The HPE ProLiant DL360 Gen10 10-Bay 2.5" is the densest 1U SFF configuration in the Gen10 family and the canonical 1U HPE Gen10 page on our site. Ten 2.5" hot-swap bays, dual-socket Intel Xeon Scalable (Skylake-SP first generation or Cascade Lake-SP second generation), 24 DDR4 DIMM slots, iLO 5 with Silicon Root of Trust, and the HPE Smart Array storage controller family in a 1U chassis optimized for compute density. This is HPE's 1U workhorse for IT teams standardized on the ProLiant line and the architectural counterpart to the Dell PowerEdge R640 - choose by your shop's vendor standardization, not by capability gaps, because at this tier the two platforms trade blows feature for feature.

For HPE shops running vSphere clusters, Hyper-V deployments, scale-out application infrastructure, Kubernetes worker pools, or any compute-primary workload where rack density and per-node power efficiency matter more than per-chassis storage capacity, the DL360 Gen10 10-Bay is the right answer. Pair it with its 2U sibling, the DL380 Gen10 16-Bay 2.5", when storage flexibility or PCIe expansion requires the larger chassis. Same processors. Same memory. Same iLO 5. Different chassis-level constraints.

To talk through a configuration, call 1-800-778-1545 or use the quote form below. We respond within 24 hours, every refurbished unit ships under our 180-day warranty, and every server runs through 12+ hour burn-in testing before it leaves the bench. Volume pricing kicks in at 5 units.

Platform Overview - Purley in 1U

The DL360 Gen10 sits on Intel's Purley platform with the LGA 3647 socket, identical to the DL380 Gen10. That means the same dual-generation processor support: 1st Gen Intel Xeon Scalable (Skylake-SP) for the original Gen10 launch, and 2nd Gen Xeon Scalable (Cascade Lake-SP) added mid-lifecycle as a drop-in upgrade with no board respin required. If you're sourcing CPUs separately or planning a future upgrade path, this matters: any Cascade Lake-SP processor in the supported TDP range works in a Skylake-era DL360, and any Skylake-SP processor works in a later-production Cascade Lake-era DL360, subject to BIOS revision compatibility.

The 1U thermal envelope is the meaningful constraint versus the 2U DL380. Both chassis support the same processor SKUs in principle, but the DL360 has tighter heatsink options and less airflow headroom. Processors above 165W TDP - Platinum 8260, 8268, 8280 territory - require careful confirmation of heatsink configuration and ambient inlet temperature. The DL360 ships with two heatsink variants: a standard heatsink for processors up to roughly 150W, and a high-performance heatsink for processors above that. For Gold 6230 (20 cores, 125W), Gold 6240 (18 cores, 150W), and similar mainstream dual-socket SKUs, the DL360 is comfortable. For top-bin Platinum parts where you genuinely need every core at every clock, the DL380 2U is the safer thermal envelope. We'll confirm the heatsink and ambient guidance at quote time based on the CPU SKU you specify.

Memory architecture is identical to the DL380: 24 DDR4 DIMM slots total across both sockets, six memory channels per CPU at two DIMMs per channel. The supported DIMM speeds depend on the processor: Gold 6200-series and Gold 5222 SKUs run DDR4-2933 at 1 DIMM per channel, the rest of the Skylake and Cascade Lake lineup runs DDR4-2666. RDIMM capacity goes up to 64 GB per slot for 1.5 TB per dual-socket system; LRDIMM goes to 128 GB per slot for 3 TB total. Intel Optane Persistent Memory 100-series is supported with M-suffix Cascade Lake CPUs (Gold 6230M, Platinum 8260M, etc.) in the documented ratios. HPE NVDIMM-N is supported on Skylake-only platforms - not Cascade Lake - and that's a vendor product matrix limitation, not an Intel one.

HPE's memory population rules apply identically to the DL360 and the DL380: DIMMs must be installed in even quantities per CPU, RDIMM and LRDIMM cannot be mixed, and only HPE DDR4 Smart Memory is qualified to run at the rated speeds. Third-party memory will physically work but typically drops to DDR4-2400 regardless of the CPU's rated speed. If you have a DDR4-2933 workload requirement, you need HPE Smart Memory. We stock HPE Smart Memory and will spec the kit at quote time.

Storage - 10 SFF Bays, the Density Pick

Ten 2.5" SAS/SATA hot-swap bays in 1U is the maximum SFF density on the DL360 Gen10 platform. Two additional bays over the more common 8-Bay configuration. The 10-Bay backplane is fully SAS/SATA-capable across all ten slots, and NVMe is supported on the original Gen10 only via dedicated NVMe expansion - native backplane NVMe across all bays is a Gen10 Plus and Gen11 feature, not original Gen10. If you need direct-attached NVMe storage at meaningful capacity, the Gen10 supports it through PCIe expansion cards routing to specific NVMe-capable bays, not as a backplane-wide capability.

