FAQ

General Questions

Why is Cruiser CM5 only?

Cruiser is designed around CM5’s USB 3.0 capabilities. CM4 lacks USB 3.0, which means:

• No 2.5GbE WAN (RTL8156BG requires USB 3.0)
• No SATA ports (JMS561 requires USB 3.0)
• No USB 3.0 ports (VL817 hub)

For full CM4 support, use the Interceptor Carrier Board.

What’s the difference between Cruiser and Interceptor?

*48V minimum for PoE models; Keel (non-PoE) can use 13V+

Choose Cruiser for: All-in-one NVR and smart home hub with optional built-in PoE

Choose Interceptor for: NAS with more SATA ports, CM4 compatibility, ATX power

What are the product variants?

Prices shown are MSRP. Visit the Exaviz Store for current pricing.

All variants include 2.5GbE WAN, M.2 NVMe, SATA, USB, HDMI, and ESP32-C6. The Keel has the same 4-port 1GbE switch as Cruiser Lite (similar to the Interceptor Carrier Board), but without PoE capability.

Power Questions

What power supply do I need?

It depends on your model:

With Interceptor PoE Board add-on:

Connector: Use a 5.5/2.1mm barrel jack (center positive). The barrel jack connector is rated for 7A maximum—do not exceed this rating.

Can I use a lower voltage (12V, 24V)?

For PoE models (Cruiser, Cruiser Lite): No. IEEE 802.3at PoE+ requires 48V minimum to deliver power to connected devices.

For Cruiser Keel (non-PoE): Yes! Since there’s no PoE circuitry, you only need enough voltage to run the DC-DC converters for the board and switch electronics:

• 13V minimum for stable operation
• 24V recommended for good efficiency
• 12V is too low (insufficient headroom for the buck converters)

The Keel’s 4-port 1GbE switch works normally at lower voltages—only PoE injection requires 48V.

What’s the maximum input voltage?

57V absolute maximum for all models (Cruiser, Cruiser Lite, Cruiser Keel).

Why use an Exaviz power supply?

Short answer: because we tested them and know they work.

When sourcing a PSU for the Cruiser, we evaluated multiple brands and selected LYD for consistent voltage regulation, thermal performance, and certification depth. Both our 4A and 6A models are UL Listed (cULus) under file number E336005 and carry FCC, CE, RoHS, and SAA certifications.

Our test results at rated load, sustained for 30 minutes:

The 6A model runs notably cooler than the 4A despite its higher output, giving you additional headroom for demanding configurations. Both idle at under 9W, making them well suited for always-on applications like NVR systems and smart home hubs.

You can find cheap 48V supplies from unknown brands on various sites. You might get lucky, or you might see voltage sag under load, poor efficiency, or a unit that runs hot and fails early. With an Exaviz LYD supply, you know exactly what you’re getting - and we back it with a 1-year warranty.

See the Power Supply Options section for full specs and pricing.

PoE Questions

How much power can each PoE port deliver?

Up to 30W per port (IEEE 802.3at PoE+). The total power budget depends on your power supply capacity.

Are the PoE expansion ports also Gigabit?

No. The 8 built-in ports are Gigabit (1GbE). The expansion ports via the Interceptor PoE Board are 100Mbps only.

Can I add more than 8 PoE ports?

Yes. Connect an Interceptor PoE Board via the FFC connector (J18) for +8 additional ports (100Mbps), for a total of 16 PoE ports.

See Interceptor PoE Board - Getting Started for installation instructions.

Storage Questions

Can I boot from NVMe?

Yes. CM5 supports NVMe boot. Flash the OS directly to the NVMe drive using:

1. USB-to-NVMe adapter on another computer
2. Boot from SD card, then flash to NVMe

Configuring boot order (CM5 with eMMC):

If you have CM5 with eMMC but want to boot from NVMe, configure the boot order using rpi-eeprom-config. Unlike CM4, CM5 allows EEPROM updates from the running system:

# Install if needed
sudo apt install rpi-eeprom

# Edit boot order
sudo rpi-eeprom-config --edit

Add or modify this line:

BOOT_ORDER=0xf2416

This sets boot order to: NVMe, then SD/eMMC, then USB, then Network, then restart the loop.

