Comino GrandoUser manual · pilot
Your unit not chosen yet
Comino GrandoUser manual · pilot

Which unit is in front of you?

Two answers and the manual shows only what applies to your machine. Both are printed on the Device Passport that came in the crate.

1Product

2Controller on your unit

Not sure? The Device Passport lists the controller generation and firmware on its first page. If you cannot find it, write to support@comino.com with the serial number.

Pick the product and the controller

Other Comino products (Heat Rejection Unit, Integration Kit, Mobile Data Center) get their own manuals later.

What changed
  • Logika 2.5 firmware 0.0.11: event log and service menu order changed
  • Logika 2.0 firmware 1.1.9: profile FAN PWM-NC added, Single PSU mode removed

Grando Server · RM-M-CRPS

Server manual for the unit in front of you

Only the Server procedures are shown, for the controller generation you chose. Chapters that differ by generation show only your version. Everything that could stop the machine from starting sits behind a confirmation or stays with service.

Picked the wrong unit? Use Change in the top bar.

0About this manual

0.1 Models covered User

ModelChassis codePower systemControllerFront panel
Grando ServerRM-M-CRPS4 hot-swap CRPS modules (Energia)Logika 2.0 or 2.5 (see Device Passport)2 buttons; Facelift: 3 buttons
Grando WorkstationRM-Mup to 3 SFX-L PSUs or 2 ATX-LLogika 2.0 or 2.52 buttons
Previous generationRM-L, RM-SSFX / ATXLogika 2.02 buttons

RM-L and RM-S are no longer produced. This manual keeps the notes that differ for them where a procedure differs (for example, extra screws on the top panel). Everything else applies to all three.

0.2 Documents that come with the unit User

  • Quick Start Guide (printed, ver. 2.0.3): placement, first power-up, on/off, coolant, noise profiles, support.
  • Device Passport (A4, folded): model, serial number, serial numbers of every component, controller generation and firmware, factory cooling profile, warranty data.
  • Unpack leaflet: how to lift the unit out of the crate, what to keep.
  • Invoice travels outside the crate with the shipping documents.

Older units shipped with a USB stick holding this manual. The USB stick in current packages holds the BIOS image only, see 5.2.

0.3 How warnings are marked User

DANGERAn imminently hazardous situation which, if not avoided, will result in death or serious injury.
WARNINGA potentially hazardous situation which, if not avoided, could result in death or serious injury.
CAUTIONA potentially hazardous situation which, if not avoided, could result in minor or moderate injury.
NOTICEA situation which, if not avoided, could result in property damage.

1Safety

1.1 Intended use User

  • Professional computer for indoor use. Not for caustic or potentially explosive environments, not for medical purposes.
  • Use only with original accessories and components; use only within the power limits in chapter 6.
  • Ambient temperature +3 °C to +38 °C, no direct sunlight, no covered vents. See chapter 4 for the full table.

1.2 Lifting and placing User

WARNING A Server with eight GPUs weighs about 55 kg net, 72 kg in the crate. Lift with two people or a lift. Keep the unit horizontal or vertical with the reservoir on top; any other orientation stops the cooling loop and voids the warranty.
  • Rack: use rails or a shelf rated for the weight; a 4U unit on handles alone will bend the rack ears.
  • Desk: use the legs kit; leave 10 cm free in front of the fans and behind the radiator.
  • Never place objects on the unit, never insert foreign objects into vents.

1.3 Electrical User

DANGER A Server has up to four independent mains inlets. Unplug all of them before opening the covers. One cord removed still leaves the unit live.
  • Use only the supplied cords. If an adapter or extension is unavoidable, have an electrician confirm the protective earth is kept.
  • Each power module can draw 15 A at 100–127 V. Branch circuits: C16 per module or C25 per pair (chapter 6).
  • Do not use the unit during a thunderstorm, near water, with wet hands, or with damaged cords.
  • Do not switch the unit on with the side covers removed.

1.4 Liquid loop User

  • The coolant is a water and propylene-glycol mixture with corrosion inhibitors and biocides. Not for drinking. Rinse skin with water; if it reaches the eyes, rinse and see a doctor.
  • Keep paper towels around the reservoir when refilling. Coolant on a GPU means a failed GPU.
  • Never run the unit with the reservoir empty or with the top panel open for long: the fans are the only thing cooling the RAM.
  • A unit that was shipped may hold pressure in the loop. Open the reservoir cap slowly.

1.5 Children and packaging User

Packaging foam and bags are a suffocation hazard. Keep them from children; keep the crate itself (chapter 14).

2Compliance & certifications

The unit is tested as a liquid-cooling unit with its control system, without the variable components (GPU, CPU, RAM); shielded cables must be used.

2.1 European Union User

Both models comply with 2014/30/EU (EMC), 2014/35/EU (LVD) and 2011/65/EU (RoHS), tested to EN 55032:2015/A11:2020, EN 55035:2017/A11:2020, EN IEC 61000-3-2:2019, EN 61000-3-3:2013/A2:2021/AC:2022-01, EN IEC 62368-1:2020, IEC 63000:2018.

Manufacturer: CMNO Logica SIA, Carnikavas street 5a, Riga, LV-1034, Latvia, reg. LV43603041824.

2.2 United States User

The Server was tested to the FCC Part 15 Class A limits (commercial installation), the Workstation to Class B (residential). Operation is subject to the two usual conditions: the device may not cause harmful interference and must accept any interference received.

