The XE9785L is a direct-liquid-cooled 3 OU compute node for Dell’s IR7044 and IR7050 Open Rack v3 racks; each node carries two AMD EPYC 9005 processors, 24 DDR5 slots and eight MI355X, B300 or B200 GPUs. It has no power supplies and no coolant distribution of its own, so this sheet covers the node and the rack, power shelf and CDU it relies on.
24 × DDR5 RDIMM · up to Not stated as a system total
Drive bays
Up to 16 · 3 drive-bay layouts
Expansion
12 × PCIe · 1 × OCP 3.0
GPUs
Up to 8 · 3 listed
Boot
Embedded BOSS / BOSS-N1
Management
iDRAC10
01 — Overview
The XE9785L at a glance
Our read of Dell’s documentation: where the XE9785L is strong, and the limits worth knowing before you spec one.
Strengths
+Three GPU boards on one node design — AMD MI355X OAM, NVIDIA HGX B300 NVL8 or HGX B200 — each carried under direct liquid cooling modules in the GPU tray.
+Liquid reaches the CPUs, the GPUs and, on the NVIDIA boards, the NVSwitch, and the rack manifold behind the node is rated to capture up to 10 kW per OU.
+Power and coolant both mate blind as the node is pushed home, through Barklip busbar clips and quick disconnects, so no per-node power cords are needed.
+At 94.6–95.4 kg a node is far lighter than the 163.65–172.3 kg air-cooled XE9785, yet keeps 24 DDR5 slots and up to twelve Gen5 x16 slots.
+Sixteen E1.S Gen5 NVMe drives at the front, or eight U.2 drives with two U.2 CEM cards alongside.
Watch-outs
!Not a standalone server: it needs an IR7044 or IR7050 ORv3 rack with a 1400 A busbar, 33 kW power shelves, blind-mate manifolds and a CDU from the list Dell sales holds.
!Much of the option list was still marked for a future release in this edition — six of the eight processors, the 256 GB DIMM, Red Hat Enterprise Linux and most OCP cards.
!Recochem PG25 is the only coolant Dell authorises, which sets the −5°C non-operating temperature floor, and DPU cards are supported only up to ASHRAE A2.
!B300 builds keep only PCIe slots 5–8, because slots 3, 4, 9 and 10 hold OSFP modules for the embedded CX-8 NICs.
!Rack elevations are given for reference only, and the guide never states how many XE9785L nodes a rack holds.
Processors8 SKUs
Cores / CPU48–192
DIMM slots24
Drive layouts3
Max bays16
GPUsup to 8
Rack power parts2 · to 33,000 W
02 — Processors
8 supported processors, mapped
The node’s processor table names the same eight EPYC 9005 SKUs as the air-cooled XE9785, but prints its own clocks and TDPs and marks six of them for a future release. In this edition only the 96-core 9655 and 64-core 9575F carry no asterisk.
Processor landscape — cores × TDP
8 SKUs · 1 family
AMD EPYC 9005 8Bubble size = L3 cache · hover a bubble for the full SKU
Processor explorer
filter · sort · compare
Processor
Cores / threads
Base
L3 cache
TDP
W / core
Memory
AMD EPYC 9965
192 / 384
2 GHz
384 MB
500 W
2.6
DDR5-6400
AMD EPYC 9845
160 / 320
2 GHz
320 MB
400 W
2.5
DDR5-6400
AMD EPYC 9825
144 / 288
2 GHz
384 MB
400 W
2.8
DDR5-6400
AMD EPYC 9755
128 / 256
2.6 GHz
512 MB
500 W
3.9
DDR5-6400
AMD EPYC 9655
96 / 192
2.6 GHz
384 MB
400 W
4.2
DDR5-6400
AMD EPYC 9575F
64 / 128
3 GHz
256 MB
400 W
6.3
DDR5-6400
AMD EPYC 9475F
48 / 96
3.6 GHz
256 MB
360 W
7.5
DDR5-6400
AMD EPYC 9455
48 / 96
2.9 GHz
256 MB
300 W
6.3
DDR5-6400
Figures on this page follow the XE9785L guide exactly: its single clock-speed column prints 2 GHz for the 9965, 9845 and 9825, and it gives 400 W for the 9845 and 9825 and 360 W for the 9475F. Dell’s XE9785 guide prints higher base clocks for seven of the eight parts and 390 W for the 9845 and 9825.
The thermal restriction tables clear four processors — 9965, 9755, 9655 and 9575F — on standard fans at 35°C, with MI355X plus twelve full-height cards or B300 plus four.