BOOT_ORDER is read right to left: the lowest nibble is tried first. The digits are 1 SD/eMMC, 2 Network, 4 USB-MSD, 6 NVMe, f restart. So 0xf2416 tries 6 (NVMe) first. Reading the value left to right inverts the whole order.

The EEPROM will be updated on the next reboot.

Why does USB boot take so long?

If your Cruiser takes 45-60 seconds to boot from a USB flash drive, the issue is the default EEPROM boot order, not the hardware. See Troubleshooting: Slow USB Boot for the full explanation and fix.

Quick fix:

echo -e "[all]\nBOOT_UART=1\nBOOT_ORDER=0xf2614" | sudo tee /tmp/boot.conf
sudo rpi-eeprom-config --apply /tmp/boot.conf
sudo reboot

This puts USB first in the boot order. The default order tries SD and NVMe first (which fail instantly), then reaches USB before the drive has finished enumerating, falls through to Network boot (30-60 second DHCP timeout), and only succeeds on USB during the second loop. Setting USB first lets the bootloader’s own startup time cover the USB enumeration delay.

Why does my SD card boot from power-on but not after a reboot?

Some fast microSD cards cannot be re-initialized after a warm reboot on a CM5 Lite, because the reboot does not power-cycle the card. See Troubleshooting: SD card boots from power-on but not after a reboot for the symptoms and the device-tree overlay that fixes it.

Can I use M.2 SATA SSDs with an adapter?

Yes. Any adapter that converts an M.2 SATA SSD into a standard SATA interface will work with the Cruiser’s SATA ports. A few things to keep in mind:

• M.2 SATA only, not NVMe - Make sure the M.2 SSD is a SATA model (B-key or B+M key), not NVMe. They look similar but are electrically different.
• No capacity limit - There is no drive capacity restriction on the Cruiser’s SATA ports.
• Speed - With HDDs, the drives are the bottleneck, not the interface. With SSDs, you’ll get around 340 MiB/s sequential, which is where the USB 3.0 bus tops out. The real SSD advantage is random I/O, which improved 16-49x over HDDs in our testing.
• Power - M.2 SATA SSDs draw very little power, so this is a non-issue.
• RAID still works - If you use two adapters (one per SATA port), the hardware RAID modes (RAID 0, RAID 1, JBOD, SPAN) work the same as with traditional drives.

Does the hardware RAID work with NVMe?

No. The hardware RAID (DIP switch) only applies to the two SATA ports.

Can I use more than 2 SATA drives?

Yes, via an M.2-to-SATA adapter card (JMB585-based). This adds up to 5 more SATA ports using the M.2 slot.

Storage expansion math:

• 2 built-in SATA
• +5 via M.2 adapter
• +USB 3.0 enclosures
• = 10+ drives possible

What’s the difference in SATA architecture vs Interceptor?

Cruiser: JMS561 on a dedicated USB 3.0 bus (not shared) Interceptor: JMB585 on shared PCIe bus

For typical 2-drive configs, Cruiser’s dedicated bus may actually be faster than Interceptor’s shared PCIe topology.

RAID Questions

How do I check that the SATA bridge is working?

Run lsusb | grep -i 152d. A 152d:9561 JMicron line is expected. Anything else, or no JMicron line, means contact support. See Troubleshooting: Is my SATA bridge working?.

Why does only one of my two SATA ports detect a drive?

Check the RAID mode before concluding the port is dead. Only JBOD (both S3 switches OFF) presents two independent drives. In RAID 1 the controller presents a single logical volume at single-drive capacity and does not expose the second drive as its own device, so a board set to RAID 1 can look exactly like a faulty second port: nothing in lsblk, and the port’s activity LED dark.

The RAID mode applies to the whole JMS561 controller, not to one port. See Hardware Interface - RAID DIP Switch for the switch positions.

