe payload figures, not measured disk benchmarks, and actual throughput is lower after controller and protocol overhead. For a simple upper-bound illustration, six SATA SSDs each sustaining 550MB/s would request about 3.3GB/s in aggregate. A card advertised at 16Gb/s combined (roughly 2GB/s before overhead) cannot sustain that total. Six HDDs each transferring 200MB/s would request 1.2GB/s, which is a different proposition.
For 1GbE and 2.5GbE file serving, the network is often the bottleneck. For local SSD pools, fast scrubs, parity checks or 10GbE transfers, controller bandwidth and lane allocation deserve closer attention. See HDD vs SATA SSD vs NVMe for NAS for the storage-media side of that decision.
SAS Connector Names: SFF-8087, SFF-8643 and SFF-8654
A common buying mistake is to order an eight-port HBA and then discover that the case backplane or breakout cables use a different connector. The numbers below refer to physical connector families, not to drive count printed on the box. Each cable must be specified for the controller generation, the destination and the direction of the connection.
| Connector | Where you may encounter it | Common use | Mistake to avoid |
|---|---|---|---|
| SFF-8087 | Older internal 6Gb/s SAS HBAs, including 9207-8i | Mini-SAS to compatible backplane or forward breakout to SATA | Assuming it fits an SFF-8643 socket |
| SFF-8643 | Many internal 12Gb/s SAS HBAs, including 9300-8i | Mini-SAS HD to matching backplane or forward breakout | Ordering an SFF-8087 lead |
| SFF-8654 | Newer SlimSAS designs, including HBA 9500-8i | x8 SlimSAS to a supported backplane/cable assembly | Assuming any generic 8-port SAS cable works |
| 7-pin SATA | Motherboard and ordinary PCIe SATA cards | Individual SATA data cable per drive | Forgetting separate SATA power |
For direct-attached SATA disks, a forward breakout cable carries signals from an HBA’s SAS host connector to individual SATA drives. A reverse breakout cable is wired for the opposite architecture, commonly SATA motherboard ports into a suitable backplane. They may look almost identical and are not interchangeable. Confirm “forward” explicitly before ordering.
An HBA data connection does not power a 3.5-inch drive. Your PSU, SATA power harness and backplane must provide the required voltages and enough start-up current. Do not treat the data cable purchase as the entire electrical design.
SAS and SATA Compatibility Is Not Symmetrical
A compatible SAS HBA can generally address SATA drives through its SATA tunnelling capability, provided the cable/backplane is appropriate. The reverse does not hold: a normal motherboard SATA socket or SATA-only PCIe card cannot operate a SAS drive. That matters when buying inexpensive recertified enterprise disks. A cheap SAS HDD is not a bargain if the rest of your build is SATA-only.
SAS also introduces concepts such as expanders and dual-port enterprise drives. Those are useful in larger storage systems, but an eight-bay home NAS does not automatically benefit from them. A simple direct-attached cable path is easier to debug than a backplane with an undocumented expander.
Backplanes, SAS Expanders and SATA Port Multipliers
A passive/direct-attach backplane mainly provides sockets and routing between each drive and its own controller lane. A SAS expander backplane can connect more drive bays than the number of physical SAS lanes on the HBA, sharing upstream bandwidth. Both can be appropriate when correctly supported. A SATA port multiplier lets multiple SATA disks share a single SATA port, which is a less attractive basis for important NAS storage because controller support, shared bandwidth and fault isolation can be problematic.
Do not infer the backplane type from the number of drive bays. A case advertised as “8-bay hot swap” may have eight individual SATA sockets, two four-lane mini-SAS connectors, an expander or proprietary wiring. The chassis manual—not the front photograph—determines what to buy. Check the exact PCB/backplane revision, because vendors sometimes change it between case revisions.
If you are planning a compact NAS, the enclosure and cooling design are part of the storage-controller choice. A tight chassis may fit an HBA physically but provide no directed airflow across its heatsink. Conversely, a case with a suitable SAS backplane can make HBA wiring cleaner than eight loose SATA cables.
