This guide documents the process of building a Dell PowerEdge R940 from an empty chassis. We’ll cover the processors, memory, processor expansion module, storage backplane, cabling, cooling, risers, network adapter, RAID/HBA controller, and power supplies required to turn a bare chassis into a functioning server.

Before you begin assembling an R940, it is important to identify the hardware required to turn the chassis into a functioning server. This section covers the major components that should be installed and verified before moving on to the rest of the build.

The goal is to start with a bare chassis and make sure you have everything necessary to power the system on, recognize the installed hardware, and eventually install an operating system.

Motherboard

The motherboard should already be installed in the chassis and securely fastened in place. It serves as the central connection point for nearly all of the server’s hardware. The R940 processor expansion module or mezzanine will also plug in to the motherboard through a set of 2 or 4 cables depending on the configuration.

The R940 is a four-socket 4U server, so the system board provides connections for up to four Intel Xeon Scalable processors along with their associated memory. 2 of the processors install directly on the motherboard. The other 2 will install onto the expansion module.

Before installing additional components, make sure the motherboard is properly seated and secured inside the chassis. A loose or incorrectly installed system board can prevent the server from powering on or cause other hardware to malfunction.

Processor expansion module

If you will be installing all 4 processors you will need the processor expansion module. This part may also be referred to as the mezzanine. The module will expand the system to support an additional 2 processors. This is a required part if you want to enable your system with the maximum number of CPUs. You can configure the R940 as 2 processor system. This will require a different set of cables then what is required to enable a system with 4 processors.

This expansion module requires 4 of these cables. These are called UPI cables. They connect to the primary motherboard.

The last cable required by the PEM is the clock cable. This looks like a mini SAS HD cable but it functions more like a signal cable for the multi-CPU architecture of the R940 rather than a storage cable.

Backplane and Cables

The storage backplane is another critical component of the R940. It provides the connections that allow the front-mounted drives to communicate with the server’s storage controller while also supplying power to the drive bays.

Dell manufactured multiple variations of this server. This is the backplane for the 8 bay SFF model. 

The backplane will have 2 cables that connect it to the motherboard. One of these cables is a signal cable. The other is a power cable.

If you’re installing a RAID or HBA controller you will also require the following cable. This is the PCIe SAS storage cable that connects the backplane to the storage controller. It is specific to the 8 Bay SFF backplane.

Before proceeding with the build, inspect the backplane and confirm that all required signal, power, and storage-controller connections are present and properly seated. These connections are essential for the server to detect and communicate with the installed drives.

Cooling

The R940 uses a combination of hot-swappable fan modules, processor heatsinks, and an air shroud to maintain proper operating temperatures throughout the system. These components work together to move air through the 4U chassis and remove heat from the processors, memory, storage components, and PCIe hardware. The fans install into the following assembly cage. 

The fan modules are a particularly important part of the R940’s cooling system because the server can contain up to four processors and a large number of expansion and storage components. The correct fan configuration therefore depends on the hardware installed in the system.

Unlike a smaller 1U or 2U PowerEdge server, the R940’s cooling requirements need to be considered alongside the processor configuration, memory population, PCIe cards, storage configuration, and other installed hardware.

The processors also require compatible heatsinks. Because the R940 supports up to four processors, a fully populated four-CPU configuration requires four processor heatsinks.

The appropriate heatsink should be matched to the processor and system configuration. Each heatsink must be properly secured to maintain adequate contact with the processor and allow heat to be transferred into the chassis airflow.

The final major component of the R940’s cooling system is the air shroud. Positioned over the motherboard, the shroud helps channel airflow from the fan modules through the areas containing the processors, memory, and other heat-producing components.

Operating the server without the appropriate air shroud can disrupt the intended airflow pattern and reduce cooling performance.

The fans, processor heatsinks, and air shroud all work together as part of the R940’s thermal management system. Before powering on a server assembled from an empty chassis, verify that the appropriate fan modules and heatsinks are installed and that the air shroud is securely in place.

