HPE Proliant DL560 G10 24-Bay 2.5"
Configure Your System:
1 Processor Required
Step 1: Choose a generation
1st Gen (Skylake) Scalable is the value entry into four-socket computing. 2nd Gen (Cascade Lake) refines the same socket: better clocks per watt and faster memory support, and our default recommendation when the price gap is modest.
You will notice the menu starts at Gold: four-socket operation is a Gold and Platinum capability on this platform, which is why the value tiers are absent.
Step 2: Choose the shape of the compute
Higher clocks, fewer cores: the lower-core Golds suit per-core licensed software, and on a four-socket box that licensing math deserves respect before anything is final. Database engines are the classic case: licensing often dwarfs the hardware price, and every core you do not license is money kept.
More cores: the high-core Golds and the Platinum tier are the DL560's whole argument, four sockets of dense silicon in two rack units, consolidation that once took half a rack. Memory bandwidth scales with the sockets too: four processors means four full sets of memory channels.
Memory capacity is the quiet reason people buy this server; if your workload is memory-bound rather than compute-bound, weight the CPU budget toward the middle and spend the difference on DIMMs.
One honest licensing note before you commit: Windows Server licenses every core in the chassis, and four populated sockets multiply that quickly. Read the OS step before finalizing, and we will run the numbers with you at quote time; call 1-800-778-1545.
2 Heat Sink Optional
What matters is that this is a 2U chassis cooling four sockets of Scalable silicon, and it takes airflow discipline seriously. Keep every drive bay blank and internal baffle installed; the DL560 cools front to back, and with this much silicon in the airstream, an open path lets air skip the components that need it.
Set expectations accordingly: a four-socket 2U under load is not a quiet machine, and it was engineered for datacenter intake air. It will run in a well-ventilated server room, but fan and component life is better in a real cold aisle, and this is not a server to rack beside a desk.
3 Memory (RAM) 48 DIMM slots, modules added in sets of 4 Required
The guidance follows the channels at scale.
Twenty-four matched DIMMs, six per processor, engages every channel once at full speed, and that is our default: full bandwidth across all four sockets with a clean doubling path. Forty-eight DIMMs doubles capacity at two per channel with no speed penalty on this platform, and it is the natural full build when the workload is the reason you bought a four-socket box.
Your processors set the speed: the newer generation and higher tiers run the DIMMs faster, part of what you buy in the processor step.
Keep the population balanced across all four processors, always; an unevenly loaded socket starves its share of the workload. And buy it as one matched set, now: upgrade hunting on a production four-socket box means matching rank and speed across dozens of slots in a maintenance window nobody wants to schedule.
Common landing spots: 384 GB (24x 16 GB) for general four-socket duty, 768 GB (24x 32 GB) is our default for the database and consolidation work this chassis is bought for, and 1.5 TB (48x 32 GB) when memory is the whole point.
4 RAID Controllers Optional
What it gives you: full hardware RAID across the backplane with a battery-protected write cache, which keeps parity arrays performing under the real write load a four-socket consolidation host generates.
The part worth knowing: the P408i also runs in HBA mode, passing drives straight through to the operating system. That covers both classic DL560 architectures with one controller, local hardware RAID for datastores and database volumes, and direct drive access for the SAN-boot or software-defined builds where local storage is a boot mirror and the real capacity lives elsewhere. If your build is software-defined or SAN-attached, tell us at quote time and we will ship it configured the right way.
Array guidance for the hardware RAID path: RAID 10 for database volumes where write latency matters, RAID 6 over RAID 5 on anything large and spinning, and a hot spare on any array whose rebuild you would rather not race. A server bought for consolidation should not have its storage be the casual part.
5 Storage Drives Select up to 24 drives (0/24 Slots Used) Optional
The DL560 Gen10 takes 2.5" SAS and SATA drives in both flavors, and for the workloads this chassis is bought for, our guidance is direct: SSDs. Database volumes and dense VM storage want flash latency, and pairing four sockets of Scalable compute with spinning-disk storage wastes the compute you paid for. On the 24-bay chassis, a full flash backplane behind the cached controller is a serious storage subsystem in its own right.