IT Mode, HBA Mode, IR Firmware and Hardware RAID
Older Broadcom/LSI controllers are commonly discussed in terms of IT (initiator/target) firmware versus IR (integrated RAID) firmware. For a TrueNAS data pool, the aim is direct access to each drive, not a controller-managed RAID array. Some controllers are sold as genuine HBAs; others are repurposed RAID cards that require a compatible non-RAID firmware configuration. These are not interchangeable product families simply because a seller writes “LSI IT mode” in the listing.
Before buying a used card, ask for its exact model, chip family, PCI vendor/device ID, firmware version, reported operating mode and a photograph of both sides. Prefer an authenticated part with documented firmware support. Avoid blind cross-flashing instructions copied from another board: incorrect firmware, board identity or boot ROM can make the card unusable. The official TrueNAS guide discusses the sas2flash, sas3flash and storcli utilities, but flashing is a separate maintenance procedure that should follow the vendor’s instructions for the specific device.
A useful test is whether the OS reports eight distinct physical disk identities and health data. A misleading “JBOD” option that creates eight virtual RAID0 devices is not the same as a clean HBA path, even if the file manager displays eight disks.
Bare-Metal TrueNAS: A Straightforward Controller Checklist
For a dedicated TrueNAS machine, onboard AHCI SATA or a supported HBA is usually simpler than mixing several cheap controllers. Keep the boot device separate from the data-pool disks. Before creating a pool, verify all intended disks are detected by model and serial, and confirm that health/temperature information is available. Check the controller driver, firmware and error logs before trusting the system with important data.
- Install the controller with the machine powered down; use the correct bracket and a PCIe slot with the required lanes.
- Connect one test drive first, verify it appears individually, then attach the rest.
- Record each disk serial against the physical bay before pool creation.
- Confirm AHCI or non-RAID/HBA firmware; do not initialise an array in a RAID BIOS.
- Inspect system logs for link resets, timeouts and repeated drive discovery before adding production data.
- Create a pool only after confirming the right disks; the pool-creation step is destructive to selected drives.
The current TrueNAS 25.10 hardware guide is a good baseline for supported storage design. It also advises against USB-connected disks as primary pool devices. USB is useful for certain backup workflows; it is not a substitute for stable, directly accessible data disks.
TrueNAS on Proxmox: Pass Through the Whole Controller
This is the scenario where a dedicated HBA most clearly earns its cost. A TrueNAS VM should own its physical storage controller through PCIe passthrough, rather than seeing a collection of virtual disks created on the Proxmox host. That avoids adding a second storage-management layer between ZFS and the disks. The TrueNAS virtualisation guidance describes full controller passthrough as the recommended architecture for important storage.
The CPU, motherboard and firmware must support IOMMU (Intel VT-d or AMD-Vi). The HBA must be in an isolatable IOMMU group and must not also contain the Proxmox boot drive or disks required by other host services. Once passed through, the HBA and its drives belong to the VM; the host should not simultaneously manage them.
# Read-only checks on a Linux/Proxmox host
lspci -nnk | grep -A3 -Ei 'SATA|SAS|Serial Attached|RAID'
find /sys/kernel/iommu_groups/ -type l | sort
lsblk -o NAME,MODEL,SERIAL,SIZE,TRAN
# Inspect one selected controller by its real PCI address:
lspci -vv -s 03:00.0 | grep -E 'LnkCap:|LnkSta:'
# Replace 03:00.0 with the actual PCI address.
These commands are read-only examples; output depends on hardware and driver. The LnkSta field can show negotiated link speed and width. Do not copy a stranger’s PCI address or vendor/device ID into a passthrough configuration. Do not blacklist a storage driver globally if the Proxmox host uses another controller driven by the same module. Review your Proxmox version’s passthrough documentation and verify the boot path before changing kernel parameters or VFIO binding.