Risers

The R940 has a substantially more extensive PCIe expansion system than the R740 because of its larger 4U chassis and four-processor design. This means many of your PCIe devices will plug in directly to the motherboard rather than the risers. If you have the processor expansion module installed the system does support up to 2 risers assuming you’re running a 4 CPU configuration.

There is a left riser and a right riser. These plug into the processor expansion module and add an additional 6 PCIe x16 slots to the system. These risers are connected to CPU3 and CPU4 so please ensure you have those processors installed onto the expansion module.

Please note the riser links above are technically part numbers for the PCB. Also required are the left riser assembly cage and right assembly cage. The PCB attaches to these metal cages. These usually come together as a pair on the second hand market but the distinction is worth noting.

Building the server

Once you have all the required components you can begin assembling the server. Aside from the server components there are a number of tools you will need to complete the assembly.

Required tools:

• Torque bit #30
• Phillips #2
• Compressed air
• Thermal paste

Installing CPU 1 and 2

Remove the lid of the R940 to expose the inner components. Your server may have the processor expansion module already installed. If this is case, first remove the risers. Then pull up on the handle and lift the expansion module into an upright position. This will allow you to access processors 1 and 2 located beneath the module. You can also remove the expansion module entirely.

You will also need to remove the fan assembly cage:

Remove the heatsinks to expose the CPU socket. Take this opportunity to inspect the CPU pins for damage and to blow out any dust or debris with an electric blower or compressed air.

Align the arrow on the corner of the CPU with the arrow on the motherboard and carefully place the CPU in the socket. You can also install the CPU into the heatsink first. I prefer this method. Often times the black plastic CPU retaining brackets become brittle from years of heat exposure. When you flip the heatsink over to install it over the socket, the CPU can fall out and damage the pins.

Apply thermal paste:

Finally, snap on the heatsink and tighten down the T30 screws.

Installing the memory

Each CPU in the PowerEdge 940 has 12 memory slots. The population rules are similar to other Dell servers. The main idea is spreading the memory across as many channels as possible and mirroring the configuration across CPUs. Always populate the white slots first. Each white slot represents a channel. Only after populating the white slots can you begin populating the black slots which represent the second DIMM for the channel. If you populate A1 it’s best practice to populate B1. Because this is a 4 CPU system you will also have slots labeled C and D. These will be located on the processor expansion module.

R940 memory population chart

DIMMs per CPU	CPU 1	CPU 2	CPU 3	CPU 4	Total (4 CPUs)
1 DIMM	A1	B1	C1	D1	4 DIMMs
2 DIMMs	A1, A2	B1, B2	C1, C2	D1, D2	8 DIMMs
4 DIMMs	A1, A2, A4, A5	B1, B2, B4, B5	C1, C2, C4, C5	D1, D2, D4, D5	16 DIMMs
6 DIMMs	A1, A2, A3, A4, A5, A6	B1, B2, B3, B4, B5, B6	C1, C2, C3, C4, C5, C6	D1, D2, D3, D4, D5, D6	24 DIMMs
8 DIMMs	A1, A2, A4, A5, A7, A8, A10, A11	B1, B2, B4, B5, B7, B8, B10, B11	C1, C2, C4, C5, C7, C8, C10, C11	D1, D2, D4, D5, D7, D8, D10, D11	32 DIMMs
12 DIMMs	A1–A12	B1–B12	C1–C12	D1–D12	48 DIMMs

R940 channel layout

Channel	CPU 1	CPU 2	CPU 3	CPU 4
Channel 0	A1 / A7	B1 / B7	C1 / C7	D1 / D7
Channel 1	A2 / A8	B2 / B8	C2 / C8	D2 / D8
Channel 2	A3 / A9	B3 / B9	C3 / C9	D3 / D9
Channel 3	A4 / A10	B4 / B10	C4 / C10	D4 / D10
Channel 4	A5 / A11	B5 / B11	C5 / C11	D5 / D11
Channel 5	A6 / A12	B6 / B12	C6 / C12	D6 / D12

Installing the NDC

On the R940 the NDC first installs into this bracket. The bracket (also called the rNDC bracket) then interfaces with the mezzanine port on the motherboard.