Each SKU is listed with 6400 MT/s memory and a 3 TB memory-capacity figure; this table carries no boost-clock column.
Both processors sit under direct liquid cooling modules served by rear quick disconnects 5 and 6.
03 — Memory
24 DIMM slots, 12 per processor
Twelve DDR5 channels per processor with one module each make 24 slots. The modules available in this edition are 96 GB and 128 GB, both dual-rank x8 at 6400 MT/s; a 256 GB module is listed for a future release.
Capacity with every slot filled
by DIMM size
Rated memory speed
from the processor table
Processor family
Up to
AMD EPYC 9005
6400 MT/s
Filled with 128 GB modules a node holds 3 TB (24 × 128 GB), the same number Dell prints as the memory capacity of every processor in its table.
Both the 96 GB and 128 GB modules are cleared on standard fans at 35°C in Dell’s MI355X and B300 thermal tables.
Registered ECC RDIMMs at 1.1 V only, and Dell cautions that the processor may run them below their rated speed.
04 — Storage
3 drive-bay layouts, up to 16 bays
Every drive is NVMe with no controller in the path: four front positions of four E1.S Gen5 drives (configuration C02) or two positions of four 2.5-inch U.2 Gen4 drives (C03). Dell’s matrix also has a C01 row with sixteen enabled slots but 0 × E1.S in each position. There is no PERC and no software RAID.
Headline layouts
front panel view
Front panel
Configuration C02: four front positions of four E1.S Gen5 NVMe drives, direct-attached. Dell’s C01 row shows the same sixteen enabled slots with 0 × E1.S per position.
SAS/SATANVMeSAS/SATA + NVMe
Controllers & boot
Embedded BOSS / BOSS-N1 — Boot. HWRAID 0/1, 2 x M.2 NVMe SSDs. Dedicated PERC: N/A; Software RAID: N/A; internal/external storage controllers: N/A - all drive configs C01-C03 list PERC Qty 0 / Storage Controller N/A (NVMe direct-attached)
Wistron U.2 CEM card — HBA. PCIe card-edge U.2 carrier, up to 2 (slots 8, 5 with MI355X; slots 6, 7 with B300); 'system supports two U.2 CEM on the PCIe expansion slot'
Embedded BOSS / BOSS-N1 (2 x M.2 NVMe SSD, HWRAID 0/1, dedicated heat sinks and thermal pads; acoustic config lists 1 x eBOSS 960 GB; eBOSS in slot 14 for MI355X/B200 configs, BOSS-N1 in slot 1 for B300 config) — boot
Each E1.S type — read-intensive, mixed-use, enterprise and data-centre read-intensive — comes in 3.84 TB or 7.68 TB, so 16 × 7.68 TB gives 122.88 TB raw.
U.2 choices are read-intensive at 3.84, 7.68 or 15.36 TB and mixed-use at 6.4 TB. Up to two U.2 CEM cards fit PCIe slots 8 and 5 with MI355X, or 6 and 7 with B300.
Boot runs from an embedded BOSS in slot 14 with MI355X or B200 (a pair of M.2 NVMe SSDs, hardware RAID 0 or 1), or from a BOSS-N1 in slot 1 with B300 (hardware RAID across two M.2 SSDs).
Empty E1.S slots, the OCP bay and unused PCIe slots all need blanks fitted.
Every documented layout
3 with drive bays
2.5-inch3.5-inchEDSFF (E1.S / E3.S)
05 — Accelerators & I/O
GPUs, PCIe and networking
Each node takes one eight-GPU board in its own GPU tray. The AMD option puts eight 288 GB Instinct MI355X OAM modules, rated 1400 W apiece, on Infinity Fabric; NVIDIA offers an HGX B300 NVL8 board (270 GB, 1100 W per GPU) or an HGX B200 board (180 GB, 1000 W), both NVLink-connected. All three sit under direct liquid cooling modules, with separate cold plates on the NVSwitch (B300, B200) and, for B300, on the CX-8 devices. Per node, 8 × 1400 W is 11.2 kW of GPU TDP with MI355X, 8.8 kW with B300 and 8 kW with B200.
Up to twelve Gen5 x16 full-height, half-length slots open on the node’s front face, and Dell’s expansion-card table pairs MI355X and B200 builds with SW1–SW4. B300 builds list only SW1 and SW2 and keep four slots, 5–8, with OSFP modules for the embedded CX-8 NICs in slots 3, 4, 9 and 10.
