HPE ProLiant DL385 Gen11 Front GPU Cages: Removal and Replacement Example

The source shows an HPE ProLiant DL385 Gen11 GPU Configure-to-order Server with GPU cages at the front. Configurations differ; use the service guide and cable diagram for the actual installed hardware. Two overview photos with management labels are replaced here by unchanged official same-model English diagrams. Other original detail photographs remain in their source order.

1. Locate the front GPU cages

The official 8-SFF front view identifies cage 1 at callout 2 and cage 2 at callout 4, with drives between them. This establishes front-cage placement without reproducing the original management label:

Official DL385Gen11 front view for8SFFdrives with two GPU riser cages; callout2is cage1and4is cage2

The original cover label gives the removal/installation diagrams and slot map. It marks left-cage slots 9–12 and right-cage slots 13–16; use the physical label and exact configuration, including the shown 180-degree right-cage orientation:

Original English chassis-cover label showing GPU-cage removal and left/right slot numbers

2. Prepare and disconnect the cage

Before internal work, follow HPE’s service preparation: protect server data, shut down, disconnect every external power cord and peripheral cable, allow hot parts to cool, and apply ESD precautions. Remove the required access/middle covers, air baffle and fan cage as specified for the configuration. A T-10 Torx screwdriver is required for the referenced GPU replacement. This is not a hot-swap GPU procedure.

The source opens the blue release latch, disconnects the GPU power connections and GPU/riser-to-system-board connections, then slowly pushes the cage from the rear toward the front. Its annotated overview describes two information connections per cage and a further connection for each left/right cage; those are source-photo counts, not universal port identities. The official procedure requires disconnecting all applicable GPU and front-riser cables before withdrawal.

Original close-up of the two boxed cable connections in the source cage example

The replacement overview below is HPE’s GPU-riser auxiliary-power routing diagram, not a translation of the source photo’s signal-cable annotation. It shows the riser power branches for the illustrated configuration:

Official same-model GPU-riser auxiliary-power cabling example; it does not identify every signal cable

A separate official illustration shows the4-dual-width/6-single-width GPU auxiliary-power arrangement. Match actual connectors, cable PNs and installed GPU options to the service guide; color and position alone are insufficient:

Official same-model4dual-width or6single-width GPU auxiliary-power cable routing

The source blue latch and cover retaining screw are shown in this detail:

Original photo highlighting the blue release/retaining hardware

3. Open the removed GPU cage

The English instruction label on the GPU cage distinguishes single-width and double-width installations. The source calls this a CPU cage in prose, but the pictured label and operation concern the GPU cage:

Original English Installing GPU label on the GPU cage

Release the cover retaining screw and remove/unlock the cover through the applicable illustrated motion. The following photos show the cover screw and opened cage:

Original blue-boxed GPU-cage cover screw

Original opened GPU cage with riser slots exposed

4. Remove the GPU fasteners and transfer the required support

Remove the front and rear GPU retaining screws for the fitted GPU type and keep them safely. Screw/bracket arrangements differ by GPU model; the following photos identify the source’s two ends:

Original GPU end bracket with two retaining screws highlighted

Original opposite GPU end with its retaining screw highlighted

The source transfers the rear metal support bracket to the replacement GPU. Retain the bracket and its specified screws when required by the matching replacement kit; do not discard them or assume every new GPU has identical mounting hardware:

Original removed GPU with highlighted reusable metal support bracket

Original repeated retaining-screw detail from the source

5. Reassemble and verify the actual fault indication

Install the supported replacement, reconnect the verified cable paths and refit the fan assembly and other removed parts through the matching service procedure. Check each fan module is fully seated; an unseated fan can prevent normal power-on. Keep fingers clear of the dense cabling and thin CPU-heatsink fins, which can cause cuts.

The source suggests checking power cabling when error 490 is reported, but its actual capture says Insufficient Fan Solution followed by 490 - System Health Error. Treat 490 as the pictured general system-health indication and inspect the IML and the full fault text; it does not identify a power cable as the sole cause:

Original English Insufficient Fan Solution and490System Health Error capture; check IML and fan seating

The stored source does not include a successful post-replacement test result. Verify the actual IML, fan population/seating, GPU detection, cabling and hardware health before returning the server to service. No physical work or server action was performed during translation.