Installing a RAID/HBA controller

The RAID or HBA controller will also install in the rear of the server directly in a PCIe slot. Be sure you have the correct SAS cable. This cable will route from the 8 bay SFF backplane, to the left side of the chassis, and finally to the rear expansion slots. You will need a full height bracket for any PCIe card you have.

This is the required cable. One end will plug into the backplane. The other end goes to the storage controller.

The rear expansion ports are locked in by a black plastic mechanism. Slide the mechanism and then lift. Once the card is installed lock the mechanism back into place.

Installing the processor expansion module

Now it’s time to install the mezzanine for processors 3 and 4. Because it blocks the installation path for the other components it is best to install this piece last. Populate the processors and memory in the exact same way you did for processors 1 and 2. You can do this when the module is out of the server or you can install the module first and then put in the processors and memory.

Before you install the module you should plug in the required UPI cables. Installing the module first will make plugging these cables in more difficult than it needs to be. There are 4 in total. The connection only goes in one way so make sure you orient them properly. On the motherboard you will see connectors for UPI cables A, B, C, and D.

In the image below you will see the 4 UPI cables. Also make note of the clock cable. 

With the clock and UPI cables plugged in it’s time to install the processor expansion module.

Lift the module using the handle on the front and guide it down to its installation point. On the rear corners of the module you will see the hinging mechanism. This latches on to the chassis.

At the bottom make sure you have the following component installed. When rotated down into position the expansion board comes into contact with these 2 small chips:

Make sure the expansion board can easily hinge up and down. You shouldn’t encounter any resistance except from the weight of the board itself.

Now plug in the UPI cables that you previously attached to the motherboard. They are assigned letters. Simply match up the letters. You will notice the cables cross over each other. It is helpful to have the expansion board in its fully upright position while plugging in the cables. 

Once you have the cables plugged in carefully bring the expansion board down to its natural sitting position.

Installing the risers

One of the last components to install are the risers. These plug in to the processor expansion module. There is a left riser and a right riser. Both have a gear-like mechanism that pulls them up or down into position and securely connects them to the riser connector.

Installing power supplies

Last and certainly not least are the power supplies. They slide in through the rear of the chassis like most PowerEdge servers.

Final picture

Here is the complete build. I do believe there is an air shroud that can be installed but I didn’t have one on hand for this particular server.

 

The HPE ProLiant DL360 Gen10 Premium 10-Bay SFF is a unique configuration equipped with an NVMe backplane that provides NVMe support across all 10 front drive bays. With the installation of a Smart Array controller, the first 8 drive bays become “universal,” supporting SAS, SATA, or NVMe drives, while bays 9 and 10 remain NVMe-only.

In this guide, we’ll go through a complete build of this server, covering everything from installing the processors and memory to the RAID controller, Smart Array battery, network daughter card, risers, PCIe cards, storage and NVMe cabling, and power supplies.

Essential Components

Before you begin assembling an HPE ProLiant DL360 Gen10 10-Bay SFF Premium server from an empty chassis, it is important to identify the hardware required to turn the chassis into a functioning server. This configuration is different from the standard 8-bay DL360 Gen10 because it uses a 10-SFF premium backplane and dedicated NVMe hardware to provide a combination of SAS, SATA, and NVMe storage support.

The goal of this section is to identify the major components that should be installed and verified before moving on to the rest of the build. Starting with the bare chassis, you will need to install and configure the motherboard, processors, memory, storage backplane and cabling, cooling system, Smart Array controller, network daughter card, PCIe risers, expansion cards, and power supplies.

Motherboard

The motherboard should already be installed in the chassis and securely fastened in place. It serves as the central connection point for the server’s processors, memory, storage, expansion cards, networking hardware, cooling system, and power delivery.