Just as common at this tier: a small local RAID 1 pair for boot and everything else on a SAN. If that is your architecture, two SSDs and the P408i-a finish this page.
Buy new, and on flash we will add the technical reason to the usual one: endurance is consumed by writes, and every drive here ships with its full write budget ahead of it. We steer you away from refurbished on drives specifically, and we say that as a company that sells refurbished servers all day.
Plan the boot volume out of your data bays: a small RAID 1 pair of SSDs in two bays is the standard pattern. A boot failure and a data failure should never be the same event.
Fair warning on capacity math: RAID overhead, hot spares, and filesystem headroom mean usable space lands well under the sum of the labels. Size for the usable number.
6 Remote Access Required
On this server, take the Advanced license, and we will make the case rather than just assert it: the dividing line between the tiers is the remote console, the ability to reach the server's screen, mount an ISO, and rebuild below the OS from anywhere. A four-socket server is bought to carry workloads a business feels when they stop, it almost always lives in a datacenter or colo rather than down the hall, and the license costs a fraction of one emergency site visit. Standard iLO on a machine like this saves a small amount of money exactly where saving money stops making sense.
iLO 5 also brings HPE's silicon root of trust and firmware protection machinery, and the Advanced tier is where its best capabilities live.
If you are deploying 5 or more units, take the Advanced license on all of them. Fleet management without remote console access does not work in practice.
7 Power Supply 2x included in build Required
2x 800W is the configurator default and the right starting point for a four-processor machine: four sockets of Scalable silicon, a serious memory bank, and a bay of SSDs with the 50 percent load target intact, where a PSU runs at peak efficiency, runs cooler, and lasts longer than one working near its limit.
When the build reaches the top of the range, Platinum silicon across four sockets, a full 24-bay backplane, a heavy DIMM count, size up rather than run the pair near its limit; the price difference is a rounding error against the rest of this machine, and we will flag it at quote review if your configuration disagrees with its power supply.
Run the cords to separate circuits or PDUs and a single upstream failure does not touch the workload. Watch the estimated TDP counter at the bottom of the page as you add components; on a four-socket build it is worth an honest look before checkout.
8 Network Cards Required
On this server the speed decision is short: a four-socket consolidation box concentrates many workloads onto one network drop, and gigabit networking is almost never the right match for that concentration. Treat 10 Gb as the floor: the dual 10 Gb options are the standard pick, production on the fast pair, management on the embedded ports, and many DL560 deployments add PCIe cards beyond that to separate storage, VM, and backup traffic onto their own links.
Then port type, and this is the one that catches people. SFP+ ports take fiber or DAC cables and talk to datacenter-grade switches. BASE-T ports are 10 Gb copper on ordinary RJ45. Neither is better; they must simply match your switch. Order SFP+ against a copper-only switch and the server racks up with nothing to plug into.
A DL560's network is usually part of a larger architecture conversation. Call 1-800-778-1545 and we will spec the whole path, switch ports included.
9 Operating System Required
The good news on this platform: the Gen10's Xeon Scalable processors are fully supported by every modern OS, so the choice is about your environment, not compatibility.
Windows Server on a four-socket server deserves clear eyes: licensing is per core across every core in the chassis, and four populated sockets multiply that quickly. Standard edition also carries limited virtualization rights, so for the VM density a DL560 is bought for, Datacenter edition with its unlimited virtualization rights is usually the honest comparison. Windows Server 2022 is the mainstream pick and 2025 is fully supported. We will run both license paths with you at quote time; on CALs, count your users and round up.
Fair warning on Evaluation Editions: free and fully functional for 180 days, then they expire. Right for labs and proof-of-concept, wrong for production.
VMware ESXi is fully supported on this platform, and a four-socket Gen10 host is a legitimate vSphere consolidation node; we will install the version your cluster runs.
Proxmox VE is the quiet star on this chassis: open-source KVM with clustering and live migration, zero per-core licensing on a server with this many cores, and full support on this hardware. Four sockets with no licensing meter running changes what this server costs to operate.
Ubuntu Server LTS covers Linux database and application duty with five years of free updates, and No OS is there for teams imaging from their own deployment infrastructure.