3Product overview

3.1 Front panel, Server User differs by generation

Service Manual illustration
Server front panel: 1 reservoir window, 2 display, 3 PB1 (PREV), 4 PB2 (NEXT), 5 power-on indicator, 6 motherboard I/O, 7 PCIe cards, 8 rack handles.
Workstation front elements
Workstation (RM-M) front elements: reservoir, display, PB1, PB2, power-on indicator, motherboard I/O, PCIe area, handles.
FaceliftThree buttons instead of two.
Logika 2.0 / 2.5Two buttons, PB1 and PB2, left and right of the display. Short press moves, holding confirms or cancels (chapter 9).
Logika 2.0 / 2.5Two buttons, PB1 and PB2, left and right of the display. Short press moves, holding confirms or cancels (chapter 9).

3.2 Rear panel and power modules, Server User

Service Manual illustration
Four CRPS modules with handles; one status LED per module.
PSU LEDMeaning
GreenNormal operation
Fast blinking green (1 s)PSU standby, AC present
Slow blinking green (3 s)Standby / low-power mode
Slow blinking amber (1 s)Warning: high temperature, fan or over-current warning, module keeps running
AmberAC cord unplugged, AC lost, or a protection tripped (UVP, OVP, OCP, OTP)
OffNo input power

3.3 Rear panel, Workstation User

#ElementPurpose
1, 3, 5Power switches, PSU 3 / 2 / 1Switch each PSU on or off
2, 4, 6Power inlets, PSU 3 / 2 / 1One cord per PSU; all PSUs must be powered
7Thermal sensorNo user interaction
8Radiator end capNo user interaction
Workstation rear panel elements
Workstation rear panel: PSU switches and inlets 1–6, thermal sensor 7, radiator end cap 8 (current FAQ drawing).

3.4 Inside the unit User

Service Manual illustration
Server: fans, up to 8 GPUs with water blocks, reservoir with integrated pumps, 360 mm radiator, CRPS board, controller. Hardware (motherboard, CPUs, RAM, drives) is as per your invoice.

Airflow goes front to back. The motherboard sits at the front so RAM and other air-cooled parts get the coldest air. The cooling loop is sealed and filled at the factory; the only user touch points are the reservoir cap and the quick disconnects on the GPUs (chapter 12).

3.5 Labels and serial numbers User

  • Unit serial number (format 500.C.xxxxx.03): sticker on the rear panel next to the power modules, repeated on the Device Passport and the invoice. Support asks for it first.
  • Legacy marking on the rear sticker tells you whether the 10-second power-up rule applies (chapter 6).
  • Controller generation and firmware: Device Passport, and the FIRMWARE status screen on the display (chapter 9).
  • Component serials (motherboard, CPU, GPUs, drives, PSUs): all on the Device Passport; you do not need to open the unit to read them.
  • Regulatory label with CE marking: rear panel.

4Specifications & environment

4.1 Server (RM-M-CRPS) User

Form factor4U, rack-mountable or standalone; 439 × 681 × 177 mm without handles
Weight4 CRPS + 4 GPUs 49 kg net / 67 kg gross; + 6 GPUs 52 / 70; + 8 GPUs 55 / 72
Motherboardup to E-ATX / EEB; single or dual socket (Intel Xeon Scalable 4th/5th gen, Xeon 6; AMD EPYC 9004/9005, Threadripper PRO)
GPUsup to 8, liquid-cooled: NVIDIA RTX PRO 6000 Blackwell Server Edition, RTX PRO 5000, H200 NVL (subject to availability), L40S, RTX 6000 Ada; AMD W7800/W7900.
Memoryup to 2 TB, depending on motherboard and CPU
Storageup to 8 M.2 NVMe; rear hot-swap cages 4× 2.5" (or 4 more instead of the 4th PSU); internal 3.5" cage up to 4× 3.5" or 12× 2.5" 7 mm
Power4× CRPS modules, hot-swap; 1600 W or 2000 W per module depending on build; redundancy 4+0 / 3+1 / 2+2; at 90–140 V each module delivers about 1000 W.
Cooling capacity4.5 kW at 20 °C intake air, Max Performance mode
Cooling loopCPU block with VRM, GPU blocks with memory and VRM; 450 ml reservoir with two 18 W pumps; one 360 mm radiator; 3× 140 mm fans (6200 rpm ultra-high-flow or 5000 rpm high-flow)
Noise39–70 dB depending on profile and load
Networkup to 2× 10 GbE on board; up to 400 GbE via PCIe
OSUbuntu, Windows 11 Pro, Windows Server

4.2 Workstation (RM-M) User

Form factor4U desktop or rack; 439 × 681 × 177 mm; 45 kg net / 70 kg gross with 6 GPUs
GPUsup to 6, liquid-cooled; GeForce RTX 5090 allowed in the Workstation
Powerup to 3× SFX-L 1000 W (3 kW) or 2× ATX-L 1500/2000 W; 110–240 V, PSUs may sit on different lines
Cooling capacity2.5 kW at 20 °C intake air
Storageup to 8 M.2, 6× 2.5", 2× 3.5"
Noise39–50 dB (3× 140 mm 3000 rpm low-noise fans) or up to 70 dB with high-flow fans

4.3 Environment User

ParameterOperatingStorage / transport
Air temperature at the intake+3 to +38 °C; above 30 °C or with a failed fan, performance may drop−5 to +50 °C
Humidity10–90 % non-condensing, change rate under 20 %/h5–95 % non-condensing
Altitudeup to 2000 m; above 900 m derate 1 °C per 300 mup to 3000 m
CoolantAlarm below +3 °C and above +60 °C at the inlet or outlet (chapter 9)Do not store below −5 °C

5Unpacking, contents, keep the crate

5.1 Before you open the crate User

WARNING Check the crate for damage and the tilt indicator before signing for delivery. If there is visible damage, photograph it, note it on the carrier's paperwork and contact support before unpacking.

