HPE ProLiant ML350 Gen10 CTO Tower Server

The refurbished ProLiant ML350 Gen10 leverages the Intel Xeon Scalable processors with up to 71% performance gain and 27% increase in cores, along with the 2933 MT/s or 2666 MT/s HPE DDR4 SmartMemory supports up to 3.0 TB and 11% faster than 2400 MT/s. The shorter re-designed rackable chassis with multiple upgrade options provides flexibility that can expand as your business needs grow. It supports 12 Gb/s SAS, NVMe SSD, embedded 4x1GbE NIC with a broad range of graphics and compute options. Supported by the HPE Pointnext industry-leading service organization, HPE ProLiant ML350 Gen10 server helps you transform to a digital business with more agility and all within your limited IT budget.

What's New for ML350 Gen10

• Supports Intel's 2nd generation Xeon Scalable processor with 16% increase in performance and HPE DDR4 SmartMemory 2933 MT/s, providing a boost in processing performance and memory speed.
• Supports mixed LFF and SFF drive cages within the same server for tiered storage, offering the flexibility to mix drive types for cost and capacity size considerations.
• HPE InfoSight provides a cloud-based analytics tool that predicts and prevents problems before your business is impacted.
• Enhanced iLO 5 security features such as Server Configuration Lock, iLO Security Dashboard and Workload Performance Advisor.
• High performance graphic processing unit (GPU) and high-speed GPU bridge support for graphic-intensive applications such as Virtual Desktop Infrastructure (VDI) and machine learning.
• Supports the new NVIDIA Quadro P2200 and P1000 GPU cards.

Perform with Unmatched Versatility

HPE ProLiant ML350 Gen10 server supports up to two Intel Xeon Scalable processors, starting from Bronze through Platinum, 4 cores expanding up to 28 core processors offering unparalleled performance.

Up to 24 DIMM slots to support the 2933 MT/s or 2600 MT/s HPE DDR4 SmartMemory, reducing data loss and downtime with the HPE Gen10 technology licensed Fast Fault Tolerance feature while increasing workload performance and power efficiency.

It supports a wide range of solutions from Azure to Docker to ClearOS along with the traditional operating systems.

GPU expansion supports up to four units to accelerate performance in VDI applications and machine learning for financial services, surveillance and security, educational and scientific research, as well as retail and medical imaging.

With the new addition of NVIDIA GV100 and NVLINK BRIDGE, NVIDIA Tesla 4 and NVIDIA RTX GPU option support, it transforms into a even more powerful AI Tower server with high-speed GPU connection, ray-tracing and AI.

Base Knowledge
The HPE ML350 Gen10 can be configured in many ways. Please take the time to figure out what you need to make this CTO a bit easier to understand. Now that you have a plan of attack lets do it!

Gen 10 has allowed the mixing of Drive Form Factors. Please select the 4 LFF CTO Server (877625-B21) as the base configuration to start with when mixing Drive Form Factors (ex. 4LFF + 8SFF + 4LFF). You might ask, can I start with the SFF Chassis and mix LFF with it? Yes! The special feature of the 8SFF Chassis (877626-B21) is the ability to add up to 8 NVMe Drives in Box 2 only with the installation of the HPE ML350 Gen10 8SFF NVMe SSD Express Bay Enablement Kit with 2x4NVMe Risers and Support Cables (874569-B21). If you don’t want the NVMe to take up a drive bay, you have the ability to install the NVMe SSD Enablement kit (P26934-B - P26936-B2). Anytime the NVMe drives are installed you will need to also install the Redundant Fan Cage Kit (874572-B21).

If all three Drive cages are installed or 2 Cages and all the Media Bays are filled. You will need to install the redundant fan kit (874572-B21) which takes the system from 2 fans to a total of 6 fans. The redundant fan kit is needed for a plethora of items on this ML350 HPE Server. We will cover that in the cooling section below.

The ML350 support up to (2) CPUs. If both CPUs are installed, you will also need to install the redundant fan kit (874572-B21) which takes the system from 2 fans to a total of 6 fans. Do not Mix Processors. When you select a processor the correct system board will be installed. 879152-001 will be installed when you select the First Gen Intel Xeon Processor and P11390-001 will be selected if you choose the Second Gen Xeon Intel Processor. If the Processor you installed has a TDP of over 85W. Then the Performance Heatsink (879150-001) will replace the standard heatsink (879468-001).

By this point, you should have the CPU and Drive Configuration figured out. Each Processor handles 12 DIMMs with a total of 24 DIMMs in a 2 Socket Configuration. After choosing the Processor, only the supported memory for the system will be visible to you on the CTO. If you have any questions, please feel free to contact us for support.

GPU Accelerators
These Chassis options comes with the Redundant Power Supply (RPS) (874571-B21). Only a Tower Mode system can support a non-hot-plug PSU and if you install a GPU with a TDP larger than 75W. You will need to provide the correct power by installing the HPE ML350 Gen10 GPU External Power Cable Kit (877628-B21). If you install a high-powered GPU, HPE recommends the installation of the 1600W RPS PSU (830272-B21) . The ML350 Gen10 Redundant Fan Cage Kit (874572-B21) is required for ALL GPU installations. Mixing of GPU s is not supported. These GPU s are not recommended for use in office environments especially under stress mode when system fans are running at full speed.