A server at this tier deserves a conversation. Tell us the workload at quote time, or call 1-800-778-1545.
WServer Warranty
HPE ProLiant DL560 Gen10 24-Bay 2.5" Drives [Gen10] Detailed Review
The HPE ProLiant DL560 Gen10 24-Bay 2.5" pairs HPE's 4-socket flagship compute platform with maximum SFF storage density - twenty-four 2.5" hot-swap bays in a 2U chassis alongside up to four Intel Xeon Scalable processors, 48 DDR4 DIMM slots, 6 TB memory ceiling with LRDIMMs, FlexibleLOM networking, iLO 5 with Silicon Root of Trust, and up to 4 HPE Flex Slot power supplies. This is a deliberately specialized configuration: 4-socket compute for scale-up workloads combined with 24-bay SFF storage for high-density database, analytics, or HCI data that lives locally rather than on a SAN.
For the full DL560 Gen10 platform documentation - including the honest framing on when 4-socket compute is and isn't the right call, Section 12 platform vocabulary (CPU/memory/PCIe/management), and Dell PowerEdge R840 cross-vendor reference - see the DL560 Gen10 8-Bay 2.5" canonical. This page focuses on what's specific to the 24-bay variant: when 24 SFF bays alongside 4-socket compute is the right tool, the bay-count-driven workload patterns, and the storage controller and power decisions that change at 24 bays.
To configure a build, call 1-800-778-1545 or use the quote form below. Every refurbished unit ships under our 180-day warranty with 12+ hour burn-in testing, and volume pricing starts at 5 units. The 24-bay configuration benefits from extra design discussion - 4-socket compute plus 24 SSDs in 2U is genuinely dense and the architectural choices have downstream operational consequences worth getting right at quote time.
When 24 SFF Bays + 4-Socket Makes Sense
The 24-Bay DL560 is a deliberately narrow configuration. Most 4-socket workloads (SAP HANA, Oracle Database, mission-critical virtualization, SQL Server Enterprise) don't need 24 local SFF drives - they either use a SAN for primary storage or a smaller number of high-performance local SSDs alongside networked storage. The DL360-class 8-Bay DL560 is the right answer for most of those builds.
The 24-Bay DL560 earns its place when both 4-socket compute AND high-density local SSD storage are genuine requirements. The specific scenarios:
- SAP HANA with large local SSD persistence layer. HANA in-memory databases benefit from local NVMe/SSD for log persistence and warm-data tiering rather than depending on SAN latency for log writes. 24 SFF bays alongside HANA-scale memory (3-6 TB DDR4 + up to 12 TB Persistent Memory on L-series CPUs) enables a complete in-memory plus fast-persistence architecture in a single chassis. The persistence layer fits in the chassis instead of crossing the SAN, which matters for HANA savepoint and log-replay latency.
- Oracle Database with local ASM diskgroups. Oracle RAC or large-instance Oracle databases where the design choice is local SSD storage rather than SAN. 24 SAS SSDs in ASM disk groups deliver high IOPS and predictable latency without the SAN dependency. Common when SAN is unavailable, undesirable for licensing/cost reasons, or simply when the database team has decided ASM-on-local-SSD is the operational pattern they want to standardize on.
- SQL Server Enterprise with extensive tempdb and log staging on local SSDs. Per-core SQL Server licensing economics already favor consolidation on 4-socket compute; pairing with 24 high-endurance SSDs lets the entire tempdb plus transaction log infrastructure live on local NVMe/SAS rather than crossing the SAN. Datafile-on-SAN plus tempdb-and-logs-on-local-SSD is a documented Microsoft pattern for performance-sensitive SQL Server deployments.
- High-density VMware HCI or vSAN ReadyNode at 4-socket scale. vSAN ReadyNode configurations at 24 SFF bays with 4-socket compute deliver high VM density per host. Fewer, larger HCI nodes reduce vSphere license count (which is per-CPU socket) and rack footprint. The 24-bay DL560 is at the high end of the vSAN ReadyNode footprint and works well when the goal is consolidating to the fewest hosts possible.