The processors and memory modules are installed directly onto the system board. PCIe risers connect to the motherboard and provide expansion slots for network adapters, storage controllers, and other PCIe devices. The front storage backplane connects to the appropriate storage controller and NVMe hardware through dedicated cabling.

The DL360 Gen10 also uses a network daughter card, which installs directly into the system board and provides the server’s integrated network interfaces. The hot-plug fan modules connect to the motherboard through dedicated fan connectors.

Before installing the remaining components, make sure the motherboard is properly seated and secured inside the chassis. A loose or incorrectly installed system board can prevent the server from powering on or cause individual components to malfunction.

Backplane and Cables

The 10-SFF premium backplane is one of the most important components in this particular DL360 Gen10 configuration. HPE identifies the 10-SFF NVMe/SAS/SATA backplane as spare part 875556-001.

This backplane provides the physical connections for all ten front drive bays. Unlike a standard 8-SFF SAS/SATA configuration, the premium configuration is designed to provide both traditional SAS/SATA storage and direct NVMe connectivity.

The ten drive bays are divided into two functional groups. Bays 1 through 8 can support a mixture of NVMe and SAS drives, while bays 9 and 10 are NVMe-only when the 10-SFF NVMe/SAS backplane option is installed.

This makes the first eight bays effectively universal from a drive-type perspective, provided the appropriate storage controller and NVMe connections are installed. The final two bays are dedicated to NVMe storage.

The backplane requires several different cable connections depending on the storage configuration. HPE’s cabling documentation identifies separate connections between the backplane, the Smart Array controller, and the NVMe riser. For the 10-SFF configuration, HPE lists dedicated cables for the NVMe connections (see below) as well as a separate SAS cable for the connection to the Smart Array controller.

For the premium configuration, particular attention should be paid to the individual cable types and their port assignments. The cables can look similar, but each cable is designed for a specific connection between the backplane, controller, or NVMe riser.

There are 3 different cables you’ll need to connect the NVME backplane to the NVME riser.

Click to see picture of NVME cable for Port 1

Click to see picture of NVME cable for Port 2

Click to see picture of NVME cable for ports 3, 4, and 5

The Smart Array controller provides the SAS/SATA storage connection, while the NVMe drives use dedicated PCIe connectivity. HPE’s DL360 Gen10 cabling documentation specifically includes configurations for a 10-SFF NVMe backplane connected to an NVMe riser and for the 10-SFF premium backplane connected to the P824i-p Smart Array controller.

Before proceeding with the rest of the build, inspect the backplane and confirm that the required backplane power cable, SAS, and NVMe connections are present. Verify each cable against its intended port before installation.

Correct cabling is particularly important on this configuration. A drive can be physically installed in a bay but still remain unavailable to the operating system if the corresponding storage controller or NVMe PCIe connection has not been properly connected.

For a complete 10-bay configuration, the storage system should therefore be treated as two interconnected subsystems:

• SAS/SATA storage — connected through the Smart Array controller.
• NVMe storage — connected through the dedicated NVMe PCIe/riser connections.

The exact cable arrangement depends on the controller and NVMe configuration being built, so the individual cable part numbers and port assignments should be verified before installation.

Cooling

The DL360 Gen10 uses a combination of hot-plug fan modules, processor heatsinks, and an air baffle to maintain proper operating temperatures throughout the 1U chassis.

The fan modules are individually installed into the chassis and connect to the system board through dedicated fan connections. HPE identifies the DL360 Gen10 fan modules as hot-plug components, allowing individual fan modules to be replaced without removing the entire server from service.

The processor heatsink mounts directly above each Intel Xeon Scalable processor and transfers heat away from the CPU into the airflow generated by the fan modules. The air baffle directs airflow through the processor, memory, and other internal components.

The DL360 Gen10 has different cooling configurations depending on the installed processors and hardware. For higher-power processors and configurations with increased thermal requirements, HPE offers the DL360 Gen10 High Performance Heat Sink Kit (873591-001) and High Performance Fan Kit (871244-B21).