A motherboard SATA controller can sometimes be passed through instead, but it may share an IOMMU group with chipset functions or include the hypervisor’s own boot disk. A separate HBA is often easier to isolate and maintain. PCI passthrough also limits certain VM operations such as live migration; it is a conscious architectural trade-off, not a free performance toggle.
Unraid: Direct Disks, Parity Checks and Replaceability
Unraid’s array identifies and manages individual physical disks. A supported non-RAID HBA makes it easier to grow an array while retaining visibility of each drive. As with TrueNAS, the relevant test is not whether a card says “RAID support” but whether Unraid can see stable drive identities, query health and reliably perform parity operations.
The official Unraid FAQ recommends non-RAID HBAs and cautions against RAID controllers that obscure drive health. Its setup guide specifies AHCI and HBA mode. For an Unraid machine that also runs Docker and VMs, consider the PCIe slot budget: a storage HBA, a 10GbE NIC and a GPU can compete for the same limited lanes on a small motherboard.
When migrating disks between ports or controllers, record the current disk assignments and serials, maintain a current backup and follow the Unraid migration procedure for the version in use. Do not casually assign a data disk to a parity slot, format a disk because it appears unmountable, or assume Linux /dev/sdX letters will remain fixed after a controller change.
SMART, Serial Numbers and Error Reporting: What to Check
Drive identity and health telemetry are essential for safe replacement. Before using a new controller with real data, inspect a test drive with tools appropriate to the operating system. On Linux, the following commands are examples for discovery and health inspection; they do not write to the disk.
# List disks using persistent identity where available
ls -l /dev/disk/by-id/
lsblk -o NAME,MODEL,SERIAL,SIZE,TRAN
# Requires smartmontools; replace the path with a real disk
sudo smartctl -a /dev/disk/by-id/ata-EXAMPLE_DISK_SERIAL
# Kernel storage errors since boot
journalctl -k --no-pager | grep -Ei 'ata[0-9]|mpt3sas|sas|reset|I/O error|timeout'
For SAS devices or unusual bridges, smartctl may need a device-type option. Some counters and temperature fields differ between SAS and SATA; a missing field is not automatically a failed disk. What matters is that the controller is stable, disks remain identifiable, and errors are investigated rather than suppressed.
Use the NAS operating system’s own disk-health and alert features as the primary operational interface. Command-line checks are useful for diagnosis, not a replacement for configured notifications and tested recovery procedures.
Cooling: The Hidden Cost of a Server HBA
Enterprise HBAs were designed for systems with directed chassis airflow. A heatsink that remains acceptable in a rack server may run much hotter in a silent Mini-ITX case. Some older controllers also consume more power than a compact SATA expansion card, though consumption varies significantly by generation and model. Do not rely on a generic “HBAs run hot” claim: look up the exact model’s thermal and airflow requirements.
The Broadcom 9500-8i specification lists 5.96W typical power for that particular card. That is not a measurement of your entire NAS and does not describe a 9300-8i or a second-hand cross-flashed adapter. A cooler-running modern HBA may still be financially unattractive if its acquisition price is high relative to the storage workload.
- Check the HBA heatsink position against the chassis airflow path, not merely whether a fan exists somewhere in the case.
- Ensure adjacent GPU/NIC cards do not block the airflow or physically foul SAS connectors.
- Check operating temperature limits and any vendor-specified airflow requirement.
- Avoid improvised exposed fan wiring or adhesive mounting near spinning fans and hot components.
- Retest disk and controller stability after closing the chassis, not only on an open bench.
Power, Spin-Up Current and Cable Quality
The HBA or SATA card handles data; it does not remove the power requirement of the disks. Several 3.5-inch HDDs starting together can draw substantially more power than they do when idle. The exact start-up current must come from each drive’s datasheet, and the PSU must also support the motherboard, fans and other peripherals. A PSU that looks generous on average watts can still be a poor match for start-up transients or badly distributed SATA power connectors.
Use reputable SATA power harnesses and avoid overloading one chain with many drives. Check the drive backplane’s power inputs and connector ratings. Some enterprise drives implement the SATA power-disable (PWDIS) feature, which can complicate use with certain legacy power connections; diagnose that from the drive and PSU documentation rather than modifying connectors blindly.
