Proxmox HBA buying guide
Best HBA for Proxmox: SAS2, SAS3, ZFS and Expansion
An HBA can turn one PCIe slot into eight, sixteen or more direct-attached drive lanes, but model numbers that look similar can hide different SAS generations, connector families and firmware behavior. For Proxmox, the key question is not which card has the biggest model number. It is whether the controller exposes disks cleanly to the host or storage VM, fits the available PCIe slot, uses the right breakout cables, stays cool, and leaves enough ports for the next round of expansion.
Quick answer
Which HBA class should most Proxmox hosts start with?
For an HDD-heavy home server, a genuine LSI/Broadcom HBA that exposes disks directly is usually easier to reason about than a hardware RAID card. The 9207-8i remains useful when eight 6Gb/s SAS/SATA lanes are enough, while the 9300-8i class moves to 12Gb/s SAS and SFF-8643 cabling. Higher-port 9305 cards reduce slot pressure. Tri-mode cards can make sense for newer mixed-media builds, but they add cost and cabling complexity that many SATA-only systems do not need.
Live Amazon hardware
Current controllers, cables and storage devices for this decision
These listings come from Sprint 4B's dedicated catalogue. HBA generations, connector families, breakout cables, cooling hardware and enterprise SSD candidates are separated. A marketplace phrase such as “IT mode” is treated as listing evidence, not proof of firmware state.
Buying decision
Buy the controller around ports, slot bandwidth and ownership of the disks
Count the drives you have now and the drives you expect to add, then map that number to internal HBA ports and chassis bays. If Proxmox itself will own ZFS, the HBA should present disks without hiding them behind a RAID virtual disk. If TrueNAS will own the pool, passing through the whole controller is often cleaner than passing individual disks. In either case, verify exact model, firmware state, connector type and airflow rather than relying on the marketplace phrase “IT mode” alone.
Interactive decision tool
Proxmox HBA Selector
Use this planner to screen controller class, ports, PCIe bandwidth, passthrough ownership, ZFS cache/allocation-device roles or expansion headroom before buying hardware. Results are planning guidance, not a substitute for motherboard lane/IOMMU documentation, exact HBA firmware verification, current OpenZFS behavior or independent backups.
Compatibility checklist
Four checks before adding an HBA or expanding storage
Decide who owns the disks
Proxmox-owned ZFS and a TrueNAS VM passed an entire HBA are different architectures. Do not let both layers manage the same disks.
Match ports and connectors
Count final drive lanes and verify SFF-8087, SFF-8643, backplane and breakout-cable direction before ordering.
Check PCIe lanes and IOMMU
A full-length slot may run electrically at fewer lanes. Passthrough also depends on platform IOMMU support and usable isolation.
Plan cooling and recovery
HBAs need airflow, and more attached disks increase the data you must protect with an independent backup and documented replacement map.
Start with the storage owner, not the card
Decide whether Proxmox or a storage VM owns the physical disks before shopping. When the Proxmox host owns ZFS, direct disk visibility makes health checks, replacement and topology easier to understand. When a TrueNAS VM owns ZFS, the cleanest design is often to pass through an entire HBA so the guest sees the controller and its attached disks as hardware. Mixing ownership across host and guest creates more recovery paths and more opportunities to attach the wrong disk to the wrong layer.
This ownership decision also affects boot design. Do not place the Proxmox boot device behind a controller you plan to pass through to a VM. Keep host boot storage on motherboard SATA, NVMe or a different controller so the hypervisor can start independently and then hand the storage HBA to the guest.
SAS2 is still enough for many hard-drive arrays
A 6Gb/s SAS link per port is not a practical limitation for ordinary hard drives, because a single HDD cannot saturate that lane in sustained real-world use. This is why older SAS2 HBAs remain useful in capacity-focused homelabs. The important limits become aggregate PCIe bandwidth, port count, expander design and cooling rather than the number printed beside each SAS lane.
SAS2 becomes less attractive when many SATA SSDs share one controller or when you want newer connector standards and denser cabling. If your expansion plan includes fast SSDs, a SAS3 or newer controller gives more headroom and is usually easier to integrate with modern SFF-8643 backplanes.