For a 10-bay NVMe configuration, cooling should be treated as an important part of the overall build rather than an optional finishing component. NVMe drives, high-power processors, storage controllers, and PCIe expansion cards can all contribute to the thermal load inside the 1U chassis.

Before powering on the completed server, make sure the appropriate fan modules, processor heatsinks, and air baffle are installed and that all required fan connections are properly seated.

Risers

The DL360 Gen10 uses several different PCIe riser configurations depending on the intended hardware. The risers provide expansion slots for Smart Array controllers, network adapters, NVMe hardware, GPUs, and other PCIe devices.

The 10-SFF NVMe configuration is particularly important because it uses a dedicated NVMe riser to provide the PCIe connectivity required by the front NVMe drives. HPE’s QuickSpecs identifies a 5-port 10-SFF NVMe Riser Kit, which provides connections for up to ten NVMe SSDs. This riser is included with certain 10-SFF chassis configurations and the 10-SFF Premium Backplane Kit.

The DL360 Gen10 also supports other riser configurations, including an x16/x8 primary riser, a low-profile x16 riser, and other PCIe configurations. The correct riser depends on the chassis configuration and the expansion hardware being installed.

When building a 10-bay premium server from an empty chassis, it is therefore important not to assume that any DL360 Gen10 riser will work with the configuration. The NVMe riser and its associated cabling are specifically required for the front NVMe storage connections.

The riser board installs into its corresponding riser cage and connects to dedicated PCIe connectors on the system board. Before installing an expansion card, verify that the riser provides the required PCIe slot type, lane width, physical height, and electrical connectivity.

For this particular build, the riser configuration should be established before installing the Smart Array controller, NVMe cabling, and other PCIe expansion cards. This helps ensure that each card is installed in the correct position and that there is sufficient room for the required cabling.

Building the server

Assuming you have all the required components listed above we can begin adding components.

Before starting, also ensure you have the following tools:

• T30 bit
• T20 bit
• T10 bit
• Phillips #2 bit
• Compressed air
• Thermal paste

Installing the processors

Open the lid to expose the inner components and remove the heatsinks or CPU covers. It is first necessary to inspect the pins prior to installing the other components. If for whatever reason the CPU pins are damaged you should highly consider not continuing the build, especially if the server will be going into any serious production environment.

Remove the heatsink to expose the CPU socket. Blow out any dust or debris with the compressed air:

Place CPU in the socket. If you prefer you can install the processor on the heatsink and then just install the heatsink. I have found sometimes the CPU can slip out of the black plastic bracket and damage the CPU pins, so I like to install the CPU into the socket first, and then install the heatsink next. Be sure to orient the arrow on the CPU with the arrow on the motherboard.

Apply thermal paste:

Install the heatsink. You will need the T30 bit for this:

Installing the memory

For this build I have 8 DIMMs in total. I’ll assign 4 to CPU1 and 4 to CPU2. In general you want to populate the white DIMMs first before populating the black dimms. Because I have 4 DIMMs per CPU I will populate DIMM slots 3, 5, 8, and 10. If you have a different amount consult the memory population guidelines below.

Click here for memory population guidelines

Number of DIMM(s) to populate per cpu	DL 360 G10 memory population guidelines
	Ch 6		Ch 5		Ch 4		Ch 1		Ch 2		Ch 3
1								8
2								8		10
3								8		10		12
4			3		5			8		10
5			3		5			8		10		12
6	1		3		5			8		10		12
7	1		3		5		7	8		10		12
8			3	4	5	6	7	8	9	10
9	1		3		5		7	8	9	10	11	12
10	1		3	4	5	6	7	8	9	10		12
11	1		3	4	5	6	7	8	9	10	11	12
12	1	2	3	4	5	6	7	8	9	10	11	12

Installing the smart array battery

At this point I like to install the smart array battery. This is the battery that maintains the RAID controller cache in the event of a power outage. Start by removing any cables attached to the backplane. If your server already has the full set of NVME and SAS cables installed, these have to be removed in order to install the battery. You will also need to remove the fans (at least the center fan.)

