Buy the NVIDIA RTX A400 4 GB: four displays at 50 W, price by written quote

RTX A400 price by written quote within one business day

The NVIDIA RTX A400 is a single-slot workstation graphics card with 4 GB of GDDR6 memory, 768 CUDA cores, four Mini DisplayPort outputs and 50 W of total board power. We supply it to companies across the EU. No price is printed here: tell us how many cards you need, and within one business day you receive a written quote with the price, the lead time and the warranty terms.

The buyers are IT departments, system builders and integrators who equip desks with several monitors, 2D CAD seats and compact office workstations, or who renew a fleet of older entry cards. Every figure of the card, with the NVIDIA document it comes from, is in our RTX A400 specs.

What the RTX A400 is ordered for

Desks with up to four displays. The four outputs are Mini DisplayPort 1.4a, and NVIDIA's datasheet of May 2024 lists four displays at 4096 × 2160 and 120 Hz, or four at 5120 × 2880 and 60 Hz. That is the card's first job: dispatch, monitoring and finance desks, and engineering seats with a drawing on every screen. The datasheet lists NVIDIA Mosaic with the footnote "Windows 10 and Linux"; with it an application can span the four screens of one card. A wall of several synchronised GPUs is a different order, because the RTX A400 is not on NVIDIA's RTX PRO Sync list of 15 October 2026; our article on four monitors with the RTX A400 and the NVIDIA RTX A1000 shows where that limit lies.

2D CAD and office graphics. On the pages of the software makers, read on 10 November 2026, 4 GB reaches the lines set for drawing work. The recommended line of AutoCAD LT 2027 names a 4 GB GPU, the basic line of AutoCAD 2027 a 2 GB GPU, and Adobe gives 4 GB of GPU memory for Photoshop on 4K displays and greater. Dassault's SOLIDWORKS 2026 certification list, read on the same day, has rows with the RTX A400 in desktop workstations. Each row names one computer model, one card and one driver, so look up your own workstation model in it. Our guide to the GPU for AutoCAD, Revit and Inventor goes through Autodesk's tables.

RT and Tensor Cores in the entry class. The T400 4GB, which NVIDIA uses as the previous generation in its comparisons, has neither. The RTX A400 has 6 RT Cores and 24 Tensor Cores, and NVIDIA's product page describes its AI use as "light AI workloads". How light is a matter of memory, which the next section puts into numbers.

Where 4 GB ends

The same pages show where the card stops. The table sets 4 GB against the memory figures of each maker.

system requirements pages and readme files of the software makers for the versions named, read 10 November 2026; memory figures only. Autodesk's AutoCAD and Inventor lines also carry a bandwidth value, and Lumion and Twinmotion set a benchmark score; neither is compared here.

Twinmotion and Lumion set their minimum at 6 GB, so the RTX A400 is not the card for these two renderers. In the rows above them it is the entry: AutoCAD LT aside, it reaches the minimum or basic line and stays below what the maker recommends.

One caution on Autodesk's lines. In AutoCAD and Inventor they carry a "Bandwidth" value next to the memory figure, 106 GB/s in the recommended lines of AutoCAD 2027 and AutoCAD LT 2027, and Autodesk does not define it. NVIDIA's datasheets give the RTX A400 a memory bandwidth of 96 GB/s and the RTX A1000 192 GB/s. Neither document says that the two figures measure the same thing, so we draw no conclusion from them, and the table compares memory only.

By our sizing rule the card is not one for language models. The rule takes 0.9 times the memory the driver reports, less 3 GiB, as the budget for model weights plus KV cache, and on a 4 GB card that leaves 0.6 GiB. We scale the reported memory from the 95.6 GiB the driver shows on the 96 GB card; that is an assumption, and the driver on the delivered card decides. Our article on AI on the RTX A1000 and RTX A400 shows what the two cards can and cannot hold.

RTX A400 or RTX A1000, row by row

Both cards share one board size and one power figure, so the choice is made in the rows where they differ.

NVIDIA datasheets for the RTX A400 (3279761, May 2024) and the RTX A1000 (April 2024); NVENC sessions from NVIDIA's Video Encode and Decode GPU Support Matrix of 16 October 2026; driver branches from NVIDIA's release notes 621.14, 597.30 and 582.99, as read on 10 November 2026.