The 10-Bay configuration earns its place over the 8-Bay in three specific scenarios:

• Ceph OSD nodes at 1U density. Ten OSDs per 1U node is a meaningful density improvement over eight when you're sizing a Ceph cluster for object storage or RBD workloads. The math at scale matters: a 12-node Ceph cluster with 10 OSDs per node is 120 OSDs; the same cluster at 8 OSDs per node is 96 OSDs. That 25% capacity difference per rack at the same rack-U cost is the reason this configuration exists.
• vSAN hybrid configurations. vSAN hybrid wants one SSD cache device per disk group and up to seven HDDs per group, with two disk groups per host as a common configuration. Ten bays gives you 2x (1 cache + 4 capacity) cleanly with two bays held back, or 2x (1 cache + 4 capacity) with the remaining two used for OS boot if you're not using M.2. vSAN all-flash configurations benefit similarly: 2x (1 cache + 4 capacity) all-SSD at 1U is a genuine vSAN ReadyNode-class density point.
• Distributed databases and Kubernetes persistent volumes. Cassandra nodes, MongoDB replica set members, Elasticsearch data nodes, or Kubernetes CSI-backed persistent volume hosts all benefit from the extra two bays when the workload's per-node storage requirement is in the 6-10 drive range. Eight bays is tight when you also need OS boot drives in the bay count; ten bays gives breathing room.

If your storage design fits in 8 bays comfortably and you're using HPE's M.2 enablement kit or a single boot drive in a bay, the 8-Bay variant is the simpler and slightly less expensive choice. The 10-Bay premium is modest but it's a real cost - pay it when the bays earn their place, not by default.

Drive options span the full Gen10 portfolio: SAS SSDs from 480 GB through 15.36 TB across read-intensive, mixed-use, and write-intensive endurance classes; SATA SSDs in mixed-use and read-intensive for cost-sensitive workloads; SAS HDDs at 10K and 15K from 600 GB through 2.4 TB for moderate-IOPS workloads; and NL-SAS in SFF form factor up to 2.4 TB for capacity tiers in SFF chassis. We carry the full HPE-branded drive line and will spec the right tier and endurance class based on your workload's read/write profile and the controller's queue depth tolerance.

Storage Controllers - Smart Array Family

The DL360 Gen10's controller options are identical to the DL380 Gen10's, configured for the 1U chassis:

• Smart Array P408i-a SR Gen10 with 2 GB FBWC. The mainstream production controller: RAID 0/1/5/6/10/50/60, 2 GB flash-backed write cache, full hardware RAID acceleration. This is the right pick for hardware-RAID storage where the OS sees a single logical drive per RAID group. The FBWC battery is a wear item with roughly a 5-year service life under typical conditions - plan a battery replacement somewhere in years 4 through 5, and watch iLO for cache module health alerts. We disclose this on every quote; a refurbished P408i-a's battery may have meaningful runtime already on it, and we either replace the cache module up front or document the battery's measured state at burn-in.
• Smart Array P816i-a SR Gen10 with 4 GB FBWC. The premium controller: same RAID modes, 4 GB FBWC, higher port count, and tri-mode SAS/SATA/NVMe support on supported drives. Specify the P816i-a when the workload is write-heavy at scale (transactional databases, write-mostly logging, video ingest) and the 2 GB cache on the P408i-a is the bottleneck. For most 1U DL360 deployments, the P408i-a is plenty; the P816i-a is the right call when you've actually measured cache pressure on a comparable workload or you need NVMe drive support alongside SAS/SATA.
• Smart Array E208i-a SR Gen10 (HBA mode). The HBA controller for software-defined storage: vSAN, Ceph, Storage Spaces Direct (S2D), ZFS-based appliances. No hardware RAID, just clean SAS HBA pass-through to the OS or hypervisor. This is the right pick for any storage-defined-in-software architecture where the hardware RAID controller would actually interfere with the SDS layer's drive-level visibility.
• S100i SR Gen10 (software RAID). The chipset-integrated software RAID solution. Adequate for boot-drive mirrors on Windows or Linux but not appropriate for production data RAID. Use it for OS boot if you don't have an M.2 enablement kit installed; use a real Smart Array P-series or HBA for data.