SAS3 changes cabling as well as link speed
The 9300-8i generation moves from the SFF-8087 connector common on 9207-class cards to SFF-8643 Mini-SAS HD. That means an existing pile of 8087 breakout cables is not automatically reusable. A controller upgrade can therefore require a cabling budget as well as a card budget.
For direct SATA drives, use the correct forward breakout cable from the HBA connector to individual SATA plugs. Reverse breakout cables are wired for the opposite direction and are a common source of “the card sees no drives” troubleshooting. The plugin keeps forward-breakout accessories separate so the article does not imply that every cable with the same connectors is interchangeable.
PCIe lane width can be the hidden bottleneck
An HBA with eight internal lanes may still be limited by the motherboard slot that feeds it. A physically x16 slot can be electrically x4, and consumer boards often share lanes between the second expansion slot, M.2 sockets and chipset devices. Read the motherboard lane-sharing table before assuming a controller will run at its nominal link width.
For HDD arrays, reduced PCIe width may still be acceptable. For many SSDs, especially during simultaneous reads, writes, scrubs or replication, aggregate controller bandwidth matters much more. The calculator estimates link-side headroom, but the motherboard manual remains the source of truth for slot wiring.
IT mode is a behavior to verify, not a magic label
In homelab discussions, “IT mode” is shorthand for firmware behavior that exposes attached drives directly instead of forcing them into RAID virtual disks. Marketplace sellers frequently put IT mode in a listing title, but that phrase does not prove firmware version, board identity or whether an OEM card was cross-flashed correctly.
On receipt, verify the controller identity, firmware package, attached-drive visibility and SMART passthrough before trusting it with irreplaceable data. Used HBAs can be excellent value, but they deserve the same validation you would give a used disk controller in a production server.
Eight ports versus sixteen or twenty-four
An 8i HBA is easy to cool and cheap to cable, but using two of them consumes two PCIe slots. A 16i or 24i board can consolidate a larger drive set into one slot, which is valuable on compact boards with only one usable x8-class expansion slot.
The tradeoff is thermal density and failure concentration. One high-port-count controller can connect much of the server, so a controller failure affects more bays. Leave physical access for replacement, document cable-to-bay mapping, and avoid packing the card against a hot GPU without directed airflow.
Cooling deserves a line item in the build
Server HBAs are commonly designed for chassis airflow that moves air across the heatsink continuously. A quiet tower or desktop case may not provide the same front-to-back pressure. A controller that is electrically compatible can still run uncomfortably hot if it sits in a stagnant pocket below a GPU.
Use the case layout, not internet folklore, to decide whether a small directed fan is useful. The goal is steady airflow across the heatsink without obstructing adjacent cards. Cooling accessories in this catalogue are intentionally separated from actual HBAs so a fan bracket can never masquerade as a storage controller.
SAS expanders are different from HBA ports
An HBA can connect to a SAS expander and reach more drives than its direct physical port count, but those drives share uplink bandwidth. This is often perfectly acceptable for a large HDD shelf and less attractive for a dense all-SSD pool.
Treat an expander as a topology decision rather than free ports. Check backplane compatibility, uplink lane count, connector standard and failure domain. For a small eight-drive server, direct breakout cables are simpler; for a chassis with many bays, an expander backplane can make cabling dramatically cleaner.
ZFS wants predictable disk visibility
ZFS is designed to manage redundancy and data placement itself. Hiding drives behind a hardware RAID virtual disk removes information and control that ZFS can otherwise use. A direct-attached HBA is therefore a natural match when the host owns a ZFS pool.
That does not mean every HBA automatically makes a good ZFS design. You still need redundant vdevs, independent backups, stable cabling, known drive identities and a recovery plan. The controller is only one layer in the storage stack.
Plan connector type before ordering cables
The 9207-8i and 9300-8i are often compared as if the upgrade were a card swap, but their internal connector families differ. That detail matters immediately if your server uses direct SATA fan-out cables or a backplane with fixed internal connectors.
