NVIDIA RTX 2000E Ada Generation specs: 16 GB GDDR6, 2,816 CUDA cores, 50 W, single slot
Eurokommerz, Vienna, since 2006: Private AI/ML · IT Managed Services · Enterprise Training · AI Hardware & Software
- The NVIDIA RTX 2000E Ada Generation has 16 GB of GDDR6 with ECC on a 128-bit interface at 224 GB/s, with 2,816 CUDA cores, 88 Tensor Cores and 22 RT Cores, according to NVIDIA's datasheet of October 2025
- The datasheet gives 50 W of total board power, a single slot at 2.7 in of height and 6.6 in of length, active cooling and a PCIe 4.0 x8 link
- Four Mini DisplayPort 1.4a outputs drive four displays at 4096 × 2160 and 120 Hz or at 5120 × 2880 and 60 Hz, or two displays at 7680 × 4320 and 60 Hz
- The dual-slot RTX 2000 Ada has the same memory and core counts; the datasheets give 8.9 TFLOPS in single precision at 50 W for the 2000E and 12.0 TFLOPS at 70 W for its sibling
- NVIDIA publishes no launch date, no compute capability and no NVENC session limit for the card, and we print no AI TOPS figure because two editions of its datasheet disagree
Supplied by Eurokommerz: NVIDIA RTX 2000E Ada Generation Check price and availability →
RTX 2000E Ada specs from NVIDIA's datasheet
The NVIDIA RTX 2000E Ada Generation puts 16 GB of GDDR6 memory with ECC and 2,816 CUDA cores into a single slot at 50 W of total board power. NVIDIA's datasheet gives a board of 2.7 by 6.6 in, a memory bandwidth of 224 GB/s and four Mini DisplayPort 1.4a outputs. The card is of the Ada Lovelace generation and connects over PCIe 4.0 x8.
Which documents describe this card
NVIDIA publishes little about this card. It has no product page and no row on NVIDIA's desktop line-up page of 8 September 2026, and we found no blog entry or press release for it. Three kinds of documents remain, and this page names the one behind every value.
- NVIDIA's datasheet 4328932, dated October 2025. It is the source of the hardware rows.
- NVIDIA's driver documents. The RTX Enterprise release notes 621.14, 597.30 and 582.99 and the Linux driver lists name the card. The MIG guide, the vGPU list, the compute capability page and the video support matrix were checked for it.
- Lenovo's product guide LP2207 for the card, updated on 6 October 2025. A system maker's document, used here for statements NVIDIA's datasheet does not make, and marked as Lenovo's each time.
A value that none of the three states is marked "not published" in the table, and no shop listing is used to fill it.
RTX 2000E Ada spec table with the document for each row
| SPECIFICATION | NVIDIA RTX 2000E ADA | DOCUMENT |
|---|---|---|
| Architecture | NVIDIA Ada Lovelace | datasheet, release notes |
| Launch date | not published | no NVIDIA announcement |
| GPU memory | 16 GB GDDR6 | datasheet |
| ECC | yes | datasheet |
| Memory interface | 128-bit | datasheet |
| Memory bandwidth | 224 GB/s | datasheet |
| CUDA cores | 2,816 | datasheet |
| Tensor Cores | 88 | datasheet |
| RT Cores | 22 | datasheet |
| Single precision | 8.9 TFLOPS | datasheet |
| Ray tracing | 20.5 TFLOPS | datasheet |
| AI TOPS | not printed here | two datasheet editions differ |
| Total board power | 50 W | datasheet |
| Auxiliary power cable | not required | Lenovo guide |
| Host interface | PCIe 4.0 x8 | datasheet |
| Height × length | 2.7 in × 6.6 in | datasheet |
| Slot width | single slot | datasheet |
| Cooling | active | datasheet |
| Display outputs | 4 × Mini DisplayPort 1.4a | datasheet |
| Content protection | HDCP 2.2 | datasheet |
| Video engines | 1 × encode, 1 × decode, AV1 | datasheet |
| NVENC session limit | not published | card not in the support matrix |
| Graphics APIs | DirectX 12, Shader Model 6.6, OpenGL 4.6, Vulkan 1.3 | datasheet |
| Compute APIs | CUDA 11.6, OpenCL 3.0, DirectCompute | datasheet |
| Mosaic | listed, Windows 10 and Linux | datasheet |
| NVLink | no | Lenovo guide |
| MIG and vGPU | not listed | MIG guide, vGPU list |
| Compute capability | not listed | CUDA page |
| Driver branches | R580, R595, R620 | release notes |
NVIDIA RTX 2000E Ada Generation datasheet 4328932 (October 2025), re-read on 10 November 2026; Lenovo Press product guide LP2207 (updated 6 October 2025); NVIDIA's MIG User Guide (4 November 2026), vGPU supported-GPU list (2 October 2026), "CUDA GPU Compute Capability" page (16 October 2026), Video Encode and Decode GPU Support Matrix (16 October 2026) and RTX Enterprise release notes 621.14 (4 November 2026).
