Replacing Quadro P1000, P620, T400 and T1000 cards with the NVIDIA RTX A400 or RTX A1000

Which card takes the slot of a Quadro P1000, T400 or T1000

For a workstation that holds a Quadro P400, P620 or P1000, a T400 or a T600, the card to order is the NVIDIA RTX A400 with 4 GB, or the NVIDIA RTX A1000 with 8 GB where the seat should gain memory. For a T1000 it is the RTX A1000. Both new cards are half-height single-slot cards, 2.7 in high where the old datasheets print 2.713 in, with a maximum power consumption of 50 W, a length of 6.4 in and four Mini DisplayPort outputs.

NVIDIA datasheets of the Quadro P400 (dated May 2018), P620 (February 2018), P1000 (May 2018), P2000 (June 2018) and P2200 (April 2019) and of the T400 (November 2021), T600 (May 2021) and T1000 (November 2021). The last column is our reading by slot, power, outputs and memory; NVIDIA publishes no such mapping.

The last column names a new card with at least the memory and the outputs of the old one. The full-height Quadro P2000 and P2200 have a section of their own.

What NVIDIA compares the two new cards with

NVIDIA names no successor for the T1000, the T400 or the Quadro P cards in the documents read for this page on 10 November 2026. In its performance claims for the new cards, "previous generation" stands for the T1000 8GB in the case of the RTX A1000 and for the T400 4GB in the case of the RTX A400, as the test notes show.

For the RTX A1000, NVIDIA's product page states up to 3.6 times the generative AI performance of the T1000 8GB; the test note names Stable Diffusion webUI v1.7.0 with TensorRT at 512 × 512, a Core i9-12900K and a pre-release build. For the RTX A400 the product page states up to 3 times the performance of the previous generation, with a test note that names the T400 4GB, the same processor and a pre-release build. These are NVIDIA's figures under NVIDIA's conditions. We know of no independent test that measures one of the old cards against its new counterpart.

On 10 November 2026 the Quadro P400, P620, P1000 and P2200 stood on NVIDIA's page of previous-generation desktop GPUs, while the T400, T600 and T1000 were still in the compare section of the current desktop line-up.

Old and new cards in NVIDIA's figures

NVIDIA datasheets as under the first table, and of the RTX A400 (3279761, May 2024) and the RTX A1000 (April 2024). "None" means that the datasheet mentions no RT or Tensor Cores.

CUDA-core counts compare only inside one generation, and this table spans more than one. The T1000 has more CUDA cores than the RTX A400, 896 against 768, and NVIDIA's single-precision figure is still higher for the RTX A400.

Two rows deserve a second look. The RTX A400 has 4 GB at 96 GB/s, twice the memory of a Quadro P400, P620 or 2 GB T400 and more bandwidth than any of them. Against a T600 or a T1000 it brings RT and Tensor Cores and a higher single-precision figure, but no more memory, and its published bandwidth is below their "up to 160 GB/s". The RTX A1000 has the 8 GB and the 50 W of a T1000 8GB with 192 GB/s and 6.7 TFLOPS, which is why T1000 and T600 seats that do 3D work belong to it.

Card length, slot and power in the existing chassis

NVIDIA datasheets as above; millimetres are our conversion. All cards are single-slot cards with active cooling.

The half-height cards have practically one height, 2.713 in on the old datasheets and 2.7 in on the new ones, and each takes one slot. NVIDIA's datasheets use the words low profile for the Quadro P400, P620 and P1000 and for the T1000, and its RTX A1000 datasheet speaks of a "low-profile GPU". The length is what changes. By our conversion the two new cards are about 18 mm longer than the low-profile Quadro P cards and about 7 mm longer than the T cards, so in a small-form-factor chassis the free length behind the slot has to be measured.