Two people lift the unit out by the handles. Keep the crate, the foam and the plastic: the crate is required for any warranty return (chapter 14) and a replacement crate is charged.

5.2 What is in the box, Server User

Service Manual illustration
Factory packing list for a Grando Server.
ItemQtyNote
Tested Grando Server1Serial number on the rear sticker and in the Device Passport
Case legs + M3×6 mushroom-head screws2 + 4for desk use
Rack handles + M3×4 Grando screws2 + 8for rack use
Rail set for 4U, 17.2" width + 16 fasteners + 16 M3×6 screws1 kitincluded with the Server
PSU power cablesup to 4one per installed module; plug type per order
USB flash drive with BIOS image1per device specification
Wi-Fi antenna1depends on motherboard
Dropping bottle, 250 ml1for refilling, chapter 11
Coolant: Aquacomputer Double Protect Ultra, 1000 ml1the only fluid you may add
Quick Start Guide, Device Passport, unpack leaflet1 each

Workstation (RM-M): legs, handles, up to 3 power cords, coolant bottle, documents.

5.3 First look User

  • Remove the transport sticker from the reservoir window; check the coolant level is visible in the window (chapter 11).
  • Check that nothing rattles inside; shipping may loosen a GPU support bar. If something moved, do not power on, contact support.
  • Compare the serial number with the Device Passport and the invoice.

6Site preparation & power

6.1 How much power User

BuildTypical draw under full AI loadPeakMains
Server, 8× RTX PRO 6000 (600 W each)4.1 kW measured in a fine-tuning runup to 6.4 kW with 4× 1600 W at 230 V4 feeds, 230 V recommended
Server, 4× GPUabout 2.5 kW3.2 kW2–4 feeds
Workstation, 6× GPUup to 3 kW3 kWup to 3 feeds, 110–240 V

6.2 Server power modules (CRPS) User

InputMinRatedMax
AC voltage90 V100–240 V264 V
Frequency47 Hz50–60 Hz63 Hz
Current per moduleunder 15 A at 100–127 V, under 10 A at 200–240 V, at full load
Inrush per module35 A
Efficiency80 Plus Platinum, up to 94 % at 50 % load
  • At 100–127 V (North America, Japan) each module delivers about 1000 W instead of 1600–2000 W. Redundancy drops one level: 2+2 behaves like 3+1, 3+1 like 4+0. Plan four feeds for an 8-GPU unit.
  • Breakers: C16 per module on its own branch, C25 if one branch feeds two modules.
  • Redundancy modes: 4+0, 3+1, 2+2. In 3+1 at least three modules must be online, in 2+2 at least two; below that the server shuts down. The mode is set at the factory per order.
  • Earth: every cord must connect to a protective earth; the four cords may come from different circuits.
Server power inlets
Server: where the power cords connect (from the compliance page).
Legacy units only (sticker on the rear): all modules must be energized within 10 seconds of each other, otherwise the unit beeps and will not start.

6.3 Workstation power supplies User

Up to three SFX-L units (750 or 1000 W) or two ATX-L units (1500 or 2000 W). Each accepts 90–264 V, 47–63 Hz. All installed PSUs must be powered for the unit to run; they may be on different lines. Each PSU has a fast-blow line fuse; if it blows, stop and contact support.

Workstation power inlets
Workstation: PSU inlets and switches.

6.4 Cords and plugs User

WhereCordRatingNote
Data centre PDUC13 to C1410 A, 120–240 Vfine at 230 V; at 120 V an 8-GPU Server needs four of them
Data centre PDU, high powerC19 to C2016 A, 120–240 Vpreferred for 2000 W modules
Wall, EuropeC13 to type E/F (Schuko)16 A, 230 Vone wall socket per module
Wall, UKC13 to type G13 A, 230 V
Wall, North AmericaC13 to NEMA 5-15 (type B)15 A, 120 Vfour separate 15 A circuits for an 8-GPU Server; in the review the unit ran from four ordinary outlets

Plug types A (ungrounded), C, D, I without earth are not allowed. Full plug table with pictures:

Plug type A
Type A
Plug type B
Type B
Plug type C
Type C
Plug type D
Type D
Plug type E
Type E
Plug type F
Type F
Plug type G
Type G
Plug type H
Type H
Plug type I
Type I
Plug type J
Type J
Plug type K
Type K
Plug type L
Type L
Plug type M
Type M
Plug type N
Type N
Plug type O
Type O
IEC cord
C13–C14 10 A
IEC cord
C13–C20 10 A
IEC cord
C19–C20 16 A
Types A, C, D and I without earth: not allowed. Pictures from the current FAQ..

6.5 Room User

  • Heat: everything the unit draws becomes room heat. An 8-GPU Server at full load needs about 4–6 kW of cooling in the room, or an open window in winter. No facility water is required.
  • Noise: 70 dB at full fan speed is a loud room; put the unit in a machine room or a corner, not next to a desk, unless you use the Silent profiles (chapter 9).
  • Air: keep the front intake and the rear radiator exhaust free; do not recirculate exhaust into the intake.

6.6 UPS and power quality User

  • An 8-GPU Server at full load needs an online (double-conversion) UPS of 6 kVA or more, with four outlets or a PDU; a line-interactive office UPS trips on the 35 A inrush per module.
  • Feed each power module from its own UPS outlet or circuit so that one tripped breaker only costs you redundancy, not the whole unit.
  • After a mains outage the unit returns to standby unless power-loss recovery is enabled (8.5); a UPS shutdown script on the OS plus recovery gives an unattended restart.
  • Power capping: the GPUs can be limited from the OS (for example 300 W per card as in the published reviews); the cooling controller does not need to know. The CRPS modules are not user-configurable.

