What Happens After Your DGX Station Arrives? The Enterprise Installation Checklist

Once a DGX Station GB300 arrives, the work shifts from installation to acceptance. Inspect the delivery and check the configuration against the quotation. Register the warranty with the right entity, run the tests you agreed in the purchase order, take over every credential and collect the documentation. Then sign off. Doing this properly matters more than usual, for three reasons. NVIDIA publishes no burn-in or acceptance procedure for the GB300 Station. Warranty and support come entirely from the OEM. And Dell's service manual implies that a fault at the superchip level means replacing the whole chassis. This checklist starts at the loading bay. Site power, placement and lifting are in our office installation checklist, and everything up to delivery is in from quotation to delivery. The suggestions below draw on NVIDIA and OEM documentation as of 9 Nov 2026; they are not an NVIDIA procedure.

What should happen between delivery and sign-off?

Source: Dell Pro Max with GB300 FCT6263 Setup and Specifications and Service Manual (en-sg, 22 Sep 2026); NVIDIA DGX Station Development Guide, software stack page (updated 30 Oct 2026); OEM specification pages for HP, ASUS and Supermicro; NVIDIA DCGM and MIG documentation. Fetched 9 Nov 2026. The stages are suggestions to agree with your supplier, not a published NVIDIA acceptance test.

What do you check when the crate arrives?

Treat it as a high-value delivery. Before signing, photograph the packaging and note any damage on the delivery note. The units are heavy: ASUS lists 27 kg net and 32 kg gross, and Supermicro 40 kg net and 51 kg gross. Every DGX Station is liquid-cooled, so lift it as the OEM instructs, with two people, and keep it upright. Match the serial number and part number on the box against the purchase order and quotation. Warranty registration depends on them.

How do you confirm you got the configuration you ordered?

Several things vary by OEM and order, so check them before power-on.

• Drives and RAID. NVIDIA describes two OS drives in software RAID 1. Data drives differ: HP's specification lists RAID 0 data volumes, and Dell's configurator shows RAID 1 on two drives and no RAID on the other two. RAID 0 or non-RAID data means a single drive failure loses that data, so note it for your backup plan.
• Display card. Confirm which RTX PRO card, if any, is fitted. It changes the GB300's power cap under NVIDIA's power-sharing scheme, from 1,600 W down to 1,250 W or 950 W depending on the card.
• Power cord. Dell's en-sg manual illustrates only US NEMA plugs and says to use the cable shipped with the computer, so confirm you received the right local cord.
• Networking accessories. Check that any QSFP cables or transceivers on the quotation for the 400 GbE ports are in the box.

Which acceptance tests should you run?

NVIDIA publishes no burn-in or acceptance procedure for the Station, so agree the tests in the purchase order and run them together with the supplier. A practical minimum uses documented commands. The first two confirm that DCGM sees the GPU and list the MIG profiles your unit supports. The last two confirm that the model you agreed on is being served:

$ dcgmi discovery -l
nvidia-smi mig -lgip
curl --fail-with-body -i http://localhost:8000/health
curl --fail-with-body -sS http://localhost:8000/v1/models

Sources: DCGM Getting Started; MIG User Guide; NVIDIA DGX Station vLLM playbook. If the supplier commissions a model, agree beforehand on which model, at what precision and with what response you expect, and record the result. The engineer's first-session steps are in our DGX Station GB300 setup guide.

Acceptance is the last point where problems are clearly the supplier's to fix. Write the tests into the purchase order, run them before sign-off and keep the output.

Who should hold the keys at handover?

By the end of handover, your organisation should hold every credential and the supplier none, unless a support arrangement says otherwise. Dell forces a BMC password change at first login, but other OEMs' defaults aren't published, so confirm each one was changed. Record the BMC IP address, the OS administrator accounts and any API keys created during commissioning. Store them sealed, then rotate any password the supplier's engineer used. Our security hardening checklist covers what to lock down next.

How should the warranty be registered?

All hardware warranty for the DGX Station GB300 comes from the OEM; NVIDIA's own page carries no warranty or support terms. Published terms are thin:

• HP Japan: three years onsite with parts.
• Dell: a default Basic Hardware Service with onsite support after remote diagnosis, per its Canada configurator.
• ASUS Singapore: onsite service after remote diagnosis, without stating years.
• Supermicro, MSI and GIGABYTE: no Station-specific terms were found.

Before sign-off, register the unit to the end-user entity named on the paperwork. Get the Singapore terms and response times in writing. Our HP vs MSI warranty and support comparison shows how much these differ.

Plan for what a repair looks like. Dell's service manual lists the fan filters, side covers and graphics card as customer-replaceable. Drives and the PSU are field-replaceable, and the chassis assembly is replaced as a whole. A fault in the GB300 itself is therefore likely to mean a whole-unit swap measured in days. Our post on the DGX Station as a single point of failure covers fallback options.

What goes in the handover pack?

A useful pack includes the as-built configuration (drives, RAID layout, display card, firmware and OS image versions) and a network diagram with the 10 GbE, BMC and any 400 GbE connections. Add the planned MIG layout, the UPS and power details, OEM support contacts and the return process. Include a note of software entitlements. NVIDIA AI Enterprise is a separate purchase. Dell includes Ubuntu Pro. HP's specification notes that no recovery media ships with the ZGX Fury, so record how a re-image is done. Before sign-off, the unit should already appear in your monitoring (see our DCGM, Prometheus and Grafana guide) and have completed its first backup. Finish with a short onboarding session for the first users and a signed acceptance certificate. Who looks after it from then on is covered in who manages a DGX Station.

Buying in Singapore

SourceIT lists DGX Station GB300 systems from ASUS, Dell, HP, MSI, GIGABYTE and Supermicro, each supplied through its OEM's channel and all price on request. Examples include the GIGABYTE W775-V10-L01 and the ASUS ExpertCenter Pro ET900N G3, and the DGX Station GB300 buyer's guide compares them. We validate the configuration, coordinate delivery with the OEM and can agree the acceptance tests and handover pack with you before the order is placed. For clean shutdowns, the APC Network Management Card 3 is S$596.25 (listed 9 Nov 2026). Size any UPS in watts against the Station's 1,600 W supply. Prices change; ask for a current quotation. Use Request for Quotation on the product page, call (65) 6978 3502, email Sales@sourceit.com.sg or use our contact page.

Frequently asked questions

Does NVIDIA provide an acceptance test for the DGX Station GB300? No. NVIDIA publishes no Station burn-in or acceptance procedure, so agree the tests with your supplier in the purchase order.

Who provides the warranty? The OEM that built the unit, such as ASUS, Dell, HP, MSI, GIGABYTE or Supermicro. Terms vary by OEM and country, so get the Singapore terms in writing.

What happens if the GB300 itself fails? Dell's service manual lists the chassis assembly as the replacement unit, which suggests a whole-unit swap. Plan a fallback for the repair period.

Should the supplier keep admin access after handover? Only if a support agreement requires it. Otherwise take over the BMC and OS credentials, rotate any password the engineer used and store them sealed.

Is NVIDIA AI Enterprise included with the Station? No. NVIDIA treats it as a separate software licence. Record which entitlements you bought in the handover pack.