Network adapters Dell lists
25 options
Adapter
Ports
Form
Broadcom 2 x 100 GbE QSFP56 OCP 3.0 (*future) (OCP slot 13; B300 only - listed in Table 39, not in MI355X Table 38 or B200 Table 40)
2 × 100GbE
OCP 3.0
NVIDIA 2 x 100 GbE QSFP56 OCP 3.0 (*future) (OCP slot 13; MI355X, B300 or B200)
2 × 100GbE
OCP 3.0
Broadcom 4 x 25 GbE SFP OCP 3.0 (*future) (OCP slot 13; MI355X or B300 only - not listed for B200)
4 × 25GbE
OCP 3.0
Intel 2 x 25 GbE SFP28 OCP 3.0 (*future) (OCP slot 13; MI355X or B300 only - not listed for B200)
2 × 25GbE
OCP 3.0
NVIDIA 2 x 25 GbE SFP28 OCP 3.0 (*future) (OCP slot 13; MI355X or B300 only - not listed for B200)
2 × 25GbE
OCP 3.0
Broadcom 2 x 25 GbE SFP28 OCP 3.0 (OCP slot 13; MI355X or B300 only - not listed for B200)
2 × 25GbE
OCP 3.0
Intel 2 x 10 GbE BT (base-T) OCP 3.0 (*future) (OCP slot 13; MI355X or B300 only - not listed for B200)
2 × 10GbE
OCP 3.0
Broadcom 2 x 10 GbE BT (base-T) OCP 3.0 (OCP slot 13; MI355X, B300 or B200)
2 × 10GbE
OCP 3.0
Intel 4 x 10 GbE OCP 3.0 and Broadcom 4 x 10 GbE OCP 3.0 (listed only in the MI355X slot-priority table 38, slot 13)
4 × 10GbE
OCP 3.0
Intel 2 x 100 GbE OCP 3.0 and Intel 4 x 25 GbE OCP 3.0 (listed only in the MI355X slot-priority table 38, slot 13)
× 100GbE / 25GbE
OCP 3.0
AMD NIC 400 Gb Q112 (PCIe; up to 12 with MI355X)
× 400GbE
PCIe
AMD NIC 400 Gb V2 (PCIe; up to 12 with MI355X)
× 400GbE
PCIe
NVIDIA ConnectX-7 OSFP 400 Gb (PCIe; up to 12 with MI355X or B200)
× 400GbE
PCIe
NVIDIA N400 2-port NIC (PCIe; up to 12 with MI355X, up to 4 with B300)
2 × 400GbE
PCIe
NVIDIA SuperNIC 400 Gb (PCIe; up to 8 with B200, up to 4 with B300)
× 400GbE
PCIe
NVIDIA DP Q112 NIC (PCIe; up to 4 with B300)
2 ×
PCIe
Broadcom 2 x 200 Gb NIC (PCIe; up to 12 with MI355X, up to 4 with B300/B200)
2 × 200GbE
PCIe
NVIDIA 2 x 100 Gb NIC (PCIe; up to 12 with MI355X/B200, up to 4 with B300)
2 × 100GbE
PCIe
Broadcom 2 x 100 Gb NIC (PCIe; up to 12 with MI355X, up to 4 with B300)
2 × 100GbE
PCIe
Intel 2 x 100 Gb NIC (PCIe; two entries; up to 12 with MI355X, up to 4 with B300)
2 × 100GbE
PCIe
NVIDIA DPU 200 Gb (PCIe; two entries; up to 4 in slots 11, 2, 8, 5 with MI355X/B200 or slots 5, 8, 6, 7 with B300)
× 200GbE
PCIe
NVIDIA DPU 400 Gb (PCIe; up to four entries; up to 4 in slots 11, 2, 8, 5 with MI355X/B200 or slots 5, 8, 6, 7 with B300)
Broadcom Thor2 NICs (acoustic reference config: 8 x Thor2 in PCI 1 and 2 x Thor2 in PCI 2 with MI355X)
×
PCIe
Front I/O board 2 x RJ45 1 GbE dedicated iDRAC Ethernet ports
2 × 1GbE
embedded
Dell’s slot tables allow up to twelve network cards with MI355X — AMD 400 Gb, NVIDIA CX7 400 Gb OSFP, NVIDIA N400 or 100–200 Gb NICs — and up to eight NVIDIA 400 Gb SuperNICs with B200. DPUs, at 200 or 400 Gb, are capped at four in every configuration.