A footnote ties both TFLOPS rows to the GPU's boost clock, so they are peak rates. The AI row is left empty on purpose. The datasheet of October 2025 and an earlier edition of the same sheet give different AI TOPS values, and in both the figure is a theoretical one with sparsity. Until the two are reconciled we print neither; the other rows above agree between the editions.
RTX 2000E Ada and RTX 2000 Ada side by side
The card without the E, the NVIDIA RTX 2000 Ada Generation, is the better-documented sibling, and the two differ in few rows.
| ROW | RTX 2000E ADA | RTX 2000 ADA |
|---|---|---|
| Memory | 16 GB GDDR6 with ECC, 224 GB/s | 16 GB GDDR6 with ECC, 224 GB/s |
| CUDA / Tensor / RT cores | 2,816 / 88 / 22 | 2,816 / 88 / 22 |
| Single precision | 8.9 TFLOPS | 12.0 TFLOPS |
| Ray tracing | 20.5 TFLOPS | 27.7 TFLOPS |
| Total board power | 50 W | 70 W |
| Slot width | single | dual |
| Height × length | 2.7 in × 6.6 in | 2.7 in × 6.6 in |
| Compute capability | not listed | 8.9 |
NVIDIA datasheets 4328932 (October 2025) and 2999417 (January 2024); NVIDIA's "CUDA GPU Compute Capability" page (16 October 2026).
The 2000E carries the same memory and the same core counts in half the slot width and with 20 W less board power. NVIDIA's own sheets put its peak single-precision rate at 74 per cent of the dual-slot card's, by our arithmetic. NVIDIA's Linux driver list gives both cards the PCI device ID 28B0 and tells them apart by the subsystem ID. The full comparison, with the older RTX A2000 and the RTX A1000, is on our page RTX 2000 Ada vs RTX 2000E Ada; the dual-slot card has its own specs and a page to buy the RTX 2000 Ada.
Single slot at half height, dimensions and fit in SFF systems
The size is given in inches: 2.7 in of height and 6.6 in of length, in a single slot. By our conversion that is roughly 69 by 168 mm; a millimetre figure and a weight are not in the datasheet. A height of 2.7 in is half-height, the format usually called low profile.
Among the 16 GB cards of this height that we supply, the RTX 2000E Ada is the only one that takes a single slot. Our comparisons of single-slot NVIDIA RTX cards and of low-profile NVIDIA RTX cards show the alternatives in one table each.
The thermal row of the datasheet reads active. The card has a fan, and descriptions that call it passively cooled do not match the datasheet. We supply the RTX 2000E Ada to companies across the EU; send us the host system and the quantity and a written quote follows within one business day.
50 W of board power and the power cable
The datasheet states a total board power of 50 W and has no row for a power connector. The statement about the cable comes from Lenovo, whose product guide for the card says it does not require an auxiliary power cable. NVIDIA's power guidelines for workstation products, a general document of July 2023 that predates the card, give 75 W as the typical rating of a PCIe slot.
NVIDIA publishes no power-supply wattage for the card, so this page prints none; 50 W is the number to enter in the power budget of the chassis. For the host link the datasheet prints PCIe 4.0 x8 and, in a footnote, a full-length PCIe Gen 4 x8 interface.
How many displays the RTX 2000E Ada drives, and at which resolution
The outputs are four Mini DisplayPort 1.4a connectors, and four displays are the most the card drives. For them the datasheet offers three combinations to choose from: four displays at 4096 × 2160 and 120 Hz, four at 5120 × 2880 and 60 Hz, or two at 7680 × 4320 and 60 Hz. An HDMI output is not present. HDCP 2.2 and DisplayPort with audio are among the listed features.
NVIDIA Mosaic is listed with the footnote "Windows 10 and Linux". The datasheet has no line on a sync board, and the card is neither on NVIDIA's RTX PRO Sync page of 15 October 2026 nor in the sync table of the 621.14 release notes. For Mosaic across two or more GPUs NVIDIA's documentation calls for that board, so the 2000E is planned as a single card with at most four displays. The feature list also names RTX Desktop Manager; the 621.14 release notes state that Release 580 was the last driver branch to support it.