- In-memory analytics with large local hot-data tier. Analytics workloads (Splunk, Elasticsearch hot-tier, in-memory data grids) that need both maximum processing capacity (4-socket) and large local SSD datasets that don't fit entirely in DRAM but are too latency-sensitive for SAN. 24 SAS SSDs as a tiered hot-data layer behind in-memory analytics is a meaningful configuration.
- Microsoft Storage Spaces Direct (S2D) at 4-socket scale. S2D requires HBA-mode storage and benefits from high drive counts per node for performance scaling. 24 NVMe or SAS SSDs in a 4-socket S2D node delivers a high-density HCI design with the per-node compute headroom to host many workloads.
If either the 4-socket compute or the 24-bay storage capacity is more than the workload actually needs, a different platform delivers better economics. The DL380 Gen10 24-Bay covers high-density storage at the dual-socket tier; the DL560 Gen10 8-Bay covers 4-socket compute with modest local storage. Pay for both 4-socket and 24-bay only when both are genuine requirements.
Storage - 24 SFF Bays
Twenty-four 2.5" SAS/SATA/NVMe hot-swap bays across three drive boxes (Box 1, Box 2, Box 3) in the front of the chassis. With the full 24-bay configuration populated, the optional Universal Media Bay (front display port plus optical drive) is not supported - the media bay occupies the same physical position as one of the drive boxes. Production 24-bay builds typically don't need the media bay; remote iLO 5 access covers the operational requirements that the media bay served on earlier platforms.
Drive options span the full Gen10 SFF portfolio: SAS SSDs (mixed-use and read-intensive endurance tiers, 480 GB through 7.68 TB), SATA SSDs for cost-optimized roles, SAS HDDs at 10K and 15K for moderate-IOPS data, NVMe SSDs via specific drive cage and riser combinations (see NVMe section below), and self-encrypting drive variants for compliance-regulated deployments. Per-drive type mixing is supported subject to controller capability.
RAID guidance at 24 SFF bays: RAID 6 is appropriate for capacity-optimized SAS/SATA SSD pools where rebuild windows on individual drive failure need to be tolerated; RAID 10 is appropriate for write-intensive workloads where the 50% capacity overhead is acceptable in exchange for write performance and shorter rebuild windows; RAID 50 or RAID 60 across multiple sub-pools (e.g. 2x RAID 6 of 12 drives, striped) balances rebuild scope against usable capacity. We discuss RAID layout in every 24-Bay quote.
Boot Drives
M.2 boot via the HPE M.2 SSD enablement option is strongly recommended at 24 bays - consuming 2 bays for OS boot mirroring wastes meaningful storage capacity in a configuration that exists specifically for high-density local SSD. M.2 boot mounts in a PCIe slot or on the CPU mezzanine board and frees all 24 SFF bays for data. Standard on our 24-Bay DL560 quotes unless customer specifies otherwise.
Storage Controllers at 24-Bay Scale
At 24 SFF bays, the storage controller decision matters more than at 8 bays - controller capability, RAID overhead, and write-cache sizing become primary design factors rather than secondary considerations:
- Smart Array P816i-a SR Gen10 (4 GB FBWC). The standard production controller for the 24-Bay configuration. 4 GB flash-backed write cache absorbs burst writes across the larger drive pool, tri-mode SAS/SATA/NVMe support handles mixed drive types. Full hardware RAID 0/1/5/6/10/50/60. Right pick for traditional hardware RAID across 24 SAS SSDs.
- Smart Array P824i-p MR Gen10 (4 GB cache, CacheCade). MR-series controller with 24 internal SAS lanes - the controller is purpose-designed for 24-drive configurations. CacheCade SSD acceleration accelerates HDD pools when the drive mix includes both SSD and HDD. The right controller when the deployment uses dedicated CacheCade SSDs to front a larger HDD pool, or when maximum lane count matters for sustained throughput.
- Smart Array E208i-a SR Gen10 plus additional E208 (HBA mode). For software-defined storage workloads (vSAN, Ceph, ZFS, Storage Spaces Direct) at 24-bay scale. Multiple HBA controllers may be required to deliver pass-through for all 24 bays; we'll spec the right combination at quote time based on backplane configuration.