With B300 the add-in list shrinks to four cards: NVIDIA DPUs, SuperNICs or N400 and DP Q112 NICs, or Broadcom, NVIDIA and Intel 100–200 Gb NICs.
The OCP 3.0 slot is Gen5 x16 with up to four ports, NC-SI and shared-LOM redirect but no SNAPI. Only the Broadcom 2 × 25 GbE and 2 × 10 GbE cards lack a future-release mark in the card list, although the feature table marks 25 GbE port speed itself as future; the B200 table lists just two OCP cards.
In Dell’s IR7044 reference rack, east-west GPU traffic goes to a Z9864F-ON 800 Gb switch (or similar), OCP links to an S5248F and RJ45 to an S3248T; the IR7050 layout sends GPU and DPU links out to a separate network rack.
06 — Power & cooling
No supplies of its own — powered by the rack
The node carries no power supplies. Six 5500 W hot-swap modules in a 1 OU ORv3 power shelf make up its 33 kW output rating, with N+1 supported, and feed the rack’s 1400 A busbar; the XE9785L draws from it through two Barklip connectors at the rear — one for the GPU module, one for the CPU module.
AC
Each shelf accepts three-phase star at 346 or 480 V rated (311–528 V), or delta and single-phase at 200–277 V rated (180–305 V), 47–63 Hz. Input is AC only, on a seven-pin 60 A connector; the shelf itself has no input fusing (each supply carries internal 40 A fuses), and Dell recommends a 60 A upstream breaker per phase.
Other parts of the guide describe that input differently — star 346–415 V and delta 200–240 V line-to-line in the power-supply section, and 180–350 VAC at 26 A in its options table — so confirm the feed against the IR7000 Installation and Service Manual that Dell points to for power requirements.
Output is 33 kW at 49–50 V and 674 A. The busbar runs at the ORv3 48 V nominal, and elsewhere the guide describes the shelf output as 51 V DC.
Efficiency is 97.2% peak and 95.0% at full load, with 20 ms hold-up at 100% load. A shelf with six modules weighs 38 kg; each module measures 640 × 73.4 × 40 mm.
Dell’s IR7044 reference build uses seven shelves for 4-makes-3 busway redundancy; the IR7050 uses N+1 shelves with no 4-makes-3 option. Each shelf’s hot-pluggable Power Management Controller speaks Redfish and can be run from OpenManage Enterprise.
Shelves, supplies and controllers must not be mixed across vendors — Dell lists Delta and Lite-On 33 kW, 60 A shelves — and short pulse loads are allowed so long as any 10-second average stays within the rating.
07 — Management, security & OS
Run it, secure it, choose the OS
Management
iDRAC10 Datacenter (accelerator management plus air and liquid cooling control)
iDRAC Direct over USB 2.0 Type-C
iDRAC RESTful API with Redfish
RACADM CLI and IPMI
iDRAC Service Module
OpenManage Enterprise with Power Manager, Services, Update Manager and APEX AIOps Observability
Integrated Rack Controller (IRC) at rack level
Red Hat Ansible modules and Terraform providers
Security
System Lockdown (iDRAC10)
Secure Enterprise Key Management and Secure Component Verification (bundled with iDRAC10 Datacenter)
TPM connector on the HPM board
TPM specification v1.2 and v2.0 listed under standards compliance
Operating systems
Canonical Ubuntu Server LTS
Red Hat Enterprise Linux (future release)
08 — Compare
XE9680 → XE9785L
Dell’s XE9785L guide compares the node with the air-cooled 6U XE9680 rather than with its own air-cooled twin. The rows come from that table.
263.2 mm high · 482 mm wide · 1008.77 mm deep with bezel
140.5 mm high · 537 mm wide · 1047.95 mm deep with fans
Weight
107.75 kg (H100/H200/H800/H20, 16 × E3.S)
94.6–95.4 kg by GPU and drive layout
Management
iDRAC9
iDRAC10
Source: Dell PowerEdge XE9785L Technical Guide (Rev. A00, November 2025), Table 16 (comparison of PowerEdge XE9785L and XE9680).
The XE9785L family side by side
The XE9785L is the rack-scale, liquid-cooled counterpart of the 10U air-cooled XE9785. The XE9780L is the Intel version of this 3 OU node, built on 6th Gen Xeon Scalable processors and limited to NVIDIA B300 or B200 boards, and the XE8712 is another tray the same IR7044/IR7050 racks accept. The XE9680 is the model Dell’s guide measures it against.