Boot drive options on the DL360 Gen10 are the HPE M.2 enablement kit (PCIe-attached M.2 carrier with SATA M.2 drive support, typically 480 GB) or a SAS/SATA SSD pair in two of the front bays under hardware RAID 1. The HPE NS204i-p NVMe boot device that ships standard on Gen10 Plus and Gen11 is not natively supported on original Gen10 - that's a generational platform line, not a configuration option. If you need NVMe boot specifically on Gen10, it's via the M.2 kit or via a PCIe-attached NVMe drive in a bay, not via NS204i-p.

Networking and PCIe

The DL360 Gen10 provides 3 PCIe Gen3 slots in a 1U chassis - a meaningfully tighter constraint than the DL380's 8 slots, and the single most important architectural difference between the 1U and 2U chassis at this tier. Slot 1 and slot 2 are full-height half-length, slot 3 is low-profile. The standard riser configuration supports two x16 slots; an alternative riser provides three slots with a x16/x8/x8 layout.

HPE's FlexibleLOM mezzanine slot is separate from the PCIe slots and does not consume one. This is HPE's equivalent of Dell's rNDC (network daughter card) on the R640 - a dedicated mezzanine for the primary network interface that leaves all three PCIe slots free for HBAs, GPUs, or additional NICs. FlexibleLOM options span 1 GbE quad-port, 10 GbE SFP+ dual-port and quad-port, 10 GbE RJ45 dual-port and quad-port, 25 GbE SFP28 dual-port, and 100 GbE QSFP28 dual-port. For HCI workloads (vSAN, S2D, Ceph) where 25 GbE has become the standard interconnect, the 25 GbE SFP28 FlexibleLOM plus a 25 GbE SFP28 PCIe NIC in slot 1 is the standard high-bandwidth configuration. We'll spec the FlexibleLOM and any additional PCIe NICs at quote time based on your network topology.

GPU support in 1U is sharply constrained relative to the DL380. The DL360 Gen10 supports a maximum of two single-width low-profile GPUs - typically NVIDIA T4 or A2 in the inference/light-compute class. Double-width GPUs (V100, A100, A40, A30) and full-height cards do not fit the 1U chassis at all. If you need GPU compute beyond two T4-class cards, this is not the right chassis - the DL380 Gen10 supports up to seven T4s or three V100s with the GPU riser kit. The DL360 is a CPU compute platform first; GPU is a secondary capability bounded by the 1U thermal and slot envelope.

Management - iLO 5 with Silicon Root of Trust

iLO 5 is HPE's out-of-band management processor, the architectural equivalent of Dell's iDRAC9 on R640/R740 generations. Full remote KVM, virtual media mounting (mount an ISO over the network for OS install), serial-over-LAN, hardware health monitoring, power and thermal telemetry, REST API (Redfish-compliant), HPE OneView integration, and Active Health System logging. iLO is on a dedicated management network port - segregate it on a management VLAN; never expose iLO to the production network or the public internet without VPN. Both have full filesystem access to the host and full power control.

Silicon Root of Trust is HPE's Gen10-and-later platform security baseline. It's a hardware-anchored chain of trust starting from the iLO 5 silicon, verifying iLO firmware, then BIOS, then OS bootloader against cryptographic measurements baked into the silicon at manufacture. This is functionally equivalent to Dell's iDRAC9 System Lockdown plus Intel Boot Guard, just architected differently. For environments with security audit requirements - PCI DSS, HIPAA, FedRAMP-aligned, or any environment where firmware tampering is a documented threat - Silicon Root of Trust is a meaningful differentiator from older Gen9 hardware that predates it.

One important note on refurbished units: iLO Advanced licensing is typically not included on refurbished Gen10 hardware. iLO Advanced unlocks the full feature set - remote KVM in particular, plus integrated remote console, video record/replay, virtual folder, and email alerting. The base iLO 5 license that ships with the hardware includes health monitoring, IPMI, and basic remote access but not the full graphical remote KVM. We can include iLO Advanced licensing on builds where it's required - call this out at quote time so we spec the license correctly.

Power and Cooling

HPE Flex Slot power supplies in the standard hot-plug redundant configuration. Wattages span 500W, 800W, 1600W, and 1600W -48V DC. Platinum efficiency on the 500W and 800W; Titanium efficiency on the 1600W (96% efficiency at 50% load, the highest 80 PLUS tier). For a typical dual-socket Gold-class DL360 with 16 DIMMs and 10 SFF SSDs, the 800W Platinum redundant pair is the standard sizing - generous headroom for any reasonable configuration and excellent efficiency at typical load. For high-TDP Platinum-CPU builds, top-bin DIMM populations, or single-PSU operation with redundancy as the failover path, step to the 1600W Titanium pair. We size the PSUs to the build and document the expected peak draw at quote time.