7Mechanical installation

Torque values are in the Service Manual; a customer tightens by hand with the supplied screws.

7.1 Orientation rule User

NOTICE Reservoir on top (vertical) or on the left when seen from the front (horizontal). These are the only two positions. Anything else lets air into the pumps, stops cooling and voids the warranty.
Allowed placement
Allowed placement: reservoir on top (vertical) or on the left (horizontal). From the Quick Start Guide.

7.2 Rack, with rails User

  1. Mount a rack handle to each side panel with four M3×4 screws.
  2. Handle mounting
    Rack handle on the side panel.
  3. Mount the inner rail to each side panel with the fasteners and M3×6 screws from the rail kit.
  4. Mount the outer rails to the rack posts; slide the unit in until the latches close.
  5. Connect the cords last; route them so a module can be pulled without unplugging its neighbours.
Inner railsOuter rails
Inner rail on the side panel; outer rail slides onto the inner rail until the lock closes.

7.3 Rack, on a shelf User

A 19" shelf rated for 60 kg or more, 4U space, handles mounted. Do not hang the unit on the handles alone.

7.4 Desk User

Mount the two legs to the right side panel with the M3×6 mushroom-head screws, stand the unit vertically with the reservoir on top, or lay it horizontally with the reservoir on the left. Leave the vents free.

Legs mounting
Legs on the right side panel: screws and washers from the kit.

8First start, on and off, power-loss recovery

8.1 Before the first start User

  • Orientation per 7.1; transport sticker off the reservoir; coolant visible in the window.
  • One cord to every installed power module. A Server with four modules needs four cords connected, even in 2+2 mode.
  • Keyboard, display and network on the motherboard I/O panel at the front.
Service Manual illustration
Connecting AC power.
Starting up
Power cords to every module (Quick Start Guide).

8.2 What happens when power is connected User differs by generation

The controller starts on its own and runs a self-test: firmware check, supply mode detection, mains presence on every module, sensor check, temperature check. If the test passes it waits in standby with fans and pumps stopped. The self-test runs every time the controller gets power and after a reset.

Logika 2.0 · firmware 1.1.x
FIRMWARE 1.1.8-241209 SYSTEM SELF-TEST … 42%SELF-TEST PASSED COOLING SYSTEM STANDBY

Up to 1.1.4 a “COMINO COOLING SYSTEM START” splash is shown first; from 1.1.5 the firmware line replaces it.

Self-test startSelf-test passedStart hint
Logika 2.0 screens from the current FAQ: self-test, passed, start hint.
Logika 2.5 · firmware 1.0.x-L2 / 1.1.0-L2
FIRMWARE 1.1.0-L2-251019 SYSTEM SELF-TEST … 42%COOLING SYSTEM STANDBYHOLD BOTH BUTTONS TO START

The standby screen shows the reason for the last stop (since 0.0.9). The start hint depends on the panel: “hold both buttons” on a two-button unit, a one-button hint on a three-button unit (setting since 1.1.0-L2).

Facelift · three buttons
COOLING SYSTEM STANDBYPRESS START TO LAUNCH

Same Logika 2.5 firmware with the one-button hint mode.

Service Manual illustration
Self-test sequence (Logika 2.5).

8.3 Turning on User

NOTICE The unit refuses to start, and logs an error, if the motherboard is not connected to the controller. On a factory-built unit this only happens after a service operation.
  1. Hold PB1 + PB2 for 2 s; a confirmation screen appears.
  2. Power rails come up, the controller waits for POWER GOOD from the PSUs.
  3. Fans and pumps are checked against minimum speed and flow.
  4. The display shows OPERATION; the motherboard starts.
  1. Hold PB1 + PB2 for 2 s; a confirmation screen appears. Or use SYSTEM TURN ON in the commands menu (9.3).
  2. Power rails come up, the controller waits for POWER GOOD from the PSUs.
  3. Fans and pumps are checked against minimum speed and flow.
  4. The display shows COOLING SYSTEM OPERATION; the motherboard starts.

If the start fails everything is switched off, the event is logged, and the message stays until you press a button; the controller returns to standby.

  1. Hold START.
  2. Same sequence as Logika 2.5.
Service Manual illustration
Start from standby (Logika 2.5).

Check that every PSU LED is green. One amber LED means one cord is out or one module tripped (chapter 13).

8.4 Shutting down User

  1. Shut the operating system down first.
  2. Hold PB1 + PB2 for 2 s (Facelift: hold START), then press PB2 to confirm.
  3. The controller asks the motherboard to finish, then cuts the power rails. Fans and pumps stop. The display returns to STANDBY.
Logika 2.0If the motherboard does not finish, a forced shutdown follows after FALL_TIME + 60 s (firmware 1.0.24 and later). Firmware 1.1.6 fixed an early forced shutdown at 55 s with a “MOBO SYNC LOST” message.
Shutdown promptForced shutdown
Logika 2.0: shutdown confirmation and forced shutdown screens (current FAQ).
Logika 2.5If the motherboard does not finish within two minutes, a forced shutdown follows.
FaceliftSame as Logika 2.5.

To cut all power, switch off or unplug every cord. Wait for all PSU LEDs to go dark before opening any cover.

8.5 After a power cut: power-loss recovery User

By default the unit stays in standby after mains power returns. To make it start by itself, enable Power loss recovery. It is a user setting in this manual: it cannot damage the unit.

