Buying RTX PRO 6000 in the EU: which edition, what comes with it and what to check first

Which RTX PRO 6000 to buy: the edition follows the machine

To buy an RTX PRO 6000 for a company, decide the edition by the machine the card goes into. The Workstation Edition (600 W, double flow-through cooler) is for a tower whose maker lists it, the Max-Q (300 W, active cooler) is for workstations with several cards, and the passive Server Edition (300 to 600 W) is for rack servers whose maker lists it for that model and slot. All three carry 96 GB of GDDR7 with ECC, so the models that fit stay the same. The edition decides power, cooling, cards per machine and vGPU, and the Server Edition’s memory bandwidth is lower, 1,597 GB/s against 1,792 GB/s.

Once the edition is clear, check the power connector and power supply, the airflow, room for dual-slot cards, the driver branch, the vBIOS if MIG is planned, and whether the workstation or server maker lists the card. The specifications side by side are in our comparison of the Workstation, Max-Q and Server editions. This guide covers what to settle before the order and what comes with cards bought for a company.

RTX PRO 6000 editions by machine, and the checks for each

NVIDIA product pages for the Workstation Edition and the Max-Q, Max-Q datasheet 5349650 (June 2026) and Server Edition product brief SP-12355-001_v02 (June 2025), all read on 10 October 2026.

NVIDIA’s product page gives the Workstation Edition 600 W, a “Double Flow Through” cooler and a size of 5.4 by 12.0 inches (about 137 by 305 mm), dual slot. A flow-through cooler blows its heat into the case, and the tower’s own fans have to move it out. Workstation makers list one or two of these cards per tower, as our article on four RTX PRO 6000 Max-Q in one workstation shows with their specifications.

The Max-Q keeps the 96 GB and the 1,792 GB/s at 300 W. NVIDIA states that it “enables up to four GPUs in a single system”, and its datasheet gives a 4.4 by 10.5 inch card, dual slot and full height, with “1x PCIe CEM5 16-pin” for power. Four of them need four dual-slot positions, each wired at x16 for full bandwidth, and a power supply that carries 1,200 W for the cards alone.

The Server Edition has no fan. NVIDIA’s product brief lists a full-height, full-length, dual-slot card with a passive thermal solution, an ambient operating temperature of 10 °C to 45 °C and a minimum airflow that must be present “whenever input power is provided to the card”. That airflow comes from a server whose maker has qualified the card for the model, slot and fan set. The models that list it as of October 2026 are in our article on RTX PRO 6000 Server Edition compatible servers.

We supply all three editions and check the platform, power and cooling of the machine before we quote. Tell us the workstation or server model and how many cards it should hold, and we say in advance if the edition will not work in it.

Power connector, power supply and airflow

The Server Edition and the Max-Q each take their power through one 16-pin connector. The Server Edition’s brief names “One PCIe 16-pin auxiliary power connector (12VHPWR auxiliary power connector)”, and the Max-Q datasheet names the CEM5 16-pin connector. NVIDIA’s product page for the Workstation Edition names no connector, so confirm with the tower maker which cable it fits for a 600 W card.

The cable also sets the power limit. NVIDIA’s brief for the Server Edition lists a 600 W mode, with 600 W maximum, and a 450 W mode, with 450 W maximum, each with a 300 W minimum. A card on a cable strapped for 450 W is capped at 450 W, as our edition comparison shows in nvidia-smi. Some servers use the lower mode on purpose. Lenovo offers the card in a version “Slot Capped at 450W” that allows four cards in its SR650a V4, where the 600 W version allows two.

For a server you already run, the cable is often a separate part. Lenovo’s product guide LP2263 states that the option “does not ship with auxiliary power cables” and that “Cables are server-specific due to length requirements”. Put the cable, and the fan kit the server maker names for 600 W cards, on the same order as the cards.

Size the power supply from the sum of the cards. Four Max-Q draw 1,200 W, two Workstation Editions 1,200 W, and eight Server Editions 4.8 kW at 600 W or 3.6 kW at 450 W, in each case before processors, memory and fans.

Driver branch, vBIOS and MIG on new cards

NVIDIA’s brief for the Server Edition gives driver R575 or later. NVIDIA’s MIG user guide, updated on 11 September 2026, lists all three editions with up to four instances and sets R575, 575.51.03 or later, as the driver minimum for MIG.

The two workstation editions need more for MIG. The guide sets a minimum vBIOS of 98.02.55.00.00 for the Workstation Edition and 98.02.6A.00.00 for the Max-Q, and their display mode must be switched from graphics to compute with DisplayModeSelector 1.72 or later. The guide warns that compute mode “will disable physical display output”, so a workstation that uses MIG on every card needs another GPU for its monitor or is run over SSH. MIG runs on Linux distributions supported by CUDA. The Server Edition ships with its display outputs off by default and has no vBIOS minimum in the guide. For a card below the minimum, the guide says “contact your reseller or system provider for assistance”, so the seller is part of the MIG setup. Ask for the vBIOS version of the delivered cards in writing if MIG is planned. The commands are in our MIG guide for RTX PRO Blackwell cards.

