NVIDIA's Most Powerful Workstation GPU Has a 600W Appetite. Is Your PC Ready?

Most office PCs and many gaming builds are not ready for an NVIDIA RTX PRO 6000 Blackwell Workstation Edition as they stand. The card draws up to 600 W through a single 16-pin 12V-2x6 connector. It is a 12-inch, dual-slot, extended-height board, and its double flow-through cooler pushes that heat into the case rather than out of the back. A host that suits it has a power supply with real headroom, a native 12V-2x6 cable or the correct adapter, a chassis with strong front-to-back airflow, and a wall circuit that can carry the whole system. If you plan more than one card, the 300 W Max-Q edition is the version NVIDIA designed for that.

What does each RTX PRO 6000 edition ask of the host?

Source: NVIDIA RTX PRO 6000 Workstation and Max-Q datasheets and quick start guide (Jul 2025); NVIDIA GeForce RTX 5090 product page (fetched 6 Nov 2026); Puget Systems part notes. NVIDIA has not published a recommended PSU wattage for either RTX PRO 6000 edition.

How big a power supply do you actually need?

Because NVIDIA gives no PSU figure for the RTX PRO 6000, size it the way an integrator would: add up the parts, then leave headroom. The numbers below are a rule of thumb, not a vendor specification.

• One Workstation Edition: 600 W for the GPU, about 350 W for a high-end workstation CPU such as a 350 W Threadripper PRO, and about 150 W for memory, storage, fans and pumps comes to roughly 1,100 W. With about 30% headroom, that points to a 1,500 to 1,600 W supply.
• Two Max-Q cards: 600 W of GPU plus the same platform comes to roughly 1,150 W, again pointing to around 1,500 W.
• Four Max-Q cards: 1,200 W of GPU plus platform comes to roughly 1,750 W, or about 2,300 W with headroom. That matches what OEMs ship: Dell's Precision 7960 offers a 2,200 W supply for four 300 W cards, and HP's Z8 Fury G6i reaches 2,700 W across two supplies.

Headroom matters because GPUs have short power spikes above their rated board power, and supplies run cooler and more efficiently below full load. Our post on whether a 1,000 W power supply improves performance explains why wattage alone doesn't make a PC faster. Here the point is the reverse: too little wattage makes a 600 W card unstable.

Is the 16-pin connector something to worry about?

It deserves care, not fear. The 12V-2x6 connector is rated for 600 W, so a Workstation Edition runs it close to its limit. In February 2025, overclocker der8auer measured about 22 A on one wire of an RTX 5090 Founders Edition cable, against 6 to 8 A expected per pin, after a user reported a melted connector. Falcon Northwest tested many cards and could not reproduce the problem. In September 2026, PCGuide reported one RTX PRO 6000 Workstation Edition cable that melted at the power-supply end; the cause was unclear. That is a single anecdotal report, not a pattern.

Practical steps:

• Use a PSU with a native 12V-2x6 cable where possible. Otherwise use only the NVIDIA adapter in the box, fed by separate 8-pin cables as NVIDIA's quick start guide shows. Check which adapter your card came with before you plan the cabling.
• Push the plug fully home on both ends and avoid sharp bends near the connector.
• Don't mix third-party extension cables into a 600 W path unless the PSU maker supports them.

A 600 W card turns the power supply, the cable and the wall socket into parts of the GPU.

Will your chassis keep it cool?

The Workstation Edition's flow-through cooler is efficient for a single card, but it exhausts hot air inside the case. Puget Systems calls it "NOT suitable for most multi-GPU configurations, though in some specially designed systems it may be possible to use two of them." The Max-Q's blower sends its heat out of the rear bracket, which is why it stacks. Even then, integrators work hard at four cards: Exxact reported that its four-card Max-Q workstation throttled without custom cooling, and VRLA Tech said a four-card Workstation Edition build failed its thermal validation while four Max-Q cards passed.