This modelSiblings and successorEach value from that model’s own verified spec file · bars scale per column
Three node builds, one per workload Dell names for the XE9785L, each using only parts our configurator allows on it and each opening there for editing. The rack — shelves, manifolds, CDU and switches — is scoped with the whole deployment rather than per node.
Foundation-model training
MI355X node on processors shipping now
Foundation-model training heads Dell’s workload list for this node. The 96-core 9655 is one of the two processors without a future-release mark and is cleared on standard fans at 35°C with MI355X, while 24 × 128 GB fills every channel for 3 TB. Sixteen 7.68 TB E1.S drives stage training data locally.
2 × AMD EPYC 9655 — 96 cores @ 2.6 GHz, 400 W each
8 × AMD Instinct MI355X 288 GB OAM, liquid-cooled
24 × 128 GB DDR5-6400 RDIMM — 3 TB
16 × 7.68 TB E1.S Gen5 NVMe — 122.88 TB raw
Embedded BOSS: 2 × M.2 NVMe (RAID 1)
Broadcom 2 × 25 GbE SFP28 OCP 3.0 (slot 13)
Rack power: ORv3 33 kW shelf of 6 × 5500 W (shared)
Canonical Ubuntu Server 24.04 LTS
Needs an IR7044 or IR7050 with power shelves, blind-mate manifolds and a CDU from Dell’s supported list. GPU-network NICs go in PCIe slots 1–12 in the order Dell’s slot-priority table sets.
Fine-tuning is the second workload Dell names. The B300 board brings embedded CX-8 NICs whose OSFP modules sit in slots 3, 4, 9 and 10, leaving slots 5–8 for a DPU or storage adapter. The 64-core 9575F, the other processor without a future mark, is printed with a 3 GHz clock speed in this guide.
Inference rounds out Dell’s list. The B200 build pairs eight 180 GB GPUs at 1000 W with all twelve PCIe slots, and Dell’s slot table lets up to eight of them take NVIDIA 400 Gb SuperNICs. Eight 7.68 TB U.2 drives in the C03 layout match the B200 row of Dell’s weight table, printed as 94.6 kg and as 212.97 lb — two figures that do not agree.
2 × AMD EPYC 9655 — 96 cores @ 2.6 GHz, 400 W each
NVIDIA HGX B200: 8 × 180 GB SXM6 at 1000 W
24 × 128 GB DDR5-6400 RDIMM — 3 TB
8 × 7.68 TB U.2 NVMe Gen4 — 61.44 TB raw
Embedded BOSS in slot 14 (RAID 1)
Broadcom 2 × 10 GbE BASE-T OCP 3.0
Rack power: ORv3 33 kW shelf of 6 × 5500 W (shared)
Canonical Ubuntu Server 24.04 LTS
The B200 slot table lists only two OCP cards (NVIDIA 2 × 100 GbE and Broadcom 2 × 10 GbE) and no U.2 CEM card, and Dell’s thermal restriction tables cover only the MI355X and B300 layouts.
3 OU direct-liquid-cooled compute node (1 OU = 48 mm) · separate GPU and HPM trays
Racks
Dell IR7044 (44 OU) or IR7050 (50 OU), 21-inch Open Rack v3
Node dimensions
537 mm wide × 140.5 mm high · 1047.95 mm deep with fans, 889.65 mm without
Node weight
95 kg (MI355X or B300, 16 × E1.S) · 95.4 kg (8 × U.2 + 2 U.2 CEM) · 94.6 kg (B200, 8 × U.2; Dell’s matching 212.97 lb figure does not agree)
Rails
L-bracket static rails for the ORv3 frame · no CMA or strain-relief bar
Cooling
Liquid
DLC modules on the CPUs and the GPU board, plus NV-switch (B300, B200) and CX-8 (B300) cold plates · six rear quick disconnects
Air
Standard fans in the GPU and HPM trays · N+1 · fan and leak sensor board · count conflicts: overview says 4 UBB + 8 HPM fans, figures number GPU fans 1–8 and HPM fans 1–4
Coolant
Recochem PG25 only (freeze point −9 to −13°C)
Rack side
21-inch UQDB06 blind-mate manifolds, 2 QDs per OU, up to 10 kW per OU · in-row or in-rack CDU (IR7050: in-row only) · external rear-door heat exchanger listed
AC only · rated star 346/480 V, delta or single-phase 200–277 V · 47–63 Hz (Table 3)
Busbar
ORv3, 1400 A, 48 V
Ports & environment
Front
USB 3.0 Type-A · USB 2.0 Type-C (iDRAC Direct) · Mini-DisplayPort · 2 × 1 GbE iDRAC RJ45
Rear
None
Video
Matrox G200, 16 MB · up to 1920 × 1200 at 60 Hz
Thermal design range
10–35°C
Altitude
3050 m operating · 12,000 m non-operating
Non-operating temperature
−5°C to 65°C · 5–95% RH
11 — FAQ
XE9785L questions, answered from the documentation
Can the XE9785L run outside a Dell IR7000 rack?