  1. In standby, hold PB2 to open the menu.
  2. Go to POWER LOSS RECOVERY, set Enabled, hold PB2 to save.
Power loss recoveryEnabled
Power loss recovery screens (Logika 2.0, current FAQ).
Logika 2.0The item lives under MENU → SERVICE → PW LOSS RECOVERY on current firmware. This manual lists it with the user commands; the menu position changes when the service items are hidden.

After an emergency stop caused by coolant or air temperature out of range, the unit does not restart until the cause is gone, even with recovery enabled. The alarm stays on (9.4).

9Cooling controller for users

9.1 Buttons User differs by generation

Service Manual illustration
Two buttons drive every menu. Short press moves, hold confirms or cancels.
ActionPB1 (PREV / BACK)PB2 (NEXT / OK)
Short pressprevious screen or itemnext screen or item
Hold over 2 sgo back, cancelenter, confirm, save
Both held 2 sswitch the system on or off; also silences a warning buzzer until the next new error
Logika 2.5Menus return to the status screens after 10 s without input, the event log after 120 s; the status screens cycle every 3 s after 5 minutes without a button press. Menu order: COMMANDS · COOLING MODE · EVENT LOG · SERVICE.
Logika 2.0Menu order: COMMANDS · SERVICE · EVENT LOG; the cooling profile is reached by holding PB2 on a status screen. No automatic return to the status screens.
Navigation
Logika 2.0 button navigation (current FAQ).
FaceliftThree buttons: up, start/confirm, down.

9.2 Status screens User differs by generation

ScreenShowsNormal range
CLNT HOT / COLD (2.0 before 1.1.8: CLNT IN / OUT)coolant temperature leaving the components (T2) and returning from the radiator (T1)+3 to +60 °C; alarm outside
AIR IN / OUTair at the intake (T3) and at the exhaust (T4)+3 to +35 °C sensor range; ambient limit +38 °C
FANS RPMeach fan channel300–3000 rpm (6200 rpm fans run higher)
PUMPS RPMpump 1 and 23500–4000 rpm
CLNT FLOWlitres per minuteonly on units with a flow sensor
UPTIMEtotal and current session hours
FIRMWAREversion and datecompare with the Device Passport
Logika 2.0Two extra screens: controller board temperature and humidity (T5). They are informational.
Logika 2.5No board-temperature or humidity screens (removed in 0.0.7); sensors are digital and need no field calibration.

9.3 Commands User

Hold PB2 to open the menu, choose COMMANDS, step to the command, hold PB2 to run it. PB1 cancels.

Commands menu
MENU: COMMANDS as shown on the display.
CommandWhat it doesAvailable in
SYSTEM TURN ONStarts the computing moduleStandby
SYSTEM SHUT DOWNAsks the OS to shut down, then cuts powerOperation
FORCED SHUT DOWNCuts power without waiting; only when the OS is frozenOperation
SYSTEM RESETHard reset of the computing moduleOperation
COOLER RESETRestarts the controller; first step for any F-codeStandby, Failure
POWER LOSS RECOVERYAuto-start after a power cut (8.5)Standby

9.4 Cooling modes (noise and performance) User differs by generation

A mode is a fan-and-pump curve against coolant temperature. The factory mode is written in the Device Passport. MENU → COOLING MODE, short press to browse, hold PB2 to apply. A mode the unit is not allowed to select right now is refused with a message on screen.

Service Manual illustration
The five modes and when each is allowed (Logika 2.5).
Profile menu
Logika 2.0: the profile line on the display (Quick Start Guide).
ModeCharacterAllowed whenReference points, coolant °C → duty %
1 · Super SilentFans stay off until 35 °C, then ramp hardintake air T3 ≤ 30 °C; reverts to Normal if T2 reaches 50 °C35→0 · 45→10 · 50→50 · 55→100
2 · SilentQuiet but always turningT3 ≤ 30 °C; setpoint T4 = 45 °C25→15 · 40→30 · 50→50 · 55→100
3 · NormalBalance of noise and card temperaturealways; setpoint T4 = 35 °C20→15 · 35→45 · 45→75 · 58→100
4 · Max PerformanceEverything at 100 %, loudest, coolestalways; factory default on Serversall points 100
5 · CustomYour own four pointshealthy T1 and T4; editable in standby onlyfactory: 10→30 · 35→50 · 45→60 · 55→100
Logika 2.0Same five modes since firmware 1.1.2 (older firmware: four, starting with Silent). Hold PB2 on a status screen to open the profile menu. Super Silent was added in 1.1.0. Curve points on 1.1.x differ slightly from the 2.5 table.
Service Manual illustration
Fan and pump duty against coolant temperature for each mode (Logika 2.5).

9.5 Warnings and codes you may see User

Every code is a letter plus three hexadecimal digits. An A-code is an alarm: the unit keeps running, the display shows the code and an intermittent buzzer sounds (0.5 Hz); hold both buttons for 2 s to silence it. An F-code is a fault: the controller stops the computing module with a continuous buzzer until the unit is powered down or the fault clears.