Write down the connector at the HBA, the connector at the backplane or drives, and the required signal direction. Order cables only after that map is clear. This simple step prevents one of the most frustrating HBA installation failures: a perfectly good controller paired with a perfectly good cable wired for the wrong use.
Leave room for the next pool, not only today’s drives
If the server has six disks today and eight bays total, buying an 8i HBA may be rational. If the chassis has twelve bays and you know a second vdev is coming, a 16-port controller or expander architecture can avoid replacing the controller later.
Future planning also includes PCIe slots. A second HBA may collide with a GPU, 10GbE NIC or accelerator. The best HBA is the one that fits the whole host topology, not just the current disk count.
Document the build before data goes onto it
Record controller model, firmware version, slot, cable mapping and drive serials. If a controller fails two years later, this document turns replacement from archaeology into a controlled procedure.
Also keep a bootable recovery path and current backups outside the pool. An HBA simplifies direct disk access, but it cannot protect against accidental deletion, filesystem mistakes, multiple-drive failure, theft or a power event that damages several components at once.
Questions people ask
Best HBA for Proxmox questions
Is the LSI 9207-8i still good for Proxmox?
Yes for many HDD-focused systems. Its eight 6Gb/s SAS/SATA lanes and PCIe 3.0 x8 host interface are still practical when you do not need SAS3 cabling or a larger port count. Verify condition and firmware on used cards.
Is the 9300-8i faster than the 9207-8i?
The 9300-8i is a 12Gb/s SAS generation card while the 9207-8i is 6Gb/s SAS. That matters more with SSDs and aggregate workloads than with individual hard drives. The 9300 also uses SFF-8643 rather than SFF-8087 cabling.
Do I need IT mode for Proxmox ZFS?
You want the disks exposed directly rather than hidden behind hardware RAID virtual disks. Sellers often call that IT mode. Verify the actual controller and firmware behavior instead of trusting the listing phrase.
Can I use SATA drives on a SAS HBA?
Common LSI/Broadcom SAS HBAs support SAS and SATA drives, but the correct breakout cable or compatible backplane is required.
Should I buy an 8-port or 16-port HBA?
Buy enough ports for the final chassis plan, not only today’s disks. A 16-port card saves a PCIe slot in larger systems but increases thermal density and controller concentration.
Does a Proxmox HBA need active cooling?
Many server HBAs expect steady chassis airflow. Desktop cases can leave the heatsink in stagnant air, so a small directed fan can be sensible when measured temperatures or case layout indicate poor airflow.
Can a hardware RAID card be used instead?
It can, but hardware RAID changes disk visibility and ownership. For ZFS, a direct-attached HBA is generally easier to reason about because ZFS manages redundancy itself.
What cable does a 9207-8i use?
The 9207-8i uses internal SFF-8087 connectors. Direct SATA drives normally need a correctly wired forward breakout cable.
What cable does a 9300-8i use?
The 9300-8i uses internal SFF-8643 Mini-SAS HD connectors. That is a different cable family from the 9207-8i.
Can I pass the whole HBA to TrueNAS?
Yes when IOMMU and grouping allow it and the Proxmox host does not need any devices on that controller. Whole-controller passthrough is often cleaner than individual-disk passthrough for a storage VM.
Official references and methodology
Verify controller, passthrough and OpenZFS behavior before deployment
Controller facts are separated from marketplace claims. Exact Broadcom/LSI model documentation is used for known interface, connector and port characteristics. Amazon data is used only for current listing evidence, price, condition, seller and images. Firmware state, authenticity and actual delivered board must be verified after purchase.
- Proxmox VE Storage Manager
- Proxmox VE PCI(e) Passthrough
- Broadcom LSI SAS 9207-8i User Guide
- Broadcom LSI SAS 9300-8i User Guide
- Broadcom SAS 9305-16i Installation Guide
- Broadcom HBA/JBOD and RAID mode reference
As an Amazon Associate, Cloudzat may earn from qualifying purchases. Prices, seller claims, controller firmware, IT-mode status, exact board revisions, cable direction, enterprise SSD features and marketplace conditions can change. Verify the exact delivered hardware before deployment.