Before buying, check three things:

1. Length and height: the Workstation Edition is 12 inches long and taller than a standard card, so smaller towers may not close.
2. Airflow path: strong intake at the front and a clear exhaust path at the top and rear.
3. Slot spacing: leave a gap between cards where you can, and keep backplate memory in moving air.

Our Workstation vs Max-Q vs Server Edition comparison covers the cooling trade-off in detail, and Four RTX PRO 6000 GPUs, 384GB VRAM looks at what a four-card build really involves.

Can a Singapore office socket handle it?

Singapore's 230 V supply helps. Workstation power supplies deliver their full rating only at 200 to 240 V: Dell's Pro Precision 9 T6 is rated 2,400 W at 200 to 240 V but 1,500 W at 115 to 199 V. The limit is the circuit. By our own arithmetic, at 230 V and 94% efficiency, a system drawing 1,600 W takes about 7.4 A from the wall, 2,400 W takes about 11.1 A and 2,700 W about 12.5 A. That is close to a single 13 A socket's rating before you add a monitor or anything else on the same circuit.

This is our calculation, not a regulation. For any workstation that can draw 2 kW or more, ask a licensed electrician to check the circuit and consider a dedicated one. Keep space heaters, kettles and other heavy loads off it.

Does it need a UPS?

For a machine that may train or render for hours, yes. A power blip mid-job costs time, and an abrupt shutdown is hard on storage. Schneider's guidance is that neither the watt nor the VA rating of a UPS may be exceeded by the load, so size it to the system's real draw plus headroom, not the GPU alone. For a single-card workstation that can draw up to about 1,600 W, look at 2,200 to 3,000 VA online units and confirm the watt rating covers the load; multi-card systems need more. Our post on whether a S$200 UPS can protect a S$10,000 workstation covers UPS types and runtime.

What if your PC isn't ready?

You have three options. Upgrade the PSU, cabling and case of an existing high-end tower. Choose the 300 W Max-Q, which is about 12% lower in peak compute by NVIDIA's specifications for half the power. Or buy an OEM workstation validated for the card, such as HP's Z8 Fury, Dell's Precision and Pro Precision towers or Lenovo's ThinkStation PX, whose maximum configurations are published. If the card is mainly for local AI, our RTX PRO 6000 verdict will help you decide whether you need it at all.

Buying in Singapore

SourceIT lists the RTX PRO 6000 Blackwell Workstation Edition at S$29,100 here and the 300 W Max-Q Workstation Edition at S$29,100 here (both listed 6 Nov 2026; prices change, so ask for a current quotation). For power protection, the APC Smart-UPS SRT 2200VA (S$2,926.25) is here, the SRT 3000VA (S$4,165.00) here and the SRT 6000VA (S$7,490.00) here, with more in our APC Smart-UPS SRT collection. We don't list standalone PC power supplies or chassis, so tell us your current workstation model and we'll check whether it can take the card, or quote a validated multi-GPU workstation instead. 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

What size power supply does an RTX PRO 6000 Workstation Edition need? NVIDIA doesn't publish one. Adding the GPU's 600 W to the CPU and the rest of the system, then leaving about 30% headroom, points to roughly 1,500 to 1,600 W for a single-card workstation.

Does the RTX PRO 6000 use the same 16-pin connector as the RTX 5090? Yes. All RTX PRO 6000 editions use one 12V-2x6 (PCIe CEM5 16-pin) connector, rated for 600 W. Use a native cable or NVIDIA's included adapter and seat it fully.

Can I put two Workstation Editions in one PC? Only in systems designed for it. The flow-through cooler heats the case, and Puget describes two as possible only in specially designed systems. For multi-GPU builds, the Max-Q edition is the usual choice.

Can a 13 A socket in Singapore run a multi-GPU workstation? A 2.4 to 2.7 kW system draws about 11 to 12.5 A at 230 V by our calculation, close to the socket's rating. Ask a licensed electrician to check the circuit for anything over 2 kW.

Is the Max-Q much slower? By NVIDIA's specifications it has about 12% lower peak compute at half the power. Puget measured it 5 to 14% slower in content-creation apps.