Dell PowerEdge R940 24-Bay 2.5"
Configure Your System:
1 Processor Required
On series: four-socket operation is what separates Gold and Platinum from the rest of the Xeon Scalable range, so the list here starts where two-socket lists end. 6100-series parts are 1st Gen (Skylake-SP); 6200 and 8200-series are 2nd Gen (Cascade Lake-SP) with higher clocks per tier and Optane PMem support. Both generations are proven in this chassis.
On selection, match the CPU's character to the workload rather than chasing the biggest number. Transactional databases and ERP platforms reward clock speed and cache: the frequency-optimized Golds are often the faster real-world choice for them than a bigger, slower Platinum. Massive consolidation and in-memory platforms reward core density and memory bandwidth, which is where the high-core-count Platinums earn their premium. Four of anything on this page is a serious amount of compute; the question is what shape.
All four CPUs must match exactly: same model, same stepping. We build them that way, and it is worth remembering if you ever expand a partially populated unit.
Not sure which shape your workload wants? Tell us what it runs at quote time, or call 1-800-778-1545 and we will spec it against the actual application.
2 Heat Sink Optional
The standard heatsink covers the moderate-wattage CPUs in the list in a normal datacenter environment.
The performance heatsink is required for the top-wattage CPUs, and a four-socket buyer is rarely shopping the bottom of the list; plan on performance heatsinks on all four sockets for any seriously configured unit. We also spec them on densely populated NVMe configurations and at ambient inlet temperatures above 30°C.
The cost difference is small against the price of the processors they protect, and it is a lot cheaper than a thermal event on the most important server in the building. Confirm all four are in the quote and move on.
3 Memory (RAM) 48 DIMM slots, modules added in sets of 4 Required
Fill all 48 slots. On this platform the case is performance first: full population keeps every channel on every socket working, which is exactly what the in-memory databases, ERP platforms, and analytics engines this chassis attracts are bandwidth-hungry for. Cost per gigabyte follows the same logic as everywhere else: 48 smaller modules reach a given capacity for meaningfully less than 24 larger ones.
One rule that matters more here than on any two-socket server: keep the configuration identical across all four sockets. These workloads are NUMA-sensitive, and an unbalanced memory map means some cores wait on remote memory while others do not. One matched set, evenly distributed, is the whole discipline.
768 GB (48× 16 GB) is the sensible floor for a server in this class
1.5 TB (48× 32 GB) is our default, and the best cost per GB in the lineup
3 TB (48× 64 GB) for large in-memory databases and analytics
Past 3 TB you are into 128 GB LRDIMMs, and the R940 is one of the few platforms where that conversation is routine rather than exotic; buyers at that capacity chose this chassis precisely because the workload cannot be split. Call 1-800-778-1545 and we will spec it.
4 RAID Controllers Optional
PERC H740P (8 GB NV cache, battery-backed) is the default, and on this platform it is barely a decision. Transactional databases, ERP platforms, and analytics scratch space all lean on the controller's non-volatile cache, and the H740P's 8 GB is what keeps local storage out of the workload's way. Against the price of the rest of this configuration, choosing a smaller controller is saving in exactly the wrong place.
PERC H730P (2 GB cache) is defensible only when local storage is an afterthought: a boot volume and little else, with the real data on a SAN or NVMe. If that describes your build, it will do the job.
Entry controllers do not belong in this chassis, full stop.
Mind the array math wherever the data lands locally: RAID 1 or 10 for transaction logs and hot data, RAID 6 for large-capacity volumes, consistent drive models within each array, and a hot spare on any array this business depends on.
Running your data path over NVMe instead? Those drives connect over PCIe and bypass the controller entirely; redundancy happens in software there, and the controller here only matters for whatever SAS or SATA drives remain. Call 1-800-778-1545 and we will spec the whole storage path against the workload.