To tell configuration from a real fault, read the controller directly with the drive in the suspect port only:

sudo raidmgr_static
# At the JMS56X> prompt:
DC C0
SS C0
EX

If the drive is listed on one of the channels, the port and link are working and the mode is the problem. If neither channel lists it with a known-good drive and cable, contact support. Full walkthrough: Troubleshooting - Only one SATA port detects a drive.

Why don’t my DIP switch changes take effect?

The most common cause: an existing RAID array must be deleted first. The DIP switches only configure a new array when no array currently exists.

To delete the existing array (see JMicron RAID Manager for tool installation):

sudo raidmgr_static
# At the JMS56X> prompt:
DR C0 R0
EX

Then power off, set switches, hold the button during power-on for 5 seconds, and release.

See Hardware Interface - RAID DIP Switch for the complete procedure.

How do I check the current RAID mode?

Use the raidmgr_static tool (see JMicron RAID Manager for installation):

sudo raidmgr_static
# At the JMS56X> prompt:
SR C0

GC only reports how many controllers are present (Total available JMS56X controllers = 1); it does not show the mode. SR C0 is the command that reports array status, and DC C0 or SS C0 list the individual drives the controller sees on each channel.

The output shows the array type:

• H/W RAID0 / H/W RAID1 / H/W JBOD — configured via DIP switches
• S/W RAID0 / S/W RAID1 / S/W JBOD — configured via software tool
• No RAID Volume — two independent drives (true JBOD)

How long do I hold the RAID button?

Hold the button (S2) for at least 5 seconds during power-on, then release. Shorter presses may not register.

Can I create SPAN mode via software?

No. SPAN is hardware-only — it can only be configured via DIP switches (SW1=OFF, SW2=ON).

The software tool (raidmgr_static) supports RAID 0, RAID 1, and JBOD, but returns “Incorrect RAID type” if you try to create SPAN.

Pressing the button while running does nothing

The RAID button only works during power-on. Pressing it while the board is running has no effect.

For live RAID changes without power cycling, use the software tool:

sudo raidmgr_static
# Create RAID 0:
CR C0 D0,1 RAID0
# Or RAID 1:
CR C0 D0,1 RAID1

Do I need to unplug the board to change RAID mode?

No. A full shutdown (sudo poweroff) is sufficient. Unplugging is not required, though it doesn’t hurt.

Networking Questions

Are the switch ports managed or unmanaged in Linux?

The ports are managed. In Linux they are exposed via the DSA (Distributed Switch Architecture) framework, which means each port (poe0–poe7, or poe0–poe15 with the add-on board) appears as its own independent network interface. The exaviz-netplan and exaviz-dkms packages configure these interfaces automatically on first boot.

Standard Linux tools work on each port individually:

# Show port status and link speed
networkctl status poe0

# Per-port traffic statistics
ethtool -S poe0

# Bring a port up or down
ip link set poe0 down

# Bridge ports into a LAN segment
ip link add br0 type bridge
ip link set poe0 master br0
ip link set poe1 master br0

# VLAN subinterface (software tagging on the Linux side)
ip link add link poe0 name poe0.100 type vlan id 100
ip link set poe0.100 up

VLAN limitation: The above creates a software VLAN subinterface in Linux, which tags traffic leaving the CM5 toward that port. However, the RTL8367RB switch chip driver (rtl8365mb) does not yet implement hardware VLAN filtering or bridge VLAN support in the upstream Linux kernel. This means the switch itself cannot segregate traffic between ports at the hardware level based on VLAN membership. Full VLAN management (802.1Q trunking, port-based VLANs) is not yet supported and will be documented when the upstream kernel driver adds this capability.

Note: /dev/pse is the PoE power interface (enable/disable power delivery, read per-port wattage). It is separate from network management and is what the Home Assistant plugin uses. Network configuration uses standard Linux tools as shown above.

Is the 1GbE switch uplink a bottleneck for NVR use?

No. The CM5’s 1GbE uplink only carries traffic destined for the CM5 itself. Port-to-port switching between connected devices is handled entirely in hardware by the RTL8367RB chips – including traffic crossing between the two chips via their direct RGMII inter-chip link. See Switch Architecture for details.

