US How a RAID controller works
A RAID controller is the hardware or software that manages a group of drives as a RAID array. It decides where each block of data goes, keeps mirror copies or parity up to date, and presents the array to the operating system as one or more logical drives, often called virtual disks. The operating system never sees the individual drives, only the volume the controller builds from them.
The controller also runs the array day to day. It detects a failed drive, switches to a hot spare, rebuilds the missing data onto the replacement and runs background checks, sometimes called patrol reads or consistency checks, to find bad sectors before they cause a second failure.
When people say RAID controller they usually mean a RAID card: a PCIe add-in card with a dedicated RAID-on-chip processor, its own cache memory and SAS, SATA or NVMe connectors. The same job can also be done by firmware in the motherboard chipset or by software in the operating system.
Types of RAID controller
RAID can be run in four places, and the choice decides where the work is done and how portable the array is.
-
//
Hardware RAID card. A dedicated processor calculates parity and manages the array, with a write cache to absorb bursts. The array is configured in the card’s own setup utility, and the operating system only needs the card’s driver.
-
//
Firmware or chipset RAID. RAID built into the motherboard, such as Intel Rapid Storage Technology or Intel VROC for NVMe on Xeon platforms. The firmware sets up the array, but the host CPU and a driver do the work.
-
//
Software RAID. The operating system manages the array: mdadm or ZFS on Linux, Storage Spaces on Windows. It needs no special hardware and the array can move to any computer running the same software.
-
//
Host bus adapter (HBA). A card with the same connectors as a RAID card but no RAID logic. It passes each drive straight to the operating system, which suits software RAID and ZFS. Many cards can switch between RAID and HBA (often called IT or JBOD) mode.
Hardware RAID vs software RAID
The choice between hardware RAID and software RAID comes down to where the parity maths runs and what happens when something fails. Hardware RAID does the work on the card, keeps a protected write cache and presents the array the same way under any operating system, so a Windows or Linux reinstall never touches it. The cost is the card itself, its cooling and the need for a replacement from the same family if it fails.
Software RAID uses the host CPU and system memory. Modern processors handle RAID 1 and 10 with little measurable load, and parity RAID costs more but is rarely the limit on a current multi-core system. It needs no special hardware, its arrays move easily between computers running the same software, and with NVMe drives it is often faster, because no card sits between the drives and the CPU. The trade-offs are that the boot drive needs extra care, the array depends on that operating system, and without a protected cache a power cut can leave parity writes incomplete.
No option is best everywhere. The table adds firmware RAID and the HBA, and compares all four approaches on the points that usually decide it.
| Approach | Who does the work | Write cache | Moving the array | Typical use |
|---|---|---|---|---|
| Hardware RAID card | Dedicated RAID-on-chip processor | Yes, on the card, needs protection | Needs a compatible controller from the same family | Parity RAID on many SAS or SATA drives, operating-system-independent arrays |
| Firmware (chipset) RAID | Host CPU, set up by firmware | No dedicated cache | Needs the same chipset family | Mirrored boot drives, simple RAID 0 or 1 |
| Software RAID | Host CPU and operating system | Uses system memory | Easy between systems with the same software | NVMe arrays, Linux servers, ZFS storage |
| HBA | None, drives passed through | No | Drives readable on any HBA | Software RAID, ZFS, storage software that wants raw drives |
What to look for in a RAID card
If a hardware RAID controller is the right fit, these are the specifications that matter. Compare them against the drives and enclosure, not just the RAID levels.
-
//
Drive interfaces. SATA-only, SAS/SATA or tri-mode, which adds NVMe. SAS controllers also accept SATA drives; our explainer on the difference between SAS and SATA covers why the reverse does not work.
-
//
Supported RAID levels. Most cards cover RAID 0, 1, 5, 6, 10, 50 and 60. Entry-level cards may stop at 0, 1 and 10.
-
//
Cache size and protection. A write-back cache speeds up parity writes, but its contents must survive a power cut. Look for a flash-backed cache with a supercapacitor or a battery backup module.
-
//
Ports and connectors. Internal ports usually use Mini-SAS HD (SFF-8643) or SlimSAS (SFF-8654) connectors, each carrying four or eight drive lanes, so check the cables match the backplane.
-
//
Host interface and drivers. Most cards use PCIe 3.0 or 4.0 x8. Check driver and management-tool support for the exact Windows or Linux release you will deploy.
-
//
Cooling. RAID-on-chip processors run hot and most cards specify a minimum airflow over the heatsink, which needs planning in compact or fanless enclosures.
Do you need a RAID controller?
Every RAID array needs a controller of some kind, but not every system needs a RAID card. For a pair of mirrored boot drives, chipset RAID or operating-system mirroring is usually enough. A hardware card earns its place when you run parity RAID across many SAS or SATA drives, when the array must work the same way under any operating system, or when a protected write cache is needed for steady, write-heavy loads such as video recording.
NVMe has changed the picture. Modern NVMe drives are so fast that a RAID card can become the bottleneck, so NVMe arrays more often use software RAID, Intel VROC or a tri-mode card designed for NVMe. Whichever you choose, protect the data in flight: a write cache, or a drive without power-loss protection on the SSD, can lose recent writes when the power fails.
RAID controllers in industrial and edge systems
At the edge, RAID is mostly about keeping a system running and its recordings intact until someone can reach it. Industrial NVRs record cameras around the clock, and many in our range support RAID across several drive bays, often RAID 1 for two drives or RAID 5 for more.
For heavier storage, some of our rugged computers combine hot-swap U.2 NVMe bays with hardware RAID, so a failed drive can be replaced on site without shutting the system down. Edge storage servers built on Intel Xeon D processors add RAID 5 and RAID 10 across ten U.2 bays. In each case, check whether RAID comes from a dedicated controller, the chipset or software, because it decides how the array is rebuilt and what happens if the board is ever replaced.
Quick answers on RAID controllers
RAID controllers: common questions
What is a RAID card?
A PCIe card that works as a hardware RAID controller. It has its own processor, cache and drive connectors, and presents the array to the operating system as one or more logical drives.
Is hardware RAID better than software RAID?
Not always. Hardware RAID offloads parity work and adds a protected cache, which suits parity arrays of SAS or SATA drives. Software RAID is cheaper to deploy, easier to move and often faster with NVMe.
What is HBA mode on a RAID controller?
A pass-through mode with RAID turned off. The card presents each drive to the operating system individually, which suits ZFS and other software RAID.
What happens if the RAID controller fails?
The data is normally still on the drives. Fit a replacement from the same controller family and it can usually import the array configuration. Software RAID arrays can be read on any compatible system.
Does a RAID controller need a battery?
Only if it uses a write-back cache. A battery or supercapacitor-backed flash module keeps cached writes safe through a power cut. Without one, run the cache in write-through mode.
We fit and configure the RAID controller, drives and array as part of the standard build, and document the RAID level and rebuild procedure before the system ships.