Loose SATA data connectors, sharply bent breakout cables and strained mini-SAS sockets can cause intermittent resets that resemble failing disks. Route cables so the side panel does not crush them, and label each data path against the drive bay. Cable quality is a small fraction of a NAS budget but a large part of its serviceability.
What Does a Complete 4-, 6- or 8-Drive Build Need?
| Drive count | Likely controller route | Data cables | Other items often forgotten |
|---|---|---|---|
| 4 × SATA HDD | 4 native AHCI ports | 4 SATA data leads | Power split across suitable PSU/backplane connections |
| 6 × SATA HDD | 6 native ports or a documented SATA expansion card | 6 SATA leads, or existing 4 plus 2 on expansion | PCIe lane allocation, bracket and spare PSU power |
| 8 × SATA HDD | 8 native ports or an 8-port SAS HBA | Typically 2 suitable four-lane forward breakout cables, or backplane-specific harness | HBA cooling, PCIe x8-class slot, SAS/SATA backplane compatibility |
| 8 × SAS HDD | SAS HBA and SAS-compatible backplane/cables | SAS-specific cabling | SATA-only motherboard/card is not sufficient |
The third column describes typical architectures, not universal cable shopping lists. An eight-bay case with an integrated backplane may require fewer or different cables. A SlimSAS x8 HBA may use a different harness from a dual-SFF-8643 HBA. Always start with the controller and backplane manuals, then choose the cable.
Five-Year Ownership Cost: The Card Is Only One Line
For a buying decision, compare the complete incremental cost of achieving the same usable NAS, not the price of a bare card. A sensible worksheet includes the controller, cables, any required fans or brackets, a PSU upgrade if needed, warranty/replacement risk, and the electricity attributable to the additional hardware. Then include the identical disks, RAM, case, boot SSD, networking, UPS and independent backup on both sides when comparing whole NAS systems.
Five-year incremental controller cost =
controller purchase price
+ correct breakout / SATA cables
+ bracket, cooling and any PSU upgrade
+ expected replacements / warranty costs
+ (average extra watts / 1000 × 24 × 365 × 5 × electricity rate)
Illustration only: 8W continuous difference at £0.25/kWh
= 0.008 × 24 × 365 × 5 × £0.25
= £87.60 over five years.
That 8W is a hypothetical difference, not a benchmark for any named card; £0.25/kWh is an illustrative unit rate, not a current tariff quotation. Measure the assembled system at the wall under representative workloads. If one option saves £40 on the purchase but uses an extra 8W continuously, the five-year electricity difference alone exceeds the saving under those assumptions. Conversely, a reused HBA that is already owned may have almost no acquisition cost.
| Cost item | Motherboard SATA | PCIe SATA card | SAS HBA |
|---|---|---|---|
| Controller acquisition | Included with motherboard | Additional card cost | New or verified used HBA cost |
| Data wiring | Individual SATA leads | Individual SATA leads | Correct forward breakout or backplane harness |
| Cooling | Case airflow | Check card requirements | Often needs deliberate heatsink airflow |
| PCIe slot and lanes | Usually none consumed by add-in card | Slot and negotiated bandwidth required | Often x8-class host slot and lane allocation |
| Power and electricity | Baseline | Measure added watts | Measure added watts; generation matters |
| Warranty and support | Motherboard vendor | Card vendor/seller | New-card vendor or used-card seller |
| Upgrade potential | Limited to board ports | Card/slot dependent | SAS backplanes and supported expanders possible |
Do not let the controller price distract from the larger purchase: a six- or eight-drive NAS also needs disks, adequate power, a suitable case, cooling, boot media and backup capacity. Our complete DIY NAS ownership-cost guide explains why apparently cheap platforms can become expensive once those parts are included.
A Safe Pre-Purchase Compatibility Checklist
- Exact board: identify motherboard model/revision, usable PCIe slot length and electrical lane count.