One encoder and one decoder, both with AV1
The video row of the datasheet gives one encode and one decode engine with AV1 encoding and decoding. NVIDIA's Video Encode and Decode GPU Support Matrix of 16 October 2026 has no row for the 2000E, which leaves two points open: the number of concurrent NVENC sessions and the 4:2:2 formats. For the RTX 2000 Ada the matrix lists an unrestricted number of sessions, which says nothing about a card it does not name. How the limits differ between the cards we supply is the subject of our article on NVENC session limits.
Driver releases that list the RTX 2000E Ada
The card runs on the NVIDIA RTX Enterprise Production Branch driver. The release notes of all three current branches list it as NVIDIA RTX 2000E Ada Generation, and so do the Linux driver lists of the same branches.
| BRANCH | WINDOWS | LINUX | RELEASED | TYPE AND END DATE |
|---|---|---|---|---|
| R620 | 621.14 | 620.70.06 | 4 November 2026 | Long Term Support Branch, September 2029 |
| R595 | 597.30 | 595.112.07 | 22 October 2026 | Production Branch, March 2027 |
| R580 | 582.99 | 580.196.07 | 21 October 2026 | Long Term Support Branch, June 2028 |
NVIDIA RTX Enterprise release notes 621.14, 597.30 and 582.99 and NVIDIA's Data Center GPU Driver release notes for the same releases; branch types and end dates from NVIDIA's data-centre driver table of 4 November 2026.
The last column comes from NVIDIA's data-centre documentation; NVIDIA publishes no end date for the Windows RTX Enterprise branches. The dates are end dates of driver branches and say nothing about how long the card itself can be ordered. On R595 the documents differ: NVIDIA's data-centre table calls R595 a Production Branch (to March 2027); the Windows release notes of 597.30 call Release 595 a New Feature Branch.
On Windows the release notes cover Windows 11 in versions 22H2 to 25H2 and Windows 10 in version 22H2; after Release 620 no further branch will support Windows 10. The Linux lists include, among others, Ubuntu 22.04, 24.04 and 26.04 LTS, RHEL and Rocky in 8.y, 9.y and 10.x, SLES 15 SP7 and 16, and Debian 12 and 13. Our guide to the RTX Enterprise driver on Windows 10 and 11 explains the branches.
The datasheet prints "CUDA 11.6", which was the version when the sheet was written. What runs today depends on the driver: R620 goes with CUDA 13.5, R595 with CUDA 13.2 and R580 with CUDA 13.0, and an application built for CUDA 13.x needs a driver from branch 580 or newer. NVIDIA's compute capability page of 16 October 2026 lists 8.9 for the RTX 2000 Ada and does not name the 2000E, so we print no figure for it. Software that asks for a stated capability level should be confirmed on the card itself.
What is not published for the RTX 2000E Ada
- A launch date. NVIDIA has announced the card in no document we found. The documents carry their own dates: the datasheet October 2025, Lenovo's guide first published on 22 April 2025 and updated on 6 October 2025. They date the documents, and no market start can be read from them.
- An availability commitment. NVIDIA's long-life statement on its embedded page of 31 July 2026 covers the embedded modules listed there, and the 2000E is not on that page.
- Operating conditions. NVIDIA's datasheet contains no temperature, humidity or vibration figure. The conditions of use therefore come from the documentation of the system the card is built into.
- MIG and vGPU. The card is in neither NVIDIA's MIG guide nor its vGPU list. On NVLink, Lenovo's guide says no.
- Entries in software makers' lists. Dassault's SOLIDWORKS 2026 certification list and D5 Render's list of supported GPUs, both read on 10 November 2026, do not name the 2000E. That means "not listed"; neither maker says the card is unsupported.
- Measurements. We know of no review or benchmark of this card that names its system, date and conditions, so the datasheet rates are the only performance figures on this page.
No laptop version, and not the Embedded GPU
NVIDIA uses the name RTX 2000 Ada Generation for two further products, a laptop GPU with 8 GB, 3,072 CUDA cores and 35 to 140 W, and an embedded module with 8 GB that its embedded page lists at 35 to 115 W. Driver lists carry them as "Laptop GPU" and "Embedded GPU". Neither exists with an E in the name. The RTX 2000E Ada