Since there are only 8 PCI-e Slots, a Double-Wide GPU takes up two slots. We can quickly run out of room if we are not carefully thinking of the entire configuration. If you have 8 NVMe Drives installed. That means you have 2 PCI-e slots being used up and they have dedicated slots of 1 and 3. While if you have 24 SFF drives installed that means you have the SAS Expander card in the dedicated PCI-e slot 4 and up to 2 Smart Array in support. In each configuration, the GPU quantity will vary.

This server supports NVIDIA NVLink technology for specific accelerators. This technology allows two matching accelerators to communicate directly through a high-speed interconnection— NVLink bridge. This interconnection effectively combines the available memory of both accelerators for increased workload performance.

The following GPUs only work with 2^nd Gen Intel Xeon Processors.

• NVIDIA T4 16GB Computational Accelerator for HPE (R0W29C)
• NVIDIA A2 16GB PCIe Non-CEC Accelerator for HPE (R9H23C)
• Both require the Redundant Fan Cage Kit (874572-B21)

Let’s Talk Cooling!
The ML350 Gen10 Redundant Fan Kit (874572-B21) takes a 2-fan system to 6 total and is required for one of the following:

• Rack Mode
• 2 Socket Configuration
• CPUs with a TDP Larger than 85W
• When all 3 Drive Cages are filled
• When ML350 Gen10 is used in ASHRAE 3 or 4 environment
• Drives with higher RPM SAS HDDs (10K, 15K, or higher)
• Any kind of NVMe support
• 2 drives are selected
• All GPUs
• InfiniBand card
• Most Performance RAID Controllers
• Computation and Graphics Accelerators
• 100 Gigabit Ethernet adapters
• Intel E810-XXVDA4 Ethernet 10/25Gb 4-port SFP28 Adapter for HPE (P08458-B21)
• HPE Ethernet 10Gb 2-port BASE-T QL41401-A2G Adapter (867707-B21)

***Pretty much anything you do will need the Fan.

Storage Controllers?
Standard on the ML350 Gen10 is the Embedded SW RAID S100i with 14 SATA ports with the ability to upgrade to an HPE modular Smart Array (AROC) and/or PCIe Standup controller card (s). When using the S100i on a LFF chassis, you’ll need HPE ML350 Gen10 Embedded SATA Cable Kit for LFF Configuration (877578-B21). If you have selected the SFF chassis, then you’ll need HPE ML350 Gen10 Embedded SATA Cable Kit for SFF Configuration (877579-B21).

If you want SAS or upgrade from the embedded S100i you have the choice between the AROC style module or a PCI-e Card. By using the AROC you will be saving a PCI-e slot for other devices. Besides the fact the AROC is not installed on a PCI-e slot the other difference is the cabling used for connecting them is different. For AROC modules you will use HPE ML350 Gen10 AROC Mini-SAS Cable Kit for LFF Configuration (874573-B21) or the HPE ML350 Gen10 AROC Mini-SAS Cable Kit for SFF Configuration (877575-B21). If you’re using a Smart Array or HBA card. You’ll use these cables, HPE ML350 Gen10 Smart Array/HBA Mini-SAS Cable Kit for LFF Configuration (874574-B21) or HPE ML350 Gen10 Smart Array/HBA Mini-SAS Cable Kit for SFF Configuration (874575-B21).

Now for powering some of the Performance RAID Controllers with Flash Back Write Cache (FBWC). You’ll have the choice between the HPE Smart Storage Battery (P01367-B21) or the HPE Smart Storage Hybrid Capacitor (P02381-B21). The main difference between these is the amount of devices they support. The HPE Smart Storage Battery (P01367-B21) is a single 96W battery that can support up to 24 devices. While the HPE Smart Storage Hybrid Capacitor (P02381-B21) can support up to three devices.

Let’s talk power!
All three models come with the Redundant Power Supply (RPS) (874571-B21 ). This RPS can hold up to 2 PSU depending on the desired power needed. The mixing of 2 different power supplies is NOT allowed. The selection of two HPE Flex Slot power supplies provide 1+1 power redundancy. The 1600W RPS is recommended when the system is configured with high power GPU. The 800W RPS will work too (per GPU) – depending on configuration. If you want to revert to the Non-Hot-Plug PSU, the RPS will be removed.

Lights Out! iLO
iLO 5 - is a remote server management processor embedded on the system boards of HPE ProLiant servers and Synergy compute modules. iLO enables the monitoring and controlling of servers from remote locations.
iLO RESTful API - is Redfish API conformance and offers simplified server management automation such as configuration and maintenance tasks based on modern industry standards.
HPE OneView Standard - can be used for inventory, health monitoring, alerting, and reporting without additional fees. It can monitor multiple HPE server generations. The user interface is like the HPE OneView Advanced version, but the software-defined functionality is not available.