- Firmware: confirm BIOS options for AHCI, IOMMU, boot devices and slot sharing.
- Controller: obtain the precise model, chipset, firmware mode and Linux/TrueNAS/Unraid support information.
- Disks: count SATA versus SAS, HDD versus SSD and any 4Kn/sector-format constraints.
- Cables: match SFF-8087, SFF-8643, SFF-8654 or SATA; verify forward versus reverse breakout.
- Backplane: distinguish direct-attach SATA, SAS expander and proprietary wiring.
- Power: check PSU capacity, drive spin-up load and the number of properly rated power connectors.
- Cooling: check heatsink airflow with the chassis closed and other expansion cards fitted.
- VM ownership: for Proxmox, confirm IOMMU grouping and that host boot/storage are not on the passed-through controller.
- Operations: confirm disk serial visibility, health reporting, backup and restore procedure before importing valuable data.
Installation and Validation Without Risking Existing Data
Before installing a controller into a NAS with existing data, back up the important datasets and record drive serials, pool layout and controller settings. Power down fully, fit the card, route cables without strain and connect only a non-critical test disk for the first boot. If you are migrating an existing pool, follow the OS vendor’s documented procedure; do not initialise or format a drive simply because the new controller changes its Linux device name.
- Boot to firmware and confirm the new controller is enumerated without switching the system into RAID mode.
- Check the PCIe link speed and lane width and compare them with the slot/card specifications.
- Verify the test disk’s model, serial, capacity and S.M.A.R.T. data in the target operating system.
- Perform a non-destructive read test or vendor diagnostic on the test disk, then inspect kernel logs for timeouts and resets.
- Add the remaining disks, recording the cable-to-bay mapping and checking all drive identities again.
- Configure disk-health notifications, scheduled checks and an independent backup before placing the NAS into normal service.
For an existing TrueNAS ZFS pool, disk labels and metadata allow import when the hardware is correctly presented, but pool import is not a substitute for a backup. For Unraid, preserve array assignments and parity configuration. The correct process differs between systems and versions, so use the relevant official migration documentation.
Troubleshooting: Drives Missing, Resetting or Slow
| Symptom | Likely checks | What not to do |
|---|---|---|
| Some SATA ports vanish after adding NVMe | Motherboard manual: shared M.2/SATA resources; BIOS configuration | Assume the disks have failed |
| HBA shows no drives | Correct breakout direction; backplane type; drive power; firmware mode | Flash random firmware immediately |
| Disk serials replaced by virtual volume names | Controller RAID/JBOD mode; driver presentation | Create ZFS pool on an opaque RAID volume |
| One drive drops during parity check | SATA/SAS cable seating; PSU transient load; drive health; controller temperature | Ignore repeated kernel I/O errors |
| SATA card slow with multiple SSDs | Aggregate controller bandwidth and negotiated PCIe lanes | Expect 6Gb/s × every port concurrently |
| TrueNAS VM cannot claim HBA | IOMMU enabled, group isolation, host disk ownership, PCIe binding | Pass through host boot controller |
| SAS disk invisible on SATA card | SAS requires SAS-capable HBA/backplane | Buy another SATA data cable |
A storage controller should fail predictably and report useful errors. Repeated link resets or disappearing drives during long reads deserve investigation before you build a large parity or ZFS array. If several disks fail simultaneously, suspect shared power, cables, controller cooling or backplane issues rather than assuming multiple independent HDD failures.
Common Questions
Is an LSI/Broadcom HBA always better than motherboard SATA?
No. For a small bare-metal NAS, native AHCI SATA ports are usually simpler and cost less. A good HBA becomes attractive when you need more ports, SAS disks, a supported SAS backplane or clean PCIe controller passthrough to a VM.
Can I use a PCIe SATA card with TrueNAS?
Yes, provided the card exposes independent disks reliably, has a suitable driver, sufficient bandwidth and stable firmware. A documented non-RAID card is preferable to an opaque bargain controller. For important storage, validate disk identity, S.M.A.R.T. and sustained-operation stability before creating a pool.