- Smart Array P408i-a SR Gen10 (2 GB FBWC). Supported on the 24-bay configuration but the 2 GB cache is smaller than ideal for 24 SSDs under heavy write load. Acceptable for primarily read-heavy or moderate-write workloads; for write-intensive workloads the P816i-a is the better default.
The HPE Smart Storage Battery is required when any P-series performance RAID controller is installed. At 24 bays the battery is essentially mandatory - write workload at this drive count makes write-cache protection a hard requirement. We include the battery on every 24-Bay quote with a P-series controller. On P824i-p MR builds, confirm CacheCade SSD requirements at quote time - the MR controller's value depends on the CacheCade SSD configuration matching the workload's read/write profile.
NVMe at 24 Bays
The DL560 Gen10 supports up to 12 NVMe SSDs (half of the 24-bay backplane) with the right combination of NVMe drive cages, PCIe risers, and Smart Array controllers. NVMe at this drive count requires PCIe lane budget that competes with other expansion (the 4-port NVMe Mezzanine card on the CPU mezzanine board doesn't consume PCIe slots but is limited to 8 NVMe drives; beyond that requires PCIe slot consumption).
Common NVMe configurations on the 24-Bay DL560:
- 8 NVMe + 16 SAS/SATA SFF. NVMe via the mezzanine card (no PCIe slot consumption), remaining 16 bays via SAS/SATA on a P816i-a or HBA. Typical hot/warm storage tiering pattern.
- 12 NVMe + 12 SAS/SATA SFF. Maximum NVMe configuration via PCIe slot consumption for additional NVMe lanes. Higher-bandwidth tier for performance-critical local storage paired with bulk SAS/SATA. Verifies feasibility at quote time given competing PCIe demand.
- 24 SAS/SATA SFF. All-SAS or all-SATA configuration with no NVMe. Simpler PCIe planning; appropriate when NVMe-tier performance isn't the design requirement.
For most production 24-Bay DL560 workloads, all-SAS-SSD is the right answer - the per-drive performance of modern SAS SSDs is high enough that the NVMe step-up isn't required, and the SAS-only configuration simplifies PCIe planning meaningfully. If NVMe is a genuine workload requirement, we'll engineer the riser and controller combination at quote time.
Power and Cooling at 24-Drive Scale
A fully-populated DL560 Gen10 24-Bay with 4x Gold 6248 (4x 150W = 600W CPU), 48 DDR4 DIMMs (approximately 150-200W), and 24 SAS SSDs (approximately 240W) plus fans and overhead draws approximately 1,400-1,700W sustained. With 4x Platinum 8280 at 205W each (820W CPU alone) and 24 NVMe SSDs, the draw rises to approximately 1,800-2,000W sustained.
At this power envelope, 1600W Titanium PSUs are mandatory and 4-PSU configurations are typically required for production HA. Recall from the canonical that 4x 1600W PSUs and the tertiary PCIe riser are mutually exclusive - on the 24-Bay configuration, 4-PSU is typically the right choice (high TDP plus production HA) and the platform delivers 6 PCIe slots maximum rather than 8.
1600W Flex Slot Titanium PSUs require high-line input (200-240VAC) - confirm rack PDU and circuit capacity before deployment. We validate power budgets including PDU phase balance for every 24-Bay DL560 quote.
Thermal envelope: 24 SAS SSDs plus 4 high-TDP CPUs in 2U is thermally aggressive. Inlet temperature spec of 10°C to 35°C standard applies but at the upper limit (32-35°C inlet), confirm specific CPU SKU support against the HPE thermal matrix. ASHRAE A3/A4 support is configuration-specific at this density; we validate thermal headroom on every quote.
Honest Limitations
- Same 4-socket platform limitations as the canonical 8-Bay. 4x 1600W PSU and tertiary PCIe riser mutually exclusive; full 48-DIMM population drops memory speed one bin; 1st and 2nd Gen Xeon Scalable cannot be mixed; 4-socket TDP requires thermal validation; not a primary GPU compute platform. See the DL560 Gen10 8-Bay canonical for full Section 12 platform vocabulary.