Installing the Smart Array Raid controller

The smart array controller is what will allow the server to recognize SAS and SATA drives. The NVME cables alone will only support NVME drives. In order for the drive bays to be truly universal (supporting SAS, SATA, and NVME) you must have both the smart array controller and NVME cables installed.

It is best to install the Smart Array controller with the cables removed from the system. The controller will sit under the cables and it’s a lot easier to install prior to NVME cable installation. It is also helpful to remove the risers although not completely necessary.

This controller has a proprietary connection to the motherboard as seen here:

There are 2 holes in the controller. Line them up with the blue pegs on the board. Tighten down the screws.

Plug in the Mini SAS cables if you have the cable installed already. We’ll cover installing the cables in the next section.

NVME and SAS cabling

Now it’s time to install the cables. The first cable you want to install is the blue Mini SAS cable shown above. This attaches from the backplane and then to the smart array controller.

This is the installation location on the backplane. Route it along the right side of the chassis. It will make a 90 degree left turn before the riser and plug in to the smart array controller.

Now that the mini SAS cable is plugged in it’s time to install the NVME cables. These cables will plug into the backplane and route to the rear of the server where they will be plugged into the dedicated NVME riser. Cables that plug in to ports 1 and 2 on the backplane will route to the right side of the server. NVME cables port 3, 4, and 5 will route to the left.

NVME cables port 1 and 2:

Here you can see how the cables route along the edge of the chassis:

NVME cables port 3, 4, and 5

Now for the last set of cables. These install into their respective ports on the backplane like ports 1 and 2. These cables however will route to the left side of the chassis.

Installing NVME riser and plugging in cables

All of these cables now plug in to the NVME riser.

Installing NDC

The NDC will install below riser 1. Remove riser 1 and remove any blank slot fillers prior to installation of NDC.

Installing PCIe card

With the riser removed it’s time to install any PCIe cards you have. The primary riser has a full height and half height slot available. Here I will install a card in the full height slot:

Installing power supplies

The very last thing to do is install the power supplies.

Cleaning up

Prior to powering on for the first time ensure everything is in its place. If you removed the fans during installation reinstall them. Ensure all cables are properly routed and neatly tucked in to the side channels. If they’re not, it may be difficult getting the lid back on. If you’ve determined that everything is in order, plug in the power supplies and power the server for the first time.

The DL325 G10 is a 1U single-socket server from HP. The server is compatible with AMD EPYC 7001 and 7002 series processors. In this guide, we’ll cover the installation of the core components, including the CPU, memory, and power supplies, as well as the cabling configurations required to enable full eight-bay NVMe support.

Essential Components

Before you begin assembling a DL325 Gen10 from an empty chassis, it is important to identify the hardware required to turn the chassis into a functioning server. This section covers the major components that should be installed and verified before moving on to the rest of the build.

The goal is to start with a bare chassis and make sure you have everything necessary to power the system on, recognize the installed hardware, and eventually install an operating system.

Motherboard

The motherboard should already be installed in the chassis and securely fastened in place. It serves as the central connection point for the server’s major hardware components.

The processor and memory modules are installed directly onto the motherboard, while the power supplies connect to it through the rear of the chassis. PCIe risers attach to the motherboard to provide expansion slots, and the front storage backplane connects to the system board and storage controllers. The hot-swap fan assembly also connects to the motherboard.

Before installing additional components, make sure the motherboard is properly seated and secured inside the chassis. A loose or incorrectly installed system board can prevent the server from powering on or cause other hardware to malfunction.

Backplane and Cables

The front NVME storage backplane is a critical component of the DL325 Gen10 8-bay configuration. It provides the connections required for the front-mounted NVME drives to communicate with the system while also supplying power to the drive bays.

The backplane requires several connections to operate correctly. The 8-bay NVMe configuration requires specific PCIe cabling between the backplane and the motherboard or PCIe expansion hardware.