Service Manual illustration
Two classes of message and the code groups (Logika 2.5).
CodeMeaningWhat you can doThen
A201 / F203Coolant or intake air below minimumMove the unit to a warmer room, waitStarts by itself when in range
A211 / F212Coolant above maximum (60 °C)Check all fans turn, intake air under 38 °C, coolant level; let it coolRestarts only after the cause is gone
F213Intake air above maximumLower the room temperature or free the intakeCOOLER RESET once in range
F20B / F20CCoolant flow too low / level too lowCheck the reservoir window, top up (chapter 11)Repeats: contact support
F10A / F10BAll fans / all pumps failedCOOLER RESET onceRepeats: contact support
A101–A108One fan or pump channel not respondingNote which channelUnit keeps running; contact support
A111–A128One channel too slow or too fastNote which channelUnit keeps running; contact support
F1FFMotherboard failed to start / sync lostCOOLER RESET, then full power cycleRepeats: contact support
A001 / A003 / A004 / F002A temperature sensor faulty or out of rangeCOOLER RESET onceRepeats: contact support
F005 / F010 / F210 / F20A / F21A / FFFFController board, memory, +12 V or PSU faultNothing on siteContact support
F21HHumidity too highDry the room; keep the unit powered but in standbyStarts when dry
Logika 2.5MENU → EVENT LOG keeps up to 250 events with the uptime stamp. Hold PB2 on an entry to read the meaning in plain words. The log returns to the status screens after 120 s.
Service Manual illustration
Reading the log and the full code table (Logika 2.5).
Logika 2.0Codes as on firmware 1.1.8; in 1.1.9 the T3 sensor fault moves from A003 to F003. The level and flow codes on 1.0.23–1.1.x may differ from the 2.5 table above.

9.6 Advanced settings Advanced

Open advanced settings confirmation required
These settings change how alarms are detected. A wrong value can hide a real leak or stop the pumps. Changes are written to the event log and visible to support. Open with PB1 + PB2 held for 5 s in standby.
SettingValuesWhy you might touch itGeneration
CLNT LEVEL SENSDisabled / Enabled / InvertedFalse level alarms on some units (serials around 830–880)2.0, 2.5
CLNT FLOW LOW0–999 l/min; 0 disables the alarmFlow sensor removed or replaced2.0, 2.5
TEMP CALIBRATION (T1–T4 offset)±x.x °C per sensor, standby onlyReadings differ from a reference thermometer2.5 (menu since 0.0.5); 2.0 via USB only
CH POWER LIMIT0–100 %Cap fan and pump power2.0 only; removed in 2.5 firmware 0.0.7
CHNL CALIBRATION (NEXT START = YES)runs once at the next startAfter a fan or pump was replaced by support's instruction2.0, 2.5
Service Manual illustration
Temperature offset calibration (Logika 2.5).

9.7 Service operations Service

These exist on the controller but are not for customers. They are performed by Comino or a certified partner, remotely over USB or on site. Some void the warranty when done outside service: temperature-protection thresholds (coolant 60 °C limit), bootloader and firmware update, EEPROM erase, channel configuration and info, tacho configuration, PWR_OK timings, work source, forced work, standalone mode, single-PSU mode (removed in 1.1.9), logo colour and hint mode (2.5).

10Monitoring and remote management

10.1 What it is User

Comino Monitoring System reads the cooling controller over USB and shows temperatures, fan and pump speeds, flow, level, alarms and uptime in a web dashboard (Grafana) on the unit itself or on a second machine. It keeps a history so support can see what happened before a fault. On units shipped with the OS it is installed and on the desktop.

10.2 Current release User

VersionDatePlatformsNeeds
1.6.124 Jul 2026Windows 10/11, Ubuntu 20.04+ (amd64, arm64), RHELcontroller firmware 1.0.2-L2 (2.5) or 1.1.x (2.0); on Ubuntu nvidia-smi and lm-sensors installed; 300 MB disk
  • Downloads: Windows, Ubuntu amd64, Ubuntu arm64, RHEL.
  • 1.6: new Windows installer that antivirus no longer flags. 1.5: information panel repeats the controller display; supports Logika 2.0, 2.3, 2.5. 1.4: level sensor; fix for fans not stopping on OS shutdown with firmware 1.1.3.
Controller firmware 1.1.0–1.1.2 is not compatible with CMS 1.3 and later. Update the controller through support first.

10.3 Install and open User

  1. Windows: run the installer; Ubuntu: sudo apt install ./rm-monitor_1.6.1_amd64.deb; RHEL: sudo rpm -i rm-monitor-1.6.1-1.x86_64.rpm.
  2. Open the dashboard shortcut, or http://localhost:3000 in a browser.
  3. For several units, install on each and point one Grafana at all of them, or use the REST API (rmmonitoring/restapi, to be merged here).

10.4 What the numbers mean User

PanelSensorNormal
Coolant in / outT1, T23–60 °C; T2 a few degrees above T1 under load
Air in / outT3, T4T4 at least 5 °C above T3 under load
Fans, pumpsrpm per channelfans 300–3000 (or up to 6200), pumps 3500–4000
Flow, levelflow sensor, level switchflow above the alarm threshold; level OK
SupplyV10.8–13.2 V
GPU / CPUfrom nvidia-smi and sensorsGPU core under 70 °C under load on a healthy loop

10.5 Remote management (BMC / IPMI) User

Server motherboards used in the Grando (ASRock Rack GENOAD8X-2T/BCM, TURIND8X-2T, Gigabyte MZ73) carry a BMC with its own Ethernet port on the motherboard I/O panel. It works with the unit switched off, as long as the power modules have mains.

  • What you can do: power the motherboard on and off, open a remote console (KVM), read sensors and the system event log, update BIOS, mount an ISO.
  • What it does not do: it does not control the cooling loop. Fans, pumps, coolant alarms and the START command live on the Comino controller (chapter 9), and the controller must be in OPERATION before the BMC can power the board.
  • First connection: the BMC takes a DHCP address by default; the MAC is on a sticker on the I/O panel. Static IP and credentials are set in the BMC web interface or with ipmitool; change the default password at first login.
  • Workstation boards (WRX90 and similar) have no BMC; use the OS tools and CMS.