5 Storage Drives Select up to 24 drives (0/24 Slots Used) Optional
Buy SSDs, without much debate on this platform. The workloads that justify four sockets are latency-sensitive, and SAS SSDs are the production pick for the controller-attached bays: higher endurance, dual-port connectivity, and sustained-write behavior that holds up under the load this chassis exists to carry. Spinning disk has no real role here; if bulk capacity is part of the picture, it belongs on a storage platform like the R740xd, not in these bays.
The R940 also takes direct-attach NVMe in a portion of its bays, and for databases and in-memory platforms that is where the chassis earns its reputation: PCIe-attached flash feeding four sockets is the configuration to beat. On NVMe, mind endurance ratings: mixed-use (3 DWPD) where the writes land, read-intensive (1 DWPD) for capacity tiers. Remember NVMe bypasses the RAID controller, so redundancy for those drives is planned in software.
Buy new, and on this server treat it as a rule rather than advice. A refurbished drive arrives with unknown miles on its endurance budget, and this is the last chassis on which to discover that mid-quarter. Our refurbished drives have honest roles elsewhere in the catalog; this is not one of them.
Boot belongs on a Dell BOSS card, available in the Add Ons section below, keeping every front bay on mission duty.
Tell us the workload at quote time and we will spec the split between NVMe and SAS, or call 1-800-778-1545.
6 Remote Access Required
Take Enterprise. The R940 is bought to carry the workloads a business cannot lose, and iDRAC9 Enterprise is the tool that keeps a bad night short: full remote KVM and virtual media mean you can reach the server's screen, mount an ISO, and rebuild from anywhere, even when the operating system is down. Express lacks the remote console, so any problem below the OS level means someone physically at the rack, and the workloads this chassis carries do not wait politely for the drive in.
Enterprise also brings automated firmware updates, two-factor authentication, and secure erase, which matters on a server that will retire holding the most sensitive data in the building.
The license is a rounding error against this configuration's price, and the first incident it shortens pays for it many times over. This is the easiest recommendation on the page.
7 Power Supply 2x included in build Required
Add it up before choosing: four CPUs at full tilt, 48 DIMMs, and a bay full of flash is a fundamentally different load than anything a two-socket server presents. The lower-wattage pairs that anchor other configurators are not the default here.
The 1600W pair is our default for a seriously configured R940: four high-wattage CPUs, full memory, and a loaded backplane, with the 50 percent load target intact.
Step down a tier only for genuinely moderate builds: lower-wattage CPUs and restrained storage.
Step up to the top tier when the build maxes every axis at once, or when your datacenter's power redundancy scheme demands that one PSU carry the entire load alone with margin to spare, which is exactly the scheme most facilities running four-socket flagships enforce.
Size for roughly 50 percent load at your expected peak: a PSU at half capacity runs at peak efficiency, runs cooler, and lasts longer than one working near its limit. When the build sits between tiers, size up; against this build sheet the price step is a rounding error. Watch the estimated TDP counter at the bottom of the page as you add components to see where your build lands.
8 Network Cards Required
A four-socket flagship concentrates a rack's worth of work into one box, which concentrates a rack's worth of traffic into this card. Treat 10 Gb as the minimum: the quad 10 Gb options (Intel X710, X550, Broadcom 57840S) are the default recommendation, giving production, backup, and management traffic their own links. The 2x 10 Gb + 2x 1 Gb combo cards are the economy pick only when a SAN carries the storage traffic on its own fabric. Quad 1 GbE has no business on this platform; the server will saturate it before breakfast.
If the R940 is serving as a database or analytics engine for many clients, consider 25 GbE from the Add Ons section; the card is a small line item on this build sheet, and it removes the network from the list of things that can bottleneck four sockets.
Then port type, and this is the one that catches people. SFP+ ports take fiber or DAC cables and talk to datacenter-grade switches. The -T models (X540, X550) are 10GBASE-T copper on ordinary RJ45 cabling. 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.