Fans are dual-rotor, fully redundant, hot-swappable. Inlet temperature spec is 10°C to 35°C ambient for standard operation; ASHRAE A3 (10°C to 40°C) supported on most configurations and A4 (5°C to 45°C) on specific reduced-CPU configurations. For colocation environments running ASHRAE A2 or stricter, the DL360 has plenty of thermal margin in any dual-socket Gold-class build. For high-density racks running hot, confirm the inlet spec at quote time - particularly for Platinum-CPU configurations approaching the 165W per-socket envelope.

Where the DL360 Gen10 Fits in 2026

The DL360 Gen10 launched in 2017 and was Cascade Lake-refreshed in 2019. By 2026, it's two platform generations behind current: Gen10 Plus (Ice Lake-SP, PCIe Gen4, 2020) and Gen11 (Sapphire Rapids / Emerald Rapids, DDR5, PCIe Gen5, 2023-2024). That generational gap is real and we won't pretend otherwise. What it means in practice:

• Per-core performance is solid; per-socket maximums are dated. A Cascade Lake Gold 6230 at 20 cores per socket / 40 cores dual-socket is still a competitive virtualization host in 2026 - VMware vSphere 8 supports it, Windows Server 2025 supports it, modern Linux supports it, container workloads run fine. What's dated is socket-level scaling: Ice Lake-SP hit 40 cores per socket, Sapphire Rapids hit 60, Emerald Rapids hit 64. If your workload's bottleneck is socket-level core count, Gen10 is behind.
• PCIe Gen3 is the I/O limit. NVMe drives, 100 GbE NICs, and modern GPUs are all Gen4 or Gen5 internally. They work in Gen3 slots at reduced lane bandwidth - a Gen4 x4 NVMe drive runs at Gen3 x4 speeds, which is roughly half its rated throughput. For storage-intensive or networking-intensive deployments where Gen4 bandwidth actually matters, Gen10 Plus or Gen11 is the right platform. For compute-primary workloads where Gen3 NVMe bandwidth is sufficient, the limit doesn't bite.
• Memory bandwidth ceilings are real. DDR4-2933 maxes out at roughly 23.5 GB/s per channel, six channels per socket, ~141 GB/s per socket peak. Ice Lake-SP brought eight channels and DDR4-3200; Sapphire Rapids brought DDR5-4800. Memory-bandwidth-bound workloads (in-memory databases, HPC, certain analytics) see meaningful uplift on newer platforms. General-purpose virtualization and application serving rarely hit the memory bandwidth ceiling on Gen10.

That said, Gen10 is widely deployed across the enterprise installed base in 2026. HPE supports Gen10 firmware updates through 2027 under standard lifecycle policy, and parts availability for both new and refurbished components is broad - this is the most-deployed ProLiant generation in service today. For HPE shops standardized on Gen10 and looking to expand existing clusters with matching hardware, the platform makes complete sense. For greenfield deployments where budget allows, Gen10 Plus or Gen11 will get you Ice Lake or newer with PCIe Gen4/5 and more headroom for the next five years.

Honest Limitations

The points worth saying out loud before you buy:

• The Smart Array FBWC battery is a wear item. Cache module batteries have a service life of roughly 5 years under typical operating conditions, often less in hot environments. A refurbished P408i-a or P816i-a's battery may have meaningful runtime already. We replace cache modules on builds where the battery is past spec, and we document the measured battery state on every unit shipped with a P-series controller.
• iLO Advanced licensing is usually not included. Refurbished Gen10 units typically ship with the base iLO 5 license, not iLO Advanced. If you need integrated remote console KVM (the graphical remote console most people associate with iLO), the iLO Advanced license is a real cost. We'll quote it explicitly when it's required.
• The DL360 is a CPU compute platform, not a GPU compute platform. Two single-width T4-class cards is the practical ceiling. If you need GPU compute at scale, the DL380 Gen10 or a purpose-built GPU server is the right answer.
• Original Gen10 is PCIe Gen3. Modern NVMe and 100 GbE cards work but at reduced bandwidth versus their Gen4/Gen5 native platforms. For PCIe-Gen3-bandwidth-bound workloads specifically, this matters.
• HPE memory rules are strict. Third-party DDR4 will run at DDR4-2400 regardless of the CPU's rated speed. HPE DDR4 Smart Memory is required for rated DDR4-2666 or DDR4-2933 operation. This is documented HPE behavior, not a defect.