Do I need IT mode for Unraid?
For older LSI controllers that offer IT and IR firmware, IT/non-RAID operation is normally the desired mode. Some modern HBAs are natively non-RAID and do not need a user-performed cross-flash. Follow the exact controller documentation rather than applying generic flashing instructions.
Will an 8-port HBA speed up eight SATA hard drives?
Not automatically. HDD mechanical performance is unchanged. A well-provisioned HBA can avoid a narrow shared host link and improve connectivity and manageability, but 1GbE/2.5GbE file transfers may already be network-limited. SSD arrays and fast local operations are more likely to expose controller bandwidth constraints.
Can a SAS HBA connect to SATA disks?
Usually yes with compatible HBA firmware, cables and backplane. A SATA-only motherboard port cannot run a SAS disk. Check the precise controller and backplane specifications; connectors that physically resemble each other are not necessarily wired for the same function.
Can I pass through individual disks to TrueNAS on Proxmox?
Some configurations expose raw disks through the hypervisor, but for important TrueNAS storage the documented recommendation is to pass through a supported entire storage controller. That gives TrueNAS direct ownership of the disks and avoids a virtual disk layer. Confirm IOMMU grouping and host boot separation first.
Do I need a SAS expander for eight drives?
Normally not with a direct eight-lane HBA and eight attached disks. An expander is useful when connecting more drive bays than the HBA’s direct lanes, but it adds cabling, bandwidth sharing and compatibility considerations. A simple eight-bay build is often better without one.
Does an HBA protect data like RAID?
No. An HBA provides connectivity and disk access. ZFS, Unraid parity or another storage layer provides selected redundancy, and independent backups provide recoverability. See our RAID 1 vs RAID 5 vs RAID 6 guide for the distinction.
The Verdict: Buy for Disk Visibility, Not Port Count
Choose motherboard SATA for a four-drive or modest six-drive bare-metal NAS when the ports are genuinely available and the firmware is configured correctly. Choose a documented non-RAID PCIe SATA card for a simple SATA-only expansion where the slot and aggregate bandwidth are sufficient. Choose a proper SAS/SATA HBA for an expandable six- to eight-drive system, SAS drives or a TrueNAS VM requiring full PCIe passthrough. Do not buy a hardware RAID controller to hide disks from ZFS or Unraid.
The most expensive controller is not necessarily the best. For a home server, predictable drive identity, stable error handling, correct cabling, sufficient airflow and a working backup plan matter more than an impressive SAS generation printed on the box.
Related NAS and Storage Guides
- DIY NAS vs Prebuilt NAS: Complete Build and Running Costs — compare the cost of the complete system, not just the controller.
- HDD vs SATA SSD vs NVMe for NAS — match storage-media throughput to controller bandwidth.
- RAID 1 vs RAID 5 vs RAID 6 — decide redundancy separately from disk connectivity.
- CMR vs SMR for NAS — select appropriate recording technology for a parity/ZFS array.
- ECC vs Non-ECC RAM for NAS and Home Servers — understand the wider platform reliability choices.
Official Documentation and Manufacturer Sources
- TrueNAS 25.10 Hardware Guide — current hardware requirements, controllers, disks and backplanes.
- TrueNAS Storage Controller Guidance — Broadcom/LSI HBAs, IT firmware and hardware RAID caveats.
- TrueNAS: Yes, You Can (Still) Virtualize TrueNAS — why full PCI passthrough matters for important storage.
- Unraid Deployment and BIOS Checklist — AHCI/HBA mode and IOMMU settings.
- Unraid FAQ: Recommended Controllers — preference for non-RAID HBAs.
- StarTech 6P6G-PCIE-SATA-CARD — ASM1166, six SATA ports, aggregate bandwidth and supported platforms.
- Broadcom HBA 9500-8i — official host interface, connector and typical power specifications.
- Broadcom Host Bus Adapter Family — current HBA product families and connectivity options.