- Universal Media Bay not supported with full 24-bay population. The media bay occupies the same physical position as one of the three drive boxes. Production 24-bay builds don't typically need the media bay; remote iLO 5 access covers the same operational requirements.
- NVMe at full 24-bay scale is PCIe-budget-limited. Maximum 12 NVMe drives requires PCIe slot consumption beyond the 4-port mezzanine card. NVMe beyond 8 drives competes with FlexibleLOM, storage controller, and any other expansion - we engineer this carefully at quote time.
- Storage controller decision matters more at 24 bays. The P408i-a (2 GB cache) is supported but undersized for write-intensive workloads at 24 SSDs. P816i-a (4 GB) is the standard recommendation; P824i-p MR for CacheCade-accelerated workloads. The wrong controller choice at 24 bays produces measurable performance loss under load.
- Single-PSU operation not appropriate. The 24-Bay DL560 draws 1.4-2.0 kW sustained - single PSU is not a production configuration at this power level. Take redundant PSU (2x or 4x 1600W) on every production build.
- Same Gen10 generational caveats apply. PCIe Gen3, DDR4-2933 ceiling, iLO Advanced licensing typically separate, FBWC battery as a wear item, third-party DDR4 limited to DDR4-2400 regardless of CPU. The DL380 Gen10 canonical and DL560 Gen10 8-Bay canonical cover these in detail and they apply identically here.
Workload Fit
| This server is right for | Consider alternatives for |
|---|---|
| ✅ SAP HANA with large local SSD persistence layer | ❌ 8 SFF bays sufficient alongside 4-socket (use DL560 8-Bay) |
| ✅ Oracle Database with local ASM diskgroups | ❌ Dual-socket sufficient with 24 SFF (use DL380 24-Bay) |
| ✅ SQL Server Enterprise with local tempdb/log on SSD | ❌ SAN-only storage architecture (use DL560 8-Bay) |
| ✅ vSAN ReadyNode at 4-socket consolidation | ❌ Budget-constrained projects |
| ✅ High-density in-memory analytics with local hot tier | ❌ Workloads requiring more than 12 NVMe drives |
| ✅ Storage Spaces Direct (S2D) at 4-socket scale | ❌ Primary GPU compute workloads (use Apollo) |
Where to Look Instead
- 8 SFF bays sufficient alongside 4-socket compute? → DL560 Gen10 8-Bay 2.5" (canonical) - same 4-socket platform at lower cost when local storage requirement is modest
- Dual-socket with 24 SFF bays? → DL380 Gen10 24-Bay 2.5" - 24 SFF capacity at the dual-socket tier, significantly lower cost
- Need 16 SFF bays at dual-socket? → DL380 Gen10 16-Bay 2.5" - the dual-socket sweet spot for medium-density SFF storage
- Dell shop alternative at the same 4-socket 2U tier? → Dell PowerEdge R840 - 2U 4-socket Purley on the Dell side, equivalent positioning, supports up to 24 SFF in similar configurations
- Need PCIe Gen4 and DDR5 at 4-socket? → Contact us for DL560 Gen11 availability when budget allows the generational step
Ready to Configure?
24-Bay DL560 configurations are sufficiently specialized that we recommend a design conversation before hardware selection. Tell us the workload (SAP HANA / Oracle / SQL Server / vSAN / analytics / S2D), licensing context, CPU and core target, memory target including any Persistent Memory requirement, storage architecture (drive type mix, RAID layout, NVMe requirement), controller preference (P816i-a vs P824i-p MR vs HBA), PSU redundancy preference, PCIe expansion requirements, and quantity. We respond within 24 hours with a validated configuration including thermal, power-budget, and PCIe-budget confirmation. Every refurbished unit ships with the Wholesale Servers 180-day warranty and 12+ hour burn-in testing, and volume pricing starts at 5 units. Call 1-800-778-1545 or use the quote form below.
Save Your Design
Click the Add to Quote button at the bottom of your screen to save your design as a draft order for future reference and to check for discounts, lead time, and availability. Most servers ship within 1-3 days.