For an 8-bay NVMe configuration, particular attention should be paid to the PCIe cables and their individual port assignments. The DL325 Gen10 uses multiple cable assemblies to connect the NVMe backplane to the system’s PCIe resources, and the cables must be connected to the correct ports for all eight drive bays to function.

The physical appearance of some of these cables can be similar, so it is important to verify the correct HPE part number and intended connection before installing them. A cable that physically fits a connector is not necessarily the correct cable for a particular backplane or port.

Before proceeding with the build, inspect the backplane and confirm that all required signal, PCIe, and power connections are present and properly seated. The 8 NVME configuration doesn’t have the traditional signal cable connecting the backplane to the motherboard. Instead, all signal cables are technically the NVME cables. You will however need the following power cable. 

Correct cabling is essential for the server to detect and communicate with the NVMe drives installed in the front bays.

To enable NVME support on all 8 front drive bays will require the following 4 cables:

Click to see NVME Cable Port 1

View the full part number here

Click to see NVME Cable Port 2

View the full part number here

Click to see NVME Cable Port 3

View the full part number here

Click to see NVME Cable Port 4

View the full part number here

Cooling

The DL325 Gen10 uses a combination of fan modules, a processor heatsink, and an air shroud to maintain proper operating temperatures throughout the system. These components work together to move air through the chassis and remove heat from the processor, memory, and other internal hardware.

The fan modules are individually installed into the chassis and connect to the system board using dedicated fan cables. HPE identifies these as hot-plug fans, allowing an individual fan to be replaced while the server is running. The replacement procedure involves seating the fan in its guide pins and reconnecting its cable to the system board.

The DL325 Gen10 uses different fan configurations depending on the chassis and cooling requirements. HPE lists 878537-001 as a standard fan module and P06973-001 as another supported fan module. The high-performance dual-rotor P04998-001 is recommended for NVME configurations.

The processor heatsink mounts directly above the AMD EPYC processor and transfers heat away from the CPU into the airflow generated by the fans. The air baffle directs airflow through the processor and memory areas, helping ensure that cooling air passes over the components that require it. HPE lists the DL325 Gen10 air baffle as P07015-001. For an 8-bay NVMe build, the cooling system should be configured with the appropriate high-performance fans, processor heatsink, and air baffle before the server is placed into service.

Risers

The DL325 Gen10 supports a primary PCIe riser and an optional secondary low-profile riser, providing additional PCIe expansion slots for network adapters, storage controllers, GPUs, and other expansion cards. HPE’s official parts documentation lists P06514-001 as the primary riser board and P07016-001 as the secondary low-profile riser.

The primary riser is installed in the main riser cage and provides two PCIe expansion slots. The riser uses one PCIe x16 slot and one PCIe x8 slot, giving the DL325 Gen10 additional expansion capability while retaining the compact 1U form factor. Keep in mind the riser cage and the actual PCB board are 2 separate part numbers. However, when buying second hand they are usually sold as a complete set and listed as part number P04969-001.

The secondary riser is an optional low-profile riser that provides an additional PCIe slot. HPE lists the DL325 Gen10 PCIe low-profile riser kit as P04849-B21, with P07016-001 being the corresponding spare-part number for the riser board.

The riser boards install into dedicated connectors on the system board and are housed in their respective riser cages. When building a DL325 Gen10 from an empty chassis, make sure the appropriate riser cage, riser board, and expansion-slot blanks are installed for the desired configuration.

Building the server

Assuming you have verified the presence of all essential components it’s time to build the server. I always start with the CPU for a number of reasons. The first reason is to inspect the CPU pins. If the CPU pins are in any way damaged I’m not going to continue adding components because I don’t want to put a damaged server into production. Bent or smashed CPU pins can cause immediate problems with memory and PCIe devices.

For this build you will need the following tools:

• T20 bit
• Phillips #2
• Compressed air
• Thermal paste

Installing the CPU

There are a few layers of the CPU assembly that sit between you and the CPU socket. The first obvious thing in your way is the heatsink. The heatsink is secured with 4 T20 screws. Unscrew all 4 screws and remove the heatsink to expose the force frame. The next layer is the force frame. This is held down by three T20 captive screws. Unscrew these screws to allow the force frame to spring out.