For the traffic that does use the CM5 uplink (camera streams being written to storage), the bandwidth math is straightforward:

The 1GbE uplink (1000 Mbps) is not saturated in any realistic IP camera scenario. Even 16 cameras at maximum 4K bitrate uses only 40% of the available capacity.

Why isn’t my WAN port working?

The 2.5GbE WAN port (eth1) is not configured by default and requires manual setup. Check if the interface exists:

ip link show | grep -E "eth1|wan|enx"

See Configuring the Upstream / WAN Port for setup instructions.

Why did my WAN port stop working after the first reboot?

On Raspberry Pi OS (Trixie), exaviz-netplan versions before 1.1.11 can leave the WAN port with no wan name and no address after the first reboot. See Troubleshooting: WAN port has no name and no address after a reboot for the upgrade and the offline fix.

Is the 2.5GbE WAN port backwards compatible?

Yes. The RTL8156BG supports 10/100/1000/2500 Mbps auto-negotiation.

Can I use the PoE ports as regular switch ports?

Yes. Devices that don’t require PoE will work normally. The PoE power is only delivered if the device negotiates for it (802.3af/at).

Can I put all the ports on one subnet, like a normal switch?

Yes. By default exaviz-netplan puts each PoE port on its own /24 with a small local DHCP server. That isolates each camera and needs zero configuration, but the ports are not on your main LAN and do not get addresses from your router. If you would rather run the board as a flat switch - all PoE ports plus the WAN uplink on one subnet served by your own router - that is a flat “switch mode.” See Configuring the Switch Network for the tested step-by-step.

What do the interface names (ext0_1, poe0@ext0_1) mean?

They reflect how the two switch chips are wired internally. The @ in ip link output shows each interface’s parent:

• eth0 is the CM5’s internal link to the switch (the DSA conduit / CPU port). Not a front-panel port.
• ext0_1 and ext0_2 are internal links on the first switch chip. ext0_1 also carries traffic to the second (cascaded) switch chip.
• poe0-poe7 are the front-panel ports. poe4@eth0-poe7@eth0 sit on the first chip; poe0@ext0_1-poe3@ext0_1 sit on the second chip, reached through ext0_1.

For normal use you only ever touch eth1 (WAN) and poe0-poe7. Leave eth0, ext0_1, and ext0_2 alone - they are the internal plumbing that must stay up for the switch to work.

I edited a Netplan file and it changed (or my port stopped working) after a reboot

On Raspberry Pi OS / Debian the Netplan back end is NetworkManager, which rewrites everything in /etc/netplan/ into its own 90-NM-*.yaml files on every boot and may drop a hand-written interface name. This is expected behavior, not a bug - it is why a bridges: stanza you add by hand does not stick. On these images, configure the network through NetworkManager directly (nmcli); those connections persist. The switch-mode recipe uses this approach. On Ubuntu (which uses systemd-networkd), hand-written Netplan files are not rewritten.

My PoE cameras can’t reach the internet (NTP, firmware updates)

In the default per-port layout, each camera is on its own isolated subnet and the board does not route or NAT between those subnets and the WAN, so the cameras have no path out. This is intentional - the per-port mode is for keeping cameras off your main network. If your cameras need internet access (NTP, cloud, updates), use switch mode so they get addresses and a gateway from your own router, or set up routing/NAT yourself if you want to keep them on separate subnets.

How do I find a connected device’s IP address?

# Check PoE port status
networkctl status poe0

# Or use ip addr
ip addr show poe0

Why doesn’t my WAN port green LED light?

The green WAN LED (1 GbE / 2.5 GbE indication) requires the exaviz-netplan package; install or update it and the green LED enables automatically. The amber LED (10/100 Mbps and 1 GbE) works without any Exaviz software. See WAN Port Green LED Not Working.

Wireless Questions

Does the WiFi work like built-in CM5 WiFi?

Yes. The ESP32-C6 uses SDIO (same interface as native CM5 WiFi) with ~200 Mbps throughput. Performance is comparable to CM5’s built-in WiFi.

See the ESP32 Wireless Module page for setup instructions.