11Maintenance: coolant, filters, cleaning

11.1 How often and how much User

  • Look at the reservoir window once a month; it takes five seconds.
  • Typical loss is about 150 ml every 3–4 months (some units: 6 months). Top up when the level is below half the window; fill to 8/10–9/10 of the window.
  • Use only the supplied coolant (Aquacomputer Double Protect Ultra). Water, car antifreeze or other fluids damage the loop and void the warranty. The dropping bottle in the box is for this.
  • Units with a level sensor show an alarm on the display when the level reaches the minimum (9.5).
Coolant level
Coolant level window and the refill bottle (Quick Start Guide).

11.2 Checking the level User

  1. Switch the unit off and let the pumps stop.
  2. Place it in the horizontal working position (reservoir on the left).
  3. Read the level in the window on the front panel.
Reservoir window
Reservoir window on the front panel; fill when below half, up to 8/10–9/10.

11.3 Topping up User differs by chassis

NOTICE Put paper towels around the reservoir. Coolant on the hardware means a failed board. Open the cap slowly; a unit that was shipped may hold pressure.
  1. Unplug all power cords. Wait for all PSU LEDs to go dark.
  2. Open the top panel: on RM-M, two M3×4 screws on top (PH2) and three M3 screws at the rear (2 mm hex); slide the panel back and lift. RM-S and RM-L have one more M3 screw at the front.
  3. Top panel screwsLift the top panel
    Top panel: screws on top and rear, then slide back and lift (current FAQ).
  4. Unscrew the reservoir cap with a 6 mm hex key. If a level sensor is fitted instead of the cap, unscrew it by hand.
  5. Add coolant with the dropping bottle until the level is at 8/10–9/10 of the window. Do not overfill; the loop needs an air cushion.
  6. Screw the cap back on hand-tight, put the panel back, reconnect the cords.
  7. Run the unit for 15 minutes vertically, then check the level horizontally again; the loop may release trapped air. Repeat once if the level dropped.
Service Manual illustration
After filling: run, re-check the level, repeat if needed (Service Manual).
FaceliftReservoir access through the front without removing the top panel.

11.4 Cleaning User

  • Front mesh and radiator: vacuum or soft brush every 3 months in a dusty room, every 6 months otherwise. Never blow compressed air into the radiator from the rear at close range, it bends the fins.
  • Exterior: dry cloth. No solvents on the display window.
  • Inside: only with power removed; do not touch the hoses and quick disconnects.

11.5 Draining the loop Service

Draining (bottom G1/4" plug) is a service operation before shipping or a GPU swap at the factory. Customers do not drain the loop; for shipping, see chapter 14.

12Customer-permitted upgrades

Which operations keep the warranty is marked; everything else goes through support.

12.1 What you may do yourself User

OperationWarrantyChapter
Add or replace drives in the rear hot-swap cages (Server)kept12.2
Add drives in the internal 3.5" / 2.5" bayskept, power off12.3
Replace a hot-swap power module (Server)kept12.4
Add an air-cooled PCIe card (network, storage)kept if the GPU support bar is refitted12.5
Add or replace a liquid-cooled GPU with a Comino blockwith support's written OK; quick disconnects make it tool-free12.6
Replace CPU, RAM, motherboard; open the loop; change the controllerservice onlychapter 16

12.1a Tools, screws and ESD User

ToolFor
PH2 screwdrivertop-panel screws (M3×4), legs (M3×6 mushroom head), rail fasteners (M3×6 countersunk)
2 mm hex keyrear screws of the top panel
6 mm hex keyreservoir cap
Grando M3×4 screws (8 in the box)rack handles

Work on an unpainted metal surface and touch the chassis before touching a card; a wrist strap is better. Keep cards in their bags until the moment you install them. No thread-locker is needed on anything a customer touches; torque wrenches are for the factory.

12.2 Rear hot-swap drives, Server User

Up to four 2.5" SATA or NVMe drives (7 mm, or two 15 mm) in the cages next to the power modules; four more can replace the fourth module on request. Pull the carrier by the latch, fit the drive with the four screws, slide back until it clicks. Hot-swap is supported by the backplane; check your OS and RAID setup before pulling a live drive.

12.3 Internal drives User

  1. Power off, unplug all cords, open the top panel (11.3) or the bottom panel.
  2. Side bay: the holder on the left panel takes 3.5" HDDs on brackets (8 screws each) or 2.5" drives on the sub-assembly.
  3. Bottom bay (RM-M): drives sit on two brackets, four M3×4 screws each; RM-S and RM-L have a different panel.
  4. Use the supplied screws; apply thread-locker only if support asks you to.
Bottom drive bayDrive brackets
Bottom 2.5" bay on RM-M: panel screws and the two drive brackets.

12.4 Power module hot-swap, Server User

  1. Unplug the cord of the failed module (amber LED).
  2. Hold the blue lever, pull the module out by its handle.
  3. Slide the new module in until it locks, plug the cord in, check the LED turns green.
Module replacement
Pull the blue lever and slide the module out (current FAQ).

The remaining modules carry the load meanwhile; in 3+1 the unit keeps running with one module out, in 2+2 with two.

12.5 PCIe cards User

Air-cooled cards go in free slots after the GPUs. GPUs are installed from the last slot backwards at the factory so that free slots start at slot 0. The GPU support bar must be refitted after any card change: aluminium bar with supports and screws, or the steel bar rotated across the side panels. Video: aluminium bar, steel bar.

12.6 Liquid-cooled GPU via quick disconnects Advanced

Open the GPU swap procedure support's OK first
Comino water blocks fit specific cards. A card without a Comino block cannot go into the loop. Ask support for the block and the OK before buying a card.
  1. Power off, unplug, open the top panel, remove the GPU support bar.
  2. Press the two quick disconnects on the block; they close dripless. Keep a towel underneath.
  3. Release the slot latch, lift the card out.
  4. Fit the new card with its block, reconnect the two couplings until they click, refit the bar.
  5. Power on, watch the flow and the GPU temperature in CMS for ten minutes.
Aluminium support barSteel support bar
GPU support bar, aluminium (supports + rod) and steel (rotated across the side panels). Refit after any card change.