On power, NVIDIA's datasheet gives a maximum power consumption of 50 W for each of the two new cards and names no auxiliary power connector. In a seat with a 30 W Quadro P400 or T400 the card's maximum rises by 20 W, with a 40 W Quadro P620 or T600 by 10 W, with a Quadro P1000 by 3 W, and with a T1000 it stays the same. NVIDIA's power guidelines for workstation products of July 2023, written before the two cards and without rows for them, put the typical rating of a PCIe slot at 75 W. NVIDIA publishes no power-supply wattage for these cards; what counts is the power the workstation maker allows for the slot. The larger cards are covered in our article on RTX power connectors.

The host interface changes on paper. The datasheets of the Quadro P and T cards give PCI Express 3.0 x16; those of the RTX A400 and RTX A1000 give PCIe 4.0 x8, with a footnote that the card uses a full-length PCIe Gen 4 x8 interface.

We supply the RTX A400 and the RTX A1000 across the EU; send us the workstation models, the cards fitted today and the quantities and we reply with a written quote within one business day.

Mini DisplayPort stays, the 8K line at 60 Hz does not

The connector type stays the same for every half-height card on this page. The Quadro P620 and P1000, the T600 and the T1000 have four Mini DisplayPort 1.4 outputs, and the RTX A400 and RTX A1000 have four Mini DisplayPort 1.4a outputs. The Quadro P400 and the T400 have three connectors; NVIDIA's line card of March 2023 notes that the T400 4GB reaches four displays only with multi-stream transport. The RTX A400 has four connectors, which NVIDIA's blog of 16 April 2024 calls a first for its series.

The resolution lines improve for most seats. The Quadro P620 and P1000 datasheets list four displays at 4096 × 2160 and 60 Hz. The RTX A400 and RTX A1000 list four at 4096 × 2160 and 120 Hz or four at 5120 × 2880 and 60 Hz.

One line goes the other way. The T600 and T1000 datasheets print two displays at 7680 × 4320 and 60 Hz. The RTX A1000 lists two at 7680 × 4320 and 30 Hz, and the RTX A400 has no 8K line. A seat that drives an 8K panel at 60 Hz from a T1000 should move to one of the 16 GB cards named further down, whose datasheets list 8K at 60 Hz or more.

NVIDIA Mosaic is listed in the datasheets of all the old cards and of both new ones. Multi-display desks are the subject of our article on four monitors with the RTX A400 and RTX A1000.

What RT Cores and Tensor Cores change for the software

NVIDIA's datasheets for the Quadro P and T cards mention neither RT Cores nor Tensor Cores. The RTX A400 has 6 RT Cores of the second generation and 24 Tensor Cores of the third, the RTX A1000 has 18 and 72. On the software makers' pages read on 10 November 2026 this shows in four places.

• SOLIDWORKS Visualize 2026. The Fast render mode asks for an NVIDIA card of at least the Turing microarchitecture with RT Cores; the Accurate mode runs from the Maxwell generation on. The minimum is 4 GB of video memory.
• Enscape. Its requirements page of 19 March 2026 names the T1000 among the cards on which hardware ray tracing is not supported.
• D5 Render 3.1. The supported list names the RTX A400 and the RTX A1000.
• Blender 5.0 and later. Cycles with OptiX uses RTX hardware ray tracing and needs a driver of at least version 575.

For video, NVIDIA's blog names seventh-generation NVENC and fifth-generation NVDEC engines for the two new cards, where its support matrix of 16 October 2026 lists sixth-generation NVENC for the Quadro P and T cards. The session limit does not move: at most 8 concurrent NVENC sessions for the half-height old cards and for both new ones, and no AV1 encoding on any of them. The datasheets of the new cards list AV1 decoding.

Driver branches during the refresh, and Windows 10

A refresh rarely happens in one week, so old and new cards run side by side for a while. Three statements in NVIDIA's documents matter for that period.