The RTX A1000 is the better order in three cases. The software names 8 GB in its recommended line, as AutoCAD 2027, Premiere and Enscape do. An 8K display is planned, which only the RTX A1000's datasheet lists. Or the seat plays several video streams at once, for which the RTX A1000 has a second decoder. For a four-display desk and a 2D CAD seat the rows that matter are identical, and the RTX A400 does the job. The step up is described on our page to buy the RTX A1000, in the RTX A1000 specs and in the comparison RTX A1000 vs RTX A400.

We supply both cards across the EU; send us the applications and the number of seats and we quote the card that fits, or both.

In place of a T400, Quadro P400 or Quadro P620

In an existing fleet the RTX A400 is a candidate for the slot of an older entry card.

NVIDIA datasheets for the T400 (November 2021), the Quadro P400 (May 2018), the Quadro P620 (February 2018) and the RTX A400 (3279761, May 2024); all four cards are single slot.

Three things change in the slot. The RTX A400 draws more, 50 W where the T400 and the Quadro P400 are listed at 30 W and the Quadro P620 at 40 W, so the power budget of the smallest desktops deserves a look. It is longer: 6.4 in against 6.137 in for the T400 and 5.7 in for the two Quadro cards. And its host interface is PCIe 4.0 x8, where the older datasheets say PCI Express 3.0 x16; name the workstation model in your request. In return a fourth connector is gained over the T400 and the Quadro P400.

For the two Quadro cards the driver is a further point. NVIDIA's release notes for driver 621.14 state that Release 580 was the last branch to support Maxwell, Pascal and Volta GPUs. The Linux driver 580.119.02 still lists the Quadro P400 and P620; whether a newer release does is shown by its table of supported products. The RTX A400 is listed in R580, R595 and R620. Release 620 is also the last branch for Windows 10, which matters to fleets that still run it. Our article on replacing Quadro P1000, T400 and T1000 cards covers the whole range of old cards.

Slot, power and software in the target system

The card needs one slot of single width with room for 2.7 in of height and 6.4 in of length, which converts to about 69 by 163 mm. A footnote in the datasheet says that it uses a full-length PCIe Gen 4 x8 interface, so tell us if the free slot is a short one. NVIDIA's datasheet gives a maximum power consumption of 50 W and names no auxiliary power connector. It names no power-supply wattage either: add 50 W to the load the system already carries and compare the sum with the rating its maker gives. The monitors connect over Mini DisplayPort, and the card has no HDMI output.

On the software side the card runs on the NVIDIA RTX Enterprise Production Branch driver. The current Windows releases 621.14 (R620, 4 November 2026), 597.30 (R595, 22 October 2026) and 582.99 (R580, 21 October 2026) list it, as do the Linux releases 620.70.06, 595.112.07 and 580.196.07. NVIDIA's release notes name Windows 11 from 22H2 to 25H2 and Windows 10 22H2, and on Linux, among others, Ubuntu 26.04, 24.04 and 22.04 LTS, RHEL and Rocky 10.x, 9.y and 8.y, SLES 16 and 15 SP7, and Debian 13 and 12.

What the quote names and what the price depends on

Our quote is a written document that reaches you within one business day. It names the price for the quantity you asked for, the lead time and the warranty terms; quantities and lead time are confirmed in writing with the quote. The card is available to order.

Three things move the price: the quantity, the delivery address and the date of the order. For that reason we print neither a list price nor an MSRP figure. A card that is to arrive inside a workstation becomes a separate line in the quote for the whole system, as our page on workstations and servers built to order describes.

A request that can be quoted without a second round names the card with its memory (RTX A400, 4 GB, to keep it apart from the RTX A4000), how many you need, the systems the cards go into, the delivery address and the company that receives the invoice.

Contract, delivery, warranty and quantities for a fleet

The contract is a single European one, and the invoice is issued in Austria. Delivery goes to any address in the EU and, through our logistics partner, to destinations outside it. The manufacturer warranty is handled through us, and a card that is dead on arrival is replaced. The first consultation is free of charge.

Larger quantities are part of what we supply, whether an IT department refits a floor of desks or a system builder fits the card into its own small-form-factor systems. Name the full number, and the quote covers it with its lead time. Our page on NVIDIA RTX cards in volume for system builders describes such orders, and the guide to the GPU for CAD workstations helps when part of the fleet needs a larger card. If the mix of seats is still open, describe the desks and the software on them, and our reply sorts them into RTX A400 and RTX A1000 seats.

FAQ