Nobody Told the Datacenter Liquidators That SAS Is Dead
Every few months someone posts in a home lab forum: “SAS is dead, just buy SATA.” And every few months you can still buy a 12TB HGST Ultrastar He12 SAS drive for $18 shipped from a decommissioned datacenter. So which is it?
Here’s the honest answer: SAS is absolutely dead for new hardware purchases. It’s also one of the best deals in storage if you know what you’re doing with used gear. Both things are true at the same time, which is a very hardware way of being confusing.
This isn’t a spec sheet comparison. It’s a guide for home labbers who are staring at an eBay listing for 24 SAS drives and a used HP SAS expander, wondering if they should pull the trigger.
The Actual Differences That Matter
Let’s get the specs out of the way fast so we can get to the real stuff.
SATA:
- 6 Gb/s (SATA III), single-port, half-duplex
- Simpler protocol, supported everywhere
- No end-to-end checksums in the protocol itself
- Cables top out around 1 meter reliably
- No native multipath
- Your NAS, your desktop, every consumer thing uses this
SAS (SAS3/12G, what you’ll actually find used):
- 12 Gb/s per port, dual-port, full-duplex
- End-to-end T10 DIF/DIX checksums — the drive verifies data integrity all the way up the stack
- Cables rated to 6–8 meters; SAS expanders extend that further
- Native multipath (dual-port drives can talk to two HBAs simultaneously)
- SAS expanders let you hang 24, 36, 48+ drives off a single HBA
- The enterprise pedigree: these drives were designed to run 24/7 in petabyte arrays
The bandwidth argument (“SAS is twice as fast!”) is mostly irrelevant in 2026. Your spinning rust — even a 7200 RPM SATA drive — tops out around 200–250 MB/s sequential. You’re not saturating a 6 Gb/s SATA link. The performance differences between SAS and SATA HDDs are noise.
What’s not noise: the expander ecosystem, multipath, cable runs, and the fact that enterprise SAS drives are often in better shape than equivalent SATA drives because they came out of maintained datacenters with proper cooling and monitoring — not someone’s dusty NAS closet.
The HBA: Your Entry Point
You’re not plugging SAS drives into a consumer motherboard. You need a Host Bus Adapter, and in 2026 the default answer is an LSI 9300-8i or 9305-16i flashed to IT mode.
IT mode means the card passes drives straight through to the OS — no fake RAID, no complexity, just “here are your drives, Linux.” This is what you want for ZFS, TrueNAS, or anything where the software manages the drives directly.
Finding these on eBay:
# What you're looking for: LSI 9300-series, IT mode flashed# 9300-8i: 8 internal SAS/SATA ports (two SFF-8643 connectors)# 9305-16i: 16 ports (four SFF-8643 connectors)# Budget: $20-40 for 9300-8i used, $40-70 for 9305-16i“IT mode” should be listed in the auction. If it’s not, you’re flashing it yourself — which is a 45-minute process involving a USB stick and some cursing, but it’s well-documented and survivable.
The connectors you’ll encounter:
- SFF-8643 (mini-SAS HD internal) — this is on your HBA, connects to backplanes
- SFF-8644 (mini-SAS HD external) — external connector variant
- SFF-8087 (mini-SAS internal) — older HBAs, supports SAS2/6G only
- SFF-8482 (legacy SAS drive direct) — rare, mostly ancient gear
For a modern SAS3/12G setup you want SFF-8643 throughout. Breakout cables go from one SFF-8643 to four individual SAS/SATA drive connectors, or you run to a backplane via a full SFF-8643 cable.
SAS Expanders: The Real Reason to Care
This is where SAS pulls away from SATA in a way that actually matters for big storage builds.
A SAS expander is a device that takes one or two SFF-8643 ports from your HBA and fans them out to 24, 36, or even 48 drive bays. The drives still talk directly to your HBA over SAS — the expander is transparent to the OS. There’s no fake RAID, no driver weirdness, just more drives.
Popular options in the used market:
- HP SAS Expander Card (part of the HP P2000/MSA family) — common, cheap (~$30–60), 24 ports
- Intel RES2SV240 — 24-port expander, very well supported, $40–80
- Supermicro BPN-SAS3 backplanes — built-in expanders in their 24-bay chassis
With one LSI 9305-16i and a pair of HP expanders, you’re looking at 48 drives hanging off a $60 HBA. Try doing that with SATA without a rat’s nest of port multipliers that Linux doesn’t fully support anyway. (SATA port multipliers are technically a thing. They’re also a thing that will make you sad at 2 AM.)
The Drives: What to Actually Buy
Used SAS drives worth buying in 2026:
HGST/Hitachi Ultrastar He Series (Helium)
- He10 (10TB), He12 (12TB), He14 (14TB) in SAS flavor
- These are helium-sealed, enterprise-grade, 7200 RPM
- eBay price range: $15–35 per drive depending on capacity and seller
- Look for “certified refurbished” or “datacenter pulls” — avoid “untested”
Seagate Exos X Series (SAS)
- Exos X14, X16, X18 in SAS variants
- Similar pricing to HGST, slightly more common
- Generally solid, run them through
smartctlbefore you trust them
What to skip:
- 15K RPM drives (SAS-only territory): They were fast in 2012. They’re loud, hot, and your NVMe cache tier does their job better.
- 10K RPM SAS: Similar story. The performance premium doesn’t justify the noise and heat in a home lab.
- Anything under 4TB: You’re wasting expander ports.
Checking What You Actually Got
When your drives show up, verify them before committing them to an array.
First, see what the OS sees:
lsblk -o NAME,TRAN,MODEL,SIZE,ROTATRAN will show sas or sata. ROTA is 1 for spinning, 0 for flash. Simple.