Beneath the force frame will be a rail frame and a carrier frame. The carrier frame (blue in these pictures) is what will attach directly to the CPU. Once the CPU is in the carrier frame it will then slide into the rail frame. The rail frame essentially aligns the CPU with the socket.

The rail frame also snaps out of place and attached to a hinge mechanism similar to the force frame.

Here is a shot of both the force frame and the rail frame in their upward position:

Remove the blue carrier frame from the rail frame and attach the CPU. Align the arrow on the carrier frame with the arrow on the CPU. Take careful stock of the condition of the carrier frame – the clips in particular. If the clips are damaged are snapped off, the carrier frame will be unable to hold the CPU in place. In this case the CPU could fall out of the frame and potentially land right on the socket and damage pins in the process. Ensure the carrier frame is strong enough to hold the CPU before installing into the rail frame.

Once the CPU is in the carrier frame you can install it in the rail frame:

The rail frame can now be swung back down and clipped into position. This will align the CPU with the socket.

Now the force frame is screwed down:

At this point it’s time to apply a bit of thermal paste:

Now install the heatsink:

Installing the memory

Installing memory can get a bit tricky for this model. HPE has different guidelines for both one and two CPU configurations. Beyond that, there’s also different rules dependent on the exact generation of AMD EPYC CPU. For example, the population rules are different between first generation (7xx1) and second generation (7xx2)  processors.

Here I am using a first generation AMD EPYC CPU so I’ll use the following table as reference.

Click to see memory population rules

HPE ProLiant DL325 Gen10 One Processor Configuration DIMM Population Order
Number of DIMM(s) to populate	Processor 1
	CH H		CH G		CH F		CH E		CH A		CH B		CH C		CH D
1																16
2	1															16
3	1											12				16
4	1				5							12				16
5	1				5							12		14		16
6	1		3		5							12		14		16
7	1		3		5					10		12		14		16
8	1		3		5		7			10		12		14		16
9	1		3		5		7			10		12		14	15	16
10	1	2	3		5		7			10		12		14	15	16
11	1	2	3		5		7			10	11	12		14	15	16
12	1	2	3		5	6	7			10	11	12		14	15	16
13	1	2	3		5	6	7			10	11	12	13	14	15	16
14	1	2	3	4	5	6	7			10	11	12	13	14	15	16
15	1	2	3	4	5	6	7		9	10	11	12	13	14	15	16
16	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16

In this case I have 2 DIMMS to install so referring to the guide I’ll populate slots 1 and 16:

Installing the NVME cables

This server requires an NVME backplane and 4 NVME cables to enable NVME support. One end of the cable attaches to the backplane and the other end goes to the motherboard. On the backplane you will see ports 1 – 4. Each cable will also be labeled appropriately based on what port they go to. Just match up the labels and you’re good to go. The cables for ports 1, 2 and 3 will route to the left side of the chassis while the cable for port 4 will route to the right side.

Here you can see the cables for ports 1, 2, and 3 plugged into the backplane. There is a metal entry point to route that routes them to the motherboard.

Look for the termination points by the blue Smart Array battery compartment.

Finally, port 4 sits by itself on the other side of the backplane and routes along the right side of the chassis instead of the left.

Installing the NDC

HPE has a dedicated spot on the motherboard to install their NDC (Network Daughter Card.) Look for the slot on the upper right side of the chassis.

Locate the installation point. You will have to remove the primary riser 1 to access the NDC location:

Align the NDC:

Firmly press the card into the slot:

Tighten the thumbscrew by hand or with a T15 bit:

Reinstall the riser.

Installing the power supplies

The last set of components to install are the power supplies. You should always install both power supplies to ensure redundancy.

To install, simply slide them in the back.

You will encounter resistance as the power supply contacts the motherboard. Firmly press the power supply until you hear the unit click into place.