Release 580 and the Quadro P cards. The release notes of driver 621.14 state that Release 580 was the last branch to support Maxwell, Pascal and Volta GPUs. The product list of the Linux driver 580.119.02 still names all five Quadro P cards of this page. The RTX A400 and RTX A1000 are listed in the newest releases of all three current branches: 582.99 for Windows in R580 (21 October 2026), 597.30 in R595 (22 October 2026) and 621.14 in R620 (4 November 2026). In R580, NVIDIA's lists therefore show both the Quadro P cards and the two new ones; before you plan a Quadro P card on a newer branch, check the supported-products table of that release. NVIDIA's data-centre table of 4 November 2026 lists R580 as a Long Term Support Branch to June 2028 and R620 as one to September 2029; for the Windows RTX Enterprise branches NVIDIA publishes no end date. The T cards go further than the Quadro P cards: NVIDIA's data-centre driver notes for R580, R595 and R620 list the T1000 8GB, T600, T400 and T400 4GB.

Release 620 is the last branch for Windows 10. The same release notes state that Release 620 will be the last branch to support Windows 10, and that it was tested with Windows 10 from version 22H2 and with Windows 11. NVIDIA's support article 5707, updated on 27 August 2025, says that after October 2025 Windows 10 driver support moves to quarterly security updates with critical bug fixes.

Release 580 was the last branch for RTX Desktop Manager. That statement is in the release notes of 621.14 as well, while the datasheets of the RTX A400 and RTX A1000 still name the tool.

Our guide to the RTX Enterprise driver on Windows 10 and 11 gives the branch dates. For certified CAD software the driver comes from the maker's list. Dassault's SOLIDWORKS 2026 certification list, read 10 November 2026, has rows with desktop workstations for the RTX A400 and the RTX A1000, each row with its own driver; our guide to the GPU for CAD workstations explains such lists.

Quadro P2000 and P2200 need a different card format

The Quadro P2000 and P2200 are full-height cards of 4.4 × 7.9 in with 5 GB, 75 W, four full-size DisplayPort 1.4 outputs and an unrestricted number of NVENC sessions in NVIDIA's matrix. No card on this page matches all of that.

• RTX A1000. 8 GB at 50 W in a single slot, with Mini DisplayPort outputs and the limit of 8 sessions.
• NVIDIA RTX 2000 Ada Generation and NVIDIA RTX PRO 2000 Blackwell. 16 GB with ECC at 70 W, 2.7 in high in two slots, with unrestricted sessions: RTX 2000 Ada specs, buy the RTX 2000 Ada, RTX PRO 2000 Blackwell specs, buy the RTX PRO 2000 Blackwell.
• NVIDIA RTX 2000E Ada Generation. 16 GB at 50 W where only one slot is free: RTX 2000E Ada specs, buy the RTX 2000E Ada.

All of these are 2.7 in high where the old card is 4.4 in, and all have Mini DisplayPort connectors where it has full-size ones, so name the slot and the monitor cables in the enquiry. Our comparison RTX A1000 vs RTX 2000 Ada vs RTX PRO 2000 sets the 8 GB and 16 GB cards side by side.

What to send us for a fleet quote

A list with four columns is enough: the workstation model, the card fitted today, the number of machines, and the displays per seat with their connectors. Add the operating system and, for certified CAD software, the release. We answer with the RTX A400 or the RTX A1000 per seat; where a seat is undecided, the comparison RTX A1000 vs RTX A400 shows which software figures 4 GB misses.

The written quote reaches you within one business day; quantities and lead time are confirmed in writing with it. The order runs on one European contract with an invoice from Austria, and the manufacturer warranty is handled through us. The cards and their sheets are on the pages to buy the RTX A400 and to buy the RTX A1000, in the RTX A400 specs and the RTX A1000 specs; larger quantities for IT departments and system builders are described in NVIDIA RTX cards in volume. If the refresh is already planned, name the old cards and the workstation models, and the written quote names a card for each seat.

FAQ