Fetch GadgetsGadget specs

eGPU dock specs: what to look for before you buy

Nine specs that decide how much graphics performance actually reaches your laptop, each one explained with a real dock as the example.

An external GPU is a compromise: desktop-class graphics delivered over a cable that was never designed for it. How much of that performance survives the trip depends on a handful of things, and the difference between a dock at $110 and one at $630 is mostly about what you have to supply yourself. These are the nine specs worth checking, with a current enclosure introduced at each one.

Host interface: the cable is the ceiling

Everything the graphics card renders has to travel back down the host cable, and that link is far slower than the PCIe slot it replaces. OCuLink carries PCIe 4.0 x4 at 64Gbps with no display tunnelling, so it is the fastest option and the one that loses the least performance, but it needs a matching port on the host and it cannot be hot-plugged safely.

Thunderbolt 5 and USB4 V2 are the flexible options, at 80Gbps and up to 120Gbps in boost mode, and both tunnel display and data alongside the PCIe traffic. Thunderbolt 4 and USB4 run at 40Gbps and cost noticeably more performance. Check what your host actually has before buying, because a dock with a Thunderbolt 5 port delivers no advantage over a cheaper Thunderbolt 4 one if the laptop only offers 40Gbps.

Enclosure or integrated GPU: what you are actually buying

Docks fall into two kinds. A bare enclosure is a box with a PCIe slot, a power supply and a host cable, and you install whatever graphics card you already own or plan to buy. An integrated dock has the GPU soldered in, permanently, and is usually a mobile-class chip in a much smaller shell.

Bare enclosures cost less to buy and can be upgraded, and they reuse a card you may already have. Integrated docks are smaller and quieter and need no installation, but the GPU cannot be changed, so the dock is only worth buying if that specific chip is what you want. If you already own a desktop card, a bare enclosure is nearly always the better value.

GPU slot: x16 mechanical, x4 electrical

Nearly every dock mounts a full-length PCIe x16 card, and nearly every one of them runs it electrically at x4. That is not a fault, it is the interface: the host cable carries four PCIe lanes, so four is all the slot can be fed. The mechanical x16 connector is there so that physically normal desktop cards fit.

The practical consequence is that a card's performance in a dock sits somewhere between its desktop and laptop results, and that a more powerful card still helps even at x4. It also means a card designed for x8 or x16 bandwidth loses proportionally more than one that was always x8. If a listing advertises an x16 slot without stating the electrical width, x4 is the safe assumption.

GPU clearance: measure the card before the dock

A dock that cannot accommodate your card is useless, and clearance is quoted as a length, a slot width and sometimes a height. Modern triple-slot cards run 300mm to 350mm long and 60mm to 70mm thick, and the largest of them exceed what many enclosures accept. The card's power connectors add to the height requirement, since a 12VHPWR plug needs bending room that a tight lid does not allow.

Check the dock's figures against the card's published dimensions, including the shroud, because the quoted card length often excludes the bracket or the power connector. An open-frame dock sidesteps the problem entirely by having no walls at all, at the cost of noise and dust. A closed, vented enclosure is quieter but constrains what fits.

Power supply: built in, supplied, or yours to buy

Graphics cards need serious power, and where that power comes from is a real difference between docks. Many include a built-in ATX supply, usually 850W with a native 12VHPWR connector, which is the simplest arrangement and the quietest, since the supply sits inside the enclosure. Some ship with an external adapter brick instead, which keeps the dock small but puts a second box on the floor.

Others include nothing and expect you to fit your own ATX or SFX unit, which is cheaper on paper and lets you reuse a spare supply, but adds a second purchase to the real cost. Check the wattage against the card's requirement plus about 150W of headroom, and check whether the supply is user-replaceable: a built-in unit that fails takes the whole dock with it.

Reverse charging: one cable for graphics and power

Power delivery back down the host cable is one of the quiet advantages of a Thunderbolt or USB4 dock, because it lets a single cable charge the laptop and feed it a graphics card at the same time. OCuLink cannot do this: it carries PCIe data only, so an OCuLink dock always means a separate charger.

Check the wattage, because it varies widely. Anything from 15W to 140W appears in this category, and a laptop with a 100W requirement will discharge under load from a 30W port. If you want dock-only desk setup, a dock that delivers at least the laptop's own charger rating is the requirement, not a nice-to-have.

Display outputs: drive the monitors from the dock

On a dock, the display outputs belong to the graphics card inside it, not to the laptop. Connecting the monitor to the dock means the frames come straight from the GPU rather than being tunnelled back to the laptop and then out again, which saves bandwidth and usually gives access to higher refresh rates and resolutions.

Check for HDMI 2.1 and DisplayPort 2.1, which between them cover high refresh 4K and 8K. Older docks with HDMI 2.0 and DisplayPort 1.4 cap out well below what current cards can render. If the dock has no video outputs of its own, the display path runs over the host cable, which works but consumes bandwidth that the graphics data also needs.

Size and portability: a dock or a second computer

Docks range from something you can throw in a bag to a case the size of a small desktop tower. An enclosure that takes a full ATX supply and a triple-slot card is around 400mm deep and weighs six kilograms, which is a permanent desk fixture. An integrated dock with a mobile GPU can be under 250mm and around 900g, which is genuinely portable.

Portability costs performance, because a mobile GPU in a small shell runs at a lower power limit than the desktop card it replaces. Decide whether the dock will move before comparing specs: if it stays on the desk, buy the bigger enclosure with better cooling and clearance, and if it travels, accept the mobile chip.

Price: the enclosure is only part of the cost

Prices run from about $110 for a bare chassis to over $2,000 for a dock with a flagship mobile GPU inside. The cheap end is more honest than it looks: a $110 dock plus a graphics card you already own costs less than an integrated unit and performs better, because a desktop card at x4 still beats a mobile chip. What the cheap end lacks is a power supply, a card and sometimes a cable.

Add up the total before comparing. A bare enclosure at $110 plus an 850W supply plus a card can easily pass the price of a complete integrated dock, and the integrated one arrives assembled, quieter and with a single warranty. Above roughly $700 you are mostly paying for the GPU itself, so compare those docks on the chip, not on the enclosure.