What comes with an RTX PRO 6000 bought for a company

NVIDIA’s English-language EU page for RTX PRO workstation graphics cards, whose text was last updated in September 2019, gives a warranty of “Three (3) years from the date of purchase”. It applies “only to the original purchasers” who bought from “an NVIDIA-authorized third party” or nvidia.com, and not to cards acquired used. The text is governed by the laws of the buyer’s country of residence and states: “Nothing in this warranty affects your legal rights”. The German page gives the same three years and the same first-buyer rule.

Neither of these two pages lists models. NVIDIA’s US page, last updated on 29 October 2018, names the RTX PRO 6000 Blackwell Workstation Edition and the Max-Q Workstation Edition with the same three years. All three pages send the owner to the local distributor or reseller first, and NVIDIA’s warranty overview says the same “for most NVIDIA products”.

The Server Edition is not listed on any of these pages, and NVIDIA’s warranty overview publishes no terms for data-centre cards. For a Server Edition, ask the server maker or the seller for the warranty term and the claims route in writing, and have both stated in the quote. The seller’s role in a claim matters here as it does for the DGX Spark, as our guide to buying a DGX Spark in the EU explains.

NVIDIA RTX PRO workstation warranty (EU English, German and US pages) and warranty overview; MIG user guide (11 September 2026); vGPU supported GPUs (2 October 2026); AI Enterprise licensing guide (2 September 2026); Server Edition brief SP-12355-001_v02; read on 10 October 2026.

The licences depend on whether the cards run virtual machines. NVIDIA’s list of GPUs supported by vGPU, updated on 2 October 2026, names the Server Edition from release 19.0 and its liquid-cooled version from release 20.0, and neither workstation edition. For virtual desktops and workstations, NVIDIA sells its virtual GPU software per concurrent user, as an annual subscription or a perpetual licence, or per GPU as an annual subscription. Compute vGPU profiles for AI workloads are licensed through NVIDIA AI Enterprise, and MIG on bare metal needs neither licence. NVIDIA AI Enterprise “is licensed on a per-GPU basis”, and the licensing guide names included subscriptions only for the H100 PCIe and NVL, the H200 NVL and the A800 40GB Active. Which workloads need it is set out in our article on NVIDIA AI Enterprise licensing.

We put the vGPU or NVIDIA AI Enterprise licences in the same quote as the cards, with manufacturer warranty handled through us. Write to us with the number of cards and how they will run, on bare metal or in virtual machines.

Buying 4 to 8 cards: workstations or a GPU server

Four Max-Q in one workstation give 384 GB in one machine at about 1,200 W for the cards. Eight cards in that form are two towers, each with its own power, heat and noise in an office, and each serving its own team. The setup suits developers who share a machine through MIG, with up to sixteen 24 GB instances per tower, reached over SSH because cards in compute mode drive no display.

Eight Server Edition cards fit one rack server from several makers, and that route changes what the company gets. vGPU becomes available, a server has out-of-band management through its BMC, and the machine sits in a rack position with its own power feed and cooling. Our AI servers built to order are assembled and load-tested before shipment, with a test report on request. There is no NVLink on the Server Edition, whose brief lists it as “Not supported”, so a model split across cards uses PCIe on either route.

For a company that plans one platform for several departments, or 4 to 8 cards across two servers for failover, the server route fits better. For a single team without a server room, the four-card workstation is the simpler option.

Checks before you order RTX PRO 6000 cards

1. Name the machine and find it on its maker’s list, for a server the exact model on the GPU support list, for a tower the edition and card count in the specification.
2. Choose the edition from that list, not from the card’s name.
3. Confirm the 16-pin cable for each card and whether the slot runs at 600 W or 450 W.
4. Add up the power of cards, processors and fans and compare it with the power supplies and the rack or office circuit.
5. Check the airflow: the tower’s fans for 600 W cards, or the server’s fan kit and the inlet temperature at the rack position.
6. Plan the driver branch, R575 or later, and the operating system releases the server maker supports.
7. If MIG is planned on Workstation or Max-Q cards, ask for the vBIOS version of the delivered cards.
8. Settle the licences and the warranty: vGPU or AI Enterprise per GPU where needed, and the warranty term and claims route in writing.

What we supply

We supply the NVIDIA RTX PRO 6000 Blackwell in all three editions, Workstation Edition, Max-Q and Server Edition, across the EU on one EU contract and invoice, with manufacturer warranty handled through us and DOA units replaced. You can order the cards on their own for machines you already run, after we check the platform, power and cooling, or built into workstations and AI servers built to order, servers with 2 to 8 cards, where we check the rack, power and airflow before we quote. NVIDIA AI Enterprise and vGPU licences come on the same invoice as the hardware. Our RTX PRO 6000 page lists the three editions, and quantities and lead time are confirmed in writing with the quote.

FAQ