Dell documents it only as a 3 OU node installed in the IR7044 or IR7050, Open Rack v3 racks with a 1400 A busbar and blind-mate liquid manifolds. The node has no power supplies of its own and takes both power and coolant through blind-mate connections at the rear, so it depends on that rack infrastructure.
What does the rack need besides the nodes?
Dell’s rack chapters list 33 kW ORv3 power shelves, the busbar, 21-inch UQDB06 blind-mate manifolds, an in-row or in-rack CDU (in-row only on the IR7050), a management switch such as the S5248 and fabric switches, plus rPDUs, power whips and an Integrated Rack Controller. The list of supported CDUs comes from Dell sales.
Which GPUs does the XE9785L support?
One eight-GPU board per node: AMD Instinct MI355X OAM (288 GB, 1400 W each, Infinity Fabric), NVIDIA HGX B300 NVL8 (270 GB, 1100 W) or NVIDIA HGX B200 (180 GB, 1000 W). The NVIDIA boards are NVLink-connected, and all three are direct-liquid-cooled in the GPU tray.
Which processors can a node use?
The guide lists eight 5th Gen EPYC 9005 SKUs from 48 to 192 cores, but at this Rev. A00 edition six carry a future-release mark. The 96-core 9655 and the 64-core 9575F, both 400 W, are the two without it.
How much memory can a node hold?
Twenty-four DDR5 RDIMM slots, one per channel. With the 128 GB module a node holds 3 TB, matching the 3 TB capacity Dell prints for each processor; a 256 GB module is listed for a future release.
What storage does the XE9785L take?
Sixteen E1.S Gen5 NVMe drives (3.84 or 7.68 TB) or eight 2.5-inch U.2 Gen4 NVMe drives, direct-attached with no RAID controller, plus up to two U.2 CEM cards in PCIe slots with MI355X or B300. Boot is a separate M.2 NVMe pair on an embedded BOSS or a BOSS-N1.
How is the XE9785L powered?
From the rack busbar. Each ORv3 power shelf holds six 5500 W modules, is rated at 33 kW output, supports N+1 and accepts three-phase or single-phase AC, and the node connects through two Barklip clips at the rear. Dell refers to the IR7000 Installation and Service Manual for detailed power requirements.
How big and heavy is one node?
It is 537 mm wide, 140.5 mm high and 1047.95 mm deep with its fans (889.65 mm without). Dell’s weight table gives 94.6 kg to 95.4 kg depending on GPU board and drives (its pound figure for the 94.6 kg B200 row, 212.97 lb, does not match), well beyond the 121 lb point at which Dell recommends a server lift.
How does it differ from the XE9785 and XE9780L?
The XE9785 is the air-cooled 10U version with its own twelve 3200 W supplies and a choice of MI355X or B300. The XE9780L shares this 3 OU rack-scale format but pairs Intel’s 6th Gen Xeon Scalable chips with a GPU choice limited to NVIDIA’s B300 and B200 boards, and Dell lists up to four x16 Gen5 slots for it.
Is the XE9785L end of life?
The November 2025 Rev. A00 guide gives no end-of-sale or end-of-support date, and many of its options were still awaiting release at that point. Our end-of-life checker records dates when Dell publishes them.
Sources & verification
Where these figures come from
Dell PowerEdge XE9785L Technical Guide (Rev. A00, November 2025) — Chapters 1–6 (IR7044/IR7050 rack, busbar, DLC manifold, power shelf, switches, rPDU); Tables 15–16 (new technologies, XE9785L vs XE9680); Tables 17–24 and Figures 36–46 (views, dimensions, weight; images from Figures 37, 40 and 45); Tables 25–40 (processors, memory, storage, OCP, PCIe slot priorities); Tables 41–43 (power, acoustics); Tables 44–54 (iDRAC, ports, environment, thermal restrictions).
Specifications are extracted from Dell’s published documents and independently re-checked line by line. Configurations are quote-based; availability and exact options are confirmed at quotation. Last reviewed 2026-09-11.