13Troubleshooting

13.1 Before you call User

Have ready: serial number (rear sticker, Device Passport), controller firmware version (status screen), the code on the display, and if you can, a photo of the display and a CMS export (chapter 10).

13.1a Visual inspection first User

At startupDuring operation
Every PSU LED green; display shows the self-test and then STANDBY; no beeping after the self-testCoolant level visible in the window; no coolant smell or drops under the unit; fans audible and steady, no rattling
Reservoir on top or on the left; nothing blocking the front intakeDisplay cycling the status screens without a code; CMS shows coolant out a few degrees above coolant in
Cords seated, no amber LED after plugging all cordsNo PSU LED blinking amber; exhaust air warm, not hot

13.2 The five most common cases User

SymptomLikely causeWhat to do
Continuous beeping right after connecting power, unit does not startLegacy unit: modules energized more than 10 s apart; or a cord missingUnplug all cords, wait 30 s, plug all in within 10 s. Check every PSU LED is green.
“Coolant level low” or “flow low” alarm, but the window shows coolantLevel sensor false alarm (some serials), trapped air after transportRun the unit 15 minutes vertically, then horizontally; silence the buzzer; if it repeats, support may set the sensor to Disabled (9.6).
Unit stopped with an F-code and nobody is on siteCritical error latchedPower-cycle all cords remotely (PDU) to reset the controller; it returns to standby. Enable power-loss recovery (8.5) to restart automatically.
Display dark, PSU LEDs greenController not powered, display cableFull power cycle; if still dark, support.
Coolant on the floor or insideLeakUnplug everything. Do not restart. Photograph, contact support the same day; leaks are covered by the warranty (chapter 14).

13.3 Operating system does not see all GPUs User

  • Check the GPU support bar is fitted and the cards are seated; shipping can unseat a card.
  • Check the RTX PRO 6000 mode: compute or graphics (display) mode is switched with NVIDIA's display-mode tool; a card in the wrong mode is missing for some workloads.
  • PCIe AER messages in the log with an unstable card: see support; the known case involves the rear U.2 riser cabling.

13.4 When to stop User

Any F-code that repeats after one COOLER RESET, any smell, any coolant outside the loop: switch off, unplug, call support. Do not open the loop, do not change thresholds.

14Warranty, RMA, returns, packaging

14.1 Terms User

  • 12 months from delivery against defects in material and workmanship.
  • Extended warranty up to 5 years is available at purchase.
  • The warranty covers the liquid loop including leaks; if a leak damages the unit, the unit is replaced. Standard response within one working day.
  • Obvious defects must be reported in writing within 14 days of receipt.
  • Void: external force, improper maintenance (including non-Comino coolant), operation outside the limits in chapter 4, changed protection thresholds, opening the loop without support's OK.

Full terms: comino.com/en/terms.

14.2 How a return works User

  1. Contact support first (chapter 16). Never ship without an RMA number.
  2. Support may ask for remote access to run the test suite (CATS) before deciding on a return.
  3. If a return is needed you receive an RMA number, the address and the shipping arrangement.
  4. Pack the unit in the original crate (14.3). A replacement crate is charged.

US customers who bought through Puget Systems contact Puget first; Puget handles the case and involves Comino.

14.3 Packing for transport User

Service Manual illustration
Factory packing: foam cradle, bag, crate (Service Manual).
  1. Shut down, unplug, let the unit cool. Do not drain the loop.
  2. Remove legs or handles if they were fitted; pack them separately in the crate.
  3. Put the unit in the plastic bag, into the foam cradle, reservoir on top. Add the accessory box.
  4. Close the crate, strap it, apply the tilt indicator if you have one. Ship upright only.

A unit shipped on its side or without the cradle arrives with a damaged loop. Keep the crate for the whole warranty period.

14.4 Your data when the unit leaves your site User

  • RAM and GPU memory are volatile: nothing survives a power-off.
  • Drives keep data. Before an RMA or a loan return, remove the drives or sanitize them (NVMe secure erase / nvme format, or full-disk encryption with the key destroyed). Comino does not wipe customer drives and does not need them to test the unit.
  • The cooling controller stores only its configuration and the event log (codes, uptime stamps), no customer data; the BMC stores its network settings and users. Reset the BMC to factory defaults if the unit changes hands.
  • Forensic and regulated customers: ask support for the written memory-volatility statement for your audit file.

15Disposal

15.1 Electronics User

The unit is marked with the crossed-out bin: do not put it in household waste. Hand it to a WEEE collection point, or return it to Comino for recycling. Packaging, broken parts and the old unit contain materials that are reused.

15.2 Coolant User

The coolant is a propylene-glycol mixture. Do not pour it into drains. Collect it in a sealed container and give it to a hazardous-waste point as glycol antifreeze.

16Support & partners

16.1 Contact User

Riga, Latvia, working days 09:00–18:00 EET. Remote diagnostics: support can read the controller and run the test suite over a remote session; have the unit online and the serial number ready.

16.2 Partners User

  • North America: Puget Systems, Auburn WA, performs final integration, testing and support for units sold there.
  • Other certified partners service units under the Comino partner program; their contacts come with your invoice.

16.3 Keep this manual current User

This page is the live version. The “What changed” block at the top lists firmware and hardware changes; the Device Passport tells you which ones apply to your unit.