Now check the drive health with smartctl. SAS drives need the -d scsi flag or just let smartctl auto-detect:
smartctl -a /dev/sdbFor a SAS drive you’re looking for:
# Key fields in smartctl output for SAS:# Transport protocol: SAS# Current Drive Temperature: (should be reasonable — under 45°C at rest)# Elements in grown defect list: 0 ← this is your main health indicator# Non-medium error count: low number# read: Errors corrected by ECC, Total uncorrected errors: 0 ← this matters mostThe “grown defect list” is your primary indicator. Zero is good. Single digits are probably fine. Double digits on a cheap drive — return it or use it for cold archival only.
For deeper SAS-specific error counters, sg3_utils gives you access to the SAS log pages that SATA drives don’t have:
# Install sg3_utils if you don't have itsudo apt install sg3_utils # Debian/Ubuntusudo dnf install sg3_utils # Fedora/RHEL
# Read the Write Error Counter log page (0x02)sg_logs --page=0x02 /dev/sdb
# Read the Read Error Counter log page (0x03)sg_logs --page=0x03 /dev/sdb
# The full error counter page dump — good for validationsg_logs --page=0x11 /dev/sdbPage 0x11 is the “Background Scan Results” page — enterprise drives run background media scans and log the results. Zero errors in the scan log on a used drive is a good sign. Several hundred isn’t necessarily fatal, but it’s a yellow flag.
You can also verify dual-port connectivity by checking the SAS address and phy information:
# Show SAS topology — requires sas2ircu or systool# Or use lsscsi for a quick viewlsscsi -t
# sg_inq for device identification including SAS addresssg_inq /dev/sdbMultipath: The Feature You Actually Want For Redundancy
This is where SAS earns its keep for production-ish home lab setups.
A dual-port SAS drive has two independent SAS ports. Connect port A to HBA #1 and port B to HBA #2. Linux’s Device Mapper Multipath (dm-multipath) presents this as a single device. If an HBA dies, the drive stays online through the other path.
sudo apt install multipath-toolssudo systemctl enable multipathd --now
# Check multipath status after connecting drives to two HBAsmultipath -llYou’d see output like:
mpatha (3600508b1001c1234567890abcdef01234) dm-2 HGST,HUH721212AL5200size=12T features='1 queue_if_no_path' hwhandler='1 alua' wp=rw|-+- policy='service-time 0' prio=50 status=active| `- 3:0:0:0 sdb 8:16 active ready running`-+- policy='service-time 0' prio=10 status=enabled `- 4:0:0:0 sdc 8:32 active ready runningThat’s one drive, two paths. HBA 3 dies? Path through HBA 4 keeps the drive online. ZFS never knows anything happened.
You can’t do this with SATA. At all. SATA is single-port by spec.
For a home lab, is this overkill? Probably. But if you’ve got two HBAs and the drives anyway — why not?
When SAS Wins
Be honest about your use case before buying a pile of drives:
SAS is the right call when:
- You’re building a JBOD with 12+ drives and want expander support. SAS expanders are reliable and well-supported in Linux. SATA port multipliers are not.
- You want long cable runs. Running a SAS cable 4 meters to a separate chassis works fine. SATA gets sketchy past 1 meter.
- You’re buying used drives and the SAS options are cheaper per terabyte than SATA equivalents. This is often true — datacenter liquidations flood the market.
- You want multipath for HBA-level redundancy without spending money on a RAID controller.
- Your target is a large cold storage pool (backups, archival) where the enterprise write endurance of SAS drives matters.
SATA is fine (or NVMe is better) when:
- You’re adding a few drives to an existing SATA NAS or server. No reason to introduce SAS complexity.
- Your workload is random IOPS — an NVMe drive running as a ZFS SLOG/L2ARC cache tier will do more for you than any spinning disk protocol argument.
- You want simplicity. SAS requires an HBA, specific cables, usually a backplane, and some initial configuration. SATA plugs into your motherboard.
- You’re buying new drives in 2026. New SAS drives are expensive and mostly targeting enterprise OEM channels. New large SATA drives (WD Gold, Seagate Ironwolf Pro) are solid and readily available.
Power Draw Reality Check
One thing home labbers underestimate: 7200 RPM SAS drives are not efficient.
A typical 12TB HGST He12 SAS draws around 5–6W idle and 9–10W during seek. Multiply by 24 drives and you’re looking at 120–240W just in storage, before you count the HBA, expander, and chassis. That’s $20–30/month in electricity depending on your rates.
Contrast with:
- Modern SATA drives in idle spindown: 0.5–1W idle on many NAS-class drives
- NVMe SSDs: 2–5W under load, milliwatts at rest
If power efficiency matters — and it does once you’re paying the electric bill — factor this in. A 24-drive SAS array is a commitment. It’s also loud, because enterprise drives don’t care about your desk proximity.
The helium HGST drives are better than their non-helium counterparts in this regard, but “better” is relative.
The Bottom Line
SAS is absolutely worth it in 2026 if your use case fits the profile: large used-drive JBOD builds, expander-based scaling, or any setup where you want dual-port redundancy without enterprise RAID controller money. The hardware is cheap, Linux support is excellent, and the drives themselves often outlast the servers they came from.
It’s not worth it if you’re adding a few drives to a NAS, if SATA drives are price-competitive in your market, or if you just want things to be simple. The SAS ecosystem — HBAs, cables, expanders, backplanes — has a real learning curve and some upfront tooling cost.
The real villain in this story isn’t SATA. It’s the home labber who buys 24 SAS drives without understanding the cabling, flashes an HBA wrong, and posts “SAS doesn’t work” in the forum. Do the reading first. Flash to IT mode. Get the right SFF-8643 cables. Run smartctl before you commit a drive to an array.
Then enjoy your 288TB of storage for the price of a mid-range GPU.