Workload Fit

Cross-Vendor Notes - DL360 Gen10 vs. Dell R640

The DL360 Gen10 and the Dell PowerEdge R640 are direct architectural counterparts. Both are 1U dual-socket Purley-platform servers. Both support 1st Gen and 2nd Gen Xeon Scalable. Both top out at 24 DDR4 DIMM slots. Both have integrated BMC-class management (iLO 5 vs. iDRAC9) with full remote console capability. Both have a dedicated network mezzanine slot (FlexibleLOM vs. rNDC) that doesn't consume a PCIe slot. Both have similar storage controller families (Smart Array P-series vs. PERC H730/H740/H840). The DL360 has 3 PCIe slots vs. the R640's 3 PCIe slots - identical. Drive bay configurations align closely (HPE 8-bay and 10-bay vs. Dell 8-bay and 10-bay).

The differences are at the vendor-ecosystem level: management software (iLO + OneView vs. iDRAC + OpenManage), licensing models for advanced features, the vendor's installed-base relationships, parts ecosystem and refresh cadence, and support contract pricing. None of these is a capability gap. Pick by your shop's vendor standardization and your existing tooling investments. If you're an HPE shop running OneView and Insight, the DL360 is the right answer. If you're a Dell shop running OpenManage Enterprise, the R640 10-Bay is the right answer. We sell both and we'll quote both honestly.

Frequently Asked

Does the DL360 Gen10 support Cascade Lake processors? Yes. The platform supports both 1st Gen Xeon Scalable (Skylake-SP) and 2nd Gen Xeon Scalable (Cascade Lake-SP) as drop-in compatible processors with appropriate BIOS revision. Cascade Lake brings hardware Spectre/Meltdown mitigations, slightly higher core counts at the top of the stack, support for Optane Persistent Memory 100-series, and minor power efficiency improvements. Most production refurbished DL360 Gen10 units we ship are Cascade Lake-equipped.

Is the DL360 Gen10 the same as the DL360 Gen10 Plus? No. Gen10 Plus is a different platform: Ice Lake-SP processors on LGA 4189, PCIe Gen4, eight memory channels per socket, and the HPE NS204i-p NVMe boot device standard. Gen10 is Purley/Skylake-Cascade Lake/PCIe Gen3. They look similar externally and share the DL360 chassis lineage, but the motherboard, processor socket, and I/O are different generations. If you've been told "DL360 Gen10" and you need PCIe Gen4 or Ice Lake, confirm the actual SKU before purchase.

Can I run VMware vSphere 8 on the DL360 Gen10? Yes. vSphere 8 supports both Skylake-SP and Cascade Lake-SP processors and HPE certifies Gen10 ProLiant hardware against vSphere 8. There are no platform-level blockers. vSphere 9 deprecates some older hardware - check VMware's HCL at the time of deployment for current support status. As of early 2026, Gen10 is fully on the vSphere 8 HCL.

What about Windows Server 2025? Supported. HPE has Gen10 firmware and driver updates qualified against Windows Server 2025; the platform meets all of WS2025's hardware baseline requirements including TPM 2.0 (optional but available on Gen10 via the HPE TPM 2.0 module). Standard build.

Does the DL360 Gen10 support NVMe drives? Yes, via the NVMe expansion kit on specific bay positions. Native NVMe across all bays is a Gen10 Plus and Gen11 feature; the original Gen10 supports NVMe on a subset of bays through PCIe lane routing from a Smart Array P816i-a (tri-mode) or a dedicated NVMe-bay enablement kit. If NVMe is core to your storage design, we can spec it; if NVMe is the dominant storage tier, Gen10 Plus is the more native platform.

Do you ship the iLO Advanced license? Optional. The base iLO 5 license that ships with refurbished Gen10 hardware does not include integrated remote console (graphical KVM), remote media, or some of the advanced telemetry features. iLO Advanced unlocks all of these. Call it out at quote time and we'll include the license SKU.

What's the warranty? 180-day Wholesale Servers warranty on every refurbished unit, covering parts and labor. Pre-shipment burn-in testing is 12+ hours minimum on every server. Extended warranty options are available; ask at quote time.

Ready to Configure?

Tell us the workload, the CPU SKU (or the per-socket core count and clock target - we'll recommend the SKU), memory capacity, storage configuration including controller preference, network topology and FlexibleLOM choice, and quantity. We respond within 24 hours, every refurbished unit ships with the Wholesale Servers 180-day warranty and 12+ hour burn-in, and volume pricing starts at 5 units. Call 1-800-778-1545 or use the quote form below to start a conversation.