Can I use CM5 without built-in WiFi?

Yes, unless you also want Zigbee. You can buy CM5 modules without WiFi and use the ESP32-C6 WiFi instead to save money. The ESP32-C6 provides WiFi 6 (2.4 GHz) with comparable performance to CM5’s built-in WiFi.

If you want Zigbee, choose a CM5 with built-in WiFi. The Zigbee firmware replaces the WiFi + Bluetooth image on the ESP32-C6, so a board running Zigbee cannot also provide WiFi from it. The ESP32-C6 also uses an external antenna connector and the board ships without an antenna.

See the ESP32 Wireless Module page for configuration details.

How do I use Zigbee?

The ESP32-C6 can act as a Zigbee coordinator without external USB dongles, which makes Cruiser a good fit for smart home applications with Home Assistant.

See the ESP32 Wireless Module page for setup instructions.

Can I use the Zigbee coordinator with Home Assistant today?

Yes. Install the Zigbee firmware with sudo exaviz-flash zigbee, then in Home Assistant add Zigbee Home Automation, choose the Ziggurat radio type and enter the ws:// address the install prints. Home Assistant 2026.7 or newer is required. zigbee2mqtt and deCONZ are not supported. Setup, antenna and troubleshooting: Zigbee.

What about Thread and Matter?

Contact support@exaviz.com for assistance with disabling ESP32 firmware encryption if you need to install custom firmware.

M.2 Slot Questions

Can I use an AI accelerator instead of NVMe?

Yes! The M.2 slot accepts:

• Hailo-8 (26 TOPS AI)
• Coral TPU
• PCIe adapters

How can I add cellular (4G/5G) connectivity?

The Cruiser’s M.2 slot uses Key-M (22mm width) for NVMe/PCIe devices. Cellular modems use Key-B (30mm width) and will not fit. For 4G/5G connectivity, use a USB dongle plugged into J9 (the front USB 3.0 port). When the Cruiser is mounted in a case, J9 faces inside the case, keeping the dongle internal.

4G USB Dongles (~$30-75)

5G USB Dongles (~$70-350)

Is PCIe Gen 3 supported?

Unofficially, yes. Add dtparam=pciex1_gen=3 to /boot/firmware/config.txt for ~1 GB/s speeds (vs ~500 MB/s for Gen 2).

Not officially supported by Raspberry Pi, but multiple users report reliable operation. See Jeff Geerling’s PCIe testing for benchmarks.

Qwiic Questions

What voltage are the Qwiic connectors?

3.3V by default. 5V is available via a solder bridge change, but this is intentionally difficult to prevent accidental damage to 3.3V-only devices.

What can I connect to Qwiic?

Any SparkFun Qwiic or Adafruit STEMMA QT device:

• Environmental sensors (temperature, humidity, air quality)
• Motion sensors (accelerometers, gyroscopes)
• Displays (OLED, LCD)
• GPIO expanders
• And many more!

Case/Mounting Questions

Will it fit in a Mini-ITX case?

The board is 170mm x 120mm (same width as Mini-ITX, but 50mm shorter). It will fit physically, but mounting holes don’t align with standard Mini-ITX patterns.

Use standoff spacers/adapters for proper mounting.

What standoff spacers do I need?

M3 standoffs, 6-10mm height. Exact requirements depend on your case.

Logging Questions

Why are there no logs from before a reboot?

Raspberry Pi OS keeps the systemd journal in RAM only (Storage=volatile), so every reboot discards all logs. This is a Raspberry Pi OS default rather than an Exaviz setting.

It matters because rebooting a misbehaving board destroys the evidence of why it misbehaved. We recommend enabling persistent logging, particularly on any board you are actively diagnosing.

Note that setting Storage=persistent in /etc/systemd/journald.conf does not work — the Raspberry Pi drop-in takes precedence and your setting is silently ignored.

See No logs from before a reboot for how to enable it, how to verify it worked, and the write-wear trade-off to weigh first.

Troubleshooting

For common issues and solutions, see the Troubleshooting page.

Last modified October 13, 2026