Call 1-800-778-1545 and we will spec the whole path, switch to server.
9 Operating System Required
Windows Server 2022 is the version to buy today, and on this chassis the edition question matters more than usual: Datacenter edition licenses unlimited virtual machines on the host, and at the consolidation density a four-socket flagship invites, Datacenter usually beats Standard's per-VM math decisively. Windows licenses by core across all four sockets, so the core counts you chose in the processor step flow straight into the license bill; we will price both editions against your configuration at quote time.
Fair warning on Evaluation Editions: free for 180 days, then they expire. Labs only. If the server is going into service, buy the license now.
VMware ESXi 8.0 is the one to deploy for virtualization duty; 6.7 and 7.0 are past end of support and only belong on hosts joining an existing cluster. Licensing is yours to bring, and at four sockets it is a serious line item; price it before you commit to the platform.
Ubuntu Server LTS and the enterprise Linux family cover the database, ERP, and analytics duty this chassis is frequently bought for, with five years of free updates on the LTS side.
Proxmox VE runs well here for buyers consolidating away from VMware economics, though most four-socket buyers arrive with the platform already decided.
No OS is there for teams imaging from their own deployment infrastructure.
Tell us the workload at quote time, or call 1-800-778-1545.
WServer Warranty
Dell PowerEdge R940 24-Bay 2.5" Drives [14th Gen] Detailed Review
The Dell PowerEdge R940 24-Bay 2.5" is the maximum-density configuration of Dell's 14th-generation 4-socket 3U scale-up platform: twenty-four 2.5" hot-swap front bays paired with up to four 2nd Generation Intel Xeon Scalable processors, up to 48 DDR4 DIMM slots, and the R940's full 13-slot PCIe budget. This is a refurbished enterprise platform, fully tested and ready for the rare workload that genuinely needs 4-socket compute, dense local storage, and heavy expansion in one chassis.
The 24-bay front is what separates this configuration from the rest of the family. Where the 8-Bay R940 keeps local storage minimal and pushes bulk capacity to SAN, the 24-Bay puts it on the front of the chassis: persistence layers, warm data, and dataset staging that need to sit next to the compute rather than across a fabric. It is a specialist build, and the storage density is the whole reason to choose it over the 8-Bay.
To configure a build, call 1-800-778-1545 and our account team will scope processors, memory population, the storage controller layout across 24 bays, and PCIe allocation against your workload. Every chassis ships with our 180-day warranty after a 12+ hour burn-in, and volume pricing starts at 5 units. We return a validated configuration within 24 hours.
When 24 Bays Is the Right Configuration
The 24-Bay R940 earns its place only when three demands land at once: 4-socket scale-up compute, dense local SFF storage, and heavy PCIe expansion. That combination is real but uncommon. SAP HANA with a large local SSD persistence layer plus Fibre Channel SAN connectivity; Oracle at 4-socket scale with active data on local SSD and dedicated HBAs for shared storage; large-scale analytics that need GPU compute, NVMe expansion, and high-density SSD staging together. If any one of the three is more than your workload needs, a simpler configuration delivers better economics. This is the page for the build that needs all three.
Storage - 24 SFF Bays
Twenty-four 2.5" SAS/SATA hot-swap bays, with NVMe support on a subset of bays depending on backplane and PCIe allocation. At full population with 7.68 TB SSDs the front delivers roughly 184 TB of raw local capacity, which is what makes this configuration viable as a self-contained persistence-plus-compute platform rather than a SAN client. The 24-bay backplane is the defining feature; everything else on this page is shared with the rest of the R940 family.
For boot, we quote the Dell BOSS-S1 card: dual mirrored M.2 SATA SSDs on a dedicated controller as a hardware RAID 1 volume. On a 24-bay chassis the BOSS card matters even more than on the 8-bay, because dedicating two of your twenty-four valuable front bays to the operating system is a waste of storage density. BOSS keeps the OS off the front entirely and frees all 24 bays for data.
Storage Controllers
The R940 accepts full-height PERC adapters only, in slot 1 (primary) and slot 6 (secondary), with no mini-PERC support. On a 24-bay build this constraint is more visible than on the 8-bay: driving 24 drives often means a dual-controller layout, and both controllers consume those dedicated full-height slots. Plan the controller and HBA topology before you finalize the PCIe card list.
The lineup: PERC H740P (8 GB NV cache, battery-backed) is our production default for write-intensive RAID across the dense front; PERC H730P (2 GB cache, battery-backed) for mixed or read-heavy work; PERC H840 when you also need to drive external SAS enclosures beyond the 24 internal bays. For software-defined storage (vSAN, Storage Spaces Direct, Ceph), the HBA330 pass-through HBA presents drives directly to the stack. We do not quote S140 software RAID for a production storage array of this size; it is a dev and test option only.
Processors
Up to four 2nd Generation Intel Xeon Scalable (Cascade Lake-SP) processors, up to 28 cores each, for as many as 112 cores and 224 threads at full quad-socket population. The platform rule that governs every R940 applies here too: a 2-CPU build runs without the Processor Expansion Module as a 2-socket, 24-DIMM, 7-PCIe-slot machine. The PEM installs automatically with 4 CPUs to unlock 48 DIMM slots and all 13 PCIe slots. A 24-bay storage build almost always wants the full platform, so plan on four processors.
Our common balanced specification is four Gold 6230 (20 cores, 2.1 GHz, 125W) for 80 cores; for maximum per-socket throughput we quote the Platinum 8260 (24 cores, 165W) or the top-bin 28-core parts. On the high-TDP SKUs, confirm the high-performance heatsink and fan configuration is specified. A missed heatsink on a 165W-plus CPU is the most common stability problem we see on scale-up builds under sustained load, and a fully populated 24-bay chassis runs warm.
Memory
Six channels per CPU, 2 DIMMs per channel, for 12 slots per socket and 48 slots across a fully populated 4-socket system. Maximum capacity is 6 TB with 64 GB LRDIMMs. Adding Intel Optane DC Persistent Memory takes the platform to up to 15.36 TB combined across DCPMM and LRDIMM, which pairs naturally with the 24-bay storage density for HANA and large in-memory database hosts that also keep a substantial persistence layer local.
Populate all six channels per CPU evenly, either 6 or 12 DIMMs per socket, for best performance. Registered ECC DDR4 only (RDIMM or LRDIMM). RDIMM gives the best balance of frequency, capacity, and rank flexibility; LRDIMM is the path to the 64 GB-and-larger modules that reach the 6 TB ceiling. Remember the PEM rule: half the DIMM slots sit on the expansion module, so a 2-CPU build is capped at 24 DIMMs regardless of the storage configuration.
Networking and PCIe Expansion
Networking runs through a Flexible Rack Network Daughter Card (rNDC) that does not consume a PCIe slot. Options: 4x 1GbE, 4x 10GbE, 2x 10GbE plus 2x 1GbE, or 2x 25GbE. For a storage-dense scale-up host we typically quote 2x 25GbE or 4x 10GbE so the network does not bottleneck the local storage.
PCIe is where the 24-bay build gets tight. The platform offers 7 slots with two processors and 13 with four (the Processor Expansion Module adds slots 8 through 13). But on a 24-bay configuration, the two full-height PERC slots are often spoken for by storage controllers, so budget the remaining slots carefully across FC HBAs, cluster interconnects, NVMe expansion, and any GPU. This is precisely the architecture where the R940's slot count over the 2U R840 pays off, and precisely the build that needs all four processors to get there.
GPU Support
Depending on riser configuration, the R940 accommodates up to 4 double-width GPUs or up to 8 single-width GPUs. On a 24-bay storage build, GPU and storage-controller slot demands compete for the same finite budget, so a GPU-heavy plus storage-dense configuration needs careful slot planning up front. For GPU-first database acceleration with maximum front storage, the dedicated R940xa (a separate 4U platform with a 1:1 CPU-to-GPU design and 32 front bays) is the better-matched chassis. The 24-bay R940 covered here is the compute-and-storage-first 3U server.
Management - iDRAC9 Generation
iDRAC9 with Lifecycle Controller, Enterprise license required for production remote management. Silicon Root of Trust for boot integrity, with TPM 2.0 module support required for NIST, CMMC, FedRAMP, HIPAA, and PCI DSS frameworks. Quick Sync 2 enables at-the-rack management over Bluetooth, and the platform integrates with Dell OpenManage Enterprise for fleet management. Specify TPM 2.0 up front in regulated environments. The iDRAC9 platform here is identical to the rest of the R940 family; see the R940 8-Bay page for the full management feature walk-through.
Power and Cooling
Dell hot-swap redundant PSUs in 1100W, 1600W, 2000W, and 2400W tiers as a 1+1 redundant pair. A 24-bay R940 at full configuration is one of the heaviest power draws in the 14th-gen portfolio: 4-socket compute, 24 drives, full memory, and a loaded PCIe complement all on the same chassis. Size for the worst case and validate rack power before deployment.
| Configuration | PSU Recommendation | Est. Peak Draw |
|---|---|---|
| Balanced (4x Gold 6230, full RAM, 24x SSD) | 2x 2000W Platinum | ~1500W |
| Performance (4x Platinum 8260, full RAM, dual HBA, 24x SSD) | 2x 2400W Platinum | ~1850W |
| Heavy (4x top-bin, full RAM, GPUs + 24x SSD) | 2x 2400W Platinum | ~2200W |
At the heavy end this chassis can approach the limit of a standard rack circuit. Confirm PDU capacity and cooling headroom before specifying.
Physical Specs & Platform Notes
- Form factor: 3U rack server, the more expandable successor to the 4U R930. The 24-bay front makes this the deepest and heaviest configuration of the platform; confirm rack depth.
- PCIe expansion: 7 slots in a 2-socket build (slots 1-7), 13 slots in a 4-socket build with the Processor Expansion Module. On a 24-bay build, two full-height slots are typically consumed by storage controllers.
- Parts availability: Strong. Shares the broad 14th-gen PowerEdge parts ecosystem, with refurbished components, rails, and accessories readily sourced. Third-party maintenance is the standard production support path in 2026 as Dell ProSupport on 14th gen approaches end of extended support.
- Accessories we recommend: the Dell BOSS-S1 boot card (especially important here to preserve all 24 front bays for data), the LCD bezel for status and Quick Sync 2 management, 3U sliding rails, and a cable management arm given the chassis depth.
- Platform notes: Full-height PERC only in slots 1 and 6, no mini-PERC. The Processor Expansion Module is mandatory for 4-socket operation and installs automatically with 4 CPUs. No 3.5" LFF support; this is a 2.5" SFF chassis.
Our Assessment
Where it excels: The 24-Bay R940 is the right answer for the rare workload that needs 4-socket compute, dense local SFF storage, and heavy PCIe expansion simultaneously. SAP HANA with a large local SSD persistence layer plus FC SAN connectivity, Oracle at 4-socket scale with local active data and dedicated HBAs, and analytics builds that combine GPU, NVMe expansion, and SSD staging are the configurations that justify it. The 24-bay front is what makes it self-contained rather than SAN-dependent.
Where to look instead: If 8 front bays cover your local storage, the R940 8-Bay 2.5" is the same platform without the storage premium. If two sockets are enough and you mainly need the 24 bays, the R740xd 24-Bay 2.5" delivers dense SFF storage on a dual-socket platform at far lower cost. If you want 4-socket plus dense storage but in 2U, the R840 24-Bay 2.5" trades PCIe headroom for rack density.
Bottom line: This is one of the most capable single-chassis configurations Dell built in the 14th generation, and one of the most specialized. Every component (the four sockets, the 24 bays, the deep PCIe budget) has to be justified by the workload, because each adds cost and complexity. When the workload genuinely needs all three, nothing in the 14th-gen lineup consolidates them better. When it does not, one of the alternatives above will serve at lower cost. These builds start with an architecture conversation, not a configurator.
Where the R940 Fits in 2026
The R940 is a 14th-generation platform, two generations behind 15th-gen scale-up hardware and three behind current 16th-gen. For greenfield mission-critical deployments with long support horizons, evaluate newer generations. For refurbished scale-up procurement in 2026 where the workload is well understood, the 24-Bay R940 delivers 4-socket Cascade Lake compute, 6 TB of memory (15.36 TB with Optane), 184 TB of raw local SFF capacity, and a 13-slot PCIe budget at a fraction of new-platform cost. Plan on third-party maintenance as the production support path, as Dell ProSupport on 14th gen is approaching end of extended support.
Honest Limitations
- The full-height-PERC-only constraint (slots 1 and 6) bites hardest here: dual storage controllers for 24 drives often consume both slots, tightening the budget for HBAs, interconnects, and GPUs.
- This is the most power-hungry and physically deepest R940 configuration. A fully loaded build can approach the limit of a standard rack circuit; rack depth, PDU capacity, and cooling all need confirming before commitment.
- A 2-CPU build drops to 24 DIMM slots and 7 PCIe slots without the Processor Expansion Module. A 24-bay storage build almost always needs the full 4-socket platform to be worth it.
- It is a specialist platform with specialist economics. If your workload does not need all three of compute, storage, and expansion at once, you are paying for capability you will not use.
- NVMe is supported only on a subset of bays depending on backplane and PCIe allocation, not across all 24 by default. Confirm the NVMe bay count your design needs up front.
Workload Fit
| The R940 24-Bay is right for | Consider alternatives for |
|---|---|
| ✅ 4-socket + 24 SFF bays + heavy PCIe at once | ❌ Any single one of those needs (simpler config wins) |
| ✅ HANA scale-up with local SSD persistence + FC SAN | ❌ 8 front bays sufficient (use the R940 8-Bay) |
| ✅ Oracle at 4-socket with local SSD + dedicated HBAs | ❌ Dual-socket sufficient (use the R740xd 24-Bay) |
| ✅ Analytics with GPU + NVMe expansion + SSD staging | ❌ 4-socket + storage but 2U needed (use the R840 24-Bay) |
Where to Look Instead
- Less local storage, same platform: Dell PowerEdge R940 8-Bay 2.5" is the mainstream R940 without the 24-bay storage premium.
- Dual-socket, storage-dense: Dell PowerEdge R740xd 24-Bay 2.5" for 24 SFF bays when two sockets are enough.
- 4-socket plus storage in 2U: Dell PowerEdge R840 24-Bay 2.5" trades PCIe headroom for rack density.
- 4-socket, minimal storage: Dell PowerEdge R840 8-Bay 2.5" for compute-first 4-socket builds in 2U.
- Cross-vendor counterpart: the closest HPE scale-up platform in our catalog is the HPE ProLiant DL580 Gen9, a 4-socket flagship. It is a 4U Gen9-era platform with far fewer front bays, so cross-shop it on workload fit and budget rather than as a like-for-like storage match.
Ready to Configure?
R940 24-Bay configurations start with an architecture discussion, not a configurator. Tell us your workload, your full PCIe card list, your HANA or Oracle licensing context, your memory target, and your rack power infrastructure. Call 1-800-778-1545 and our account team will return a validated 24-Bay configuration within 24 hours. Every build carries our 180-day warranty and a documented 12+ hour burn-in, with volume pricing from 5 units. This is a configured, quoted scale-up platform: we scope it with you before you commit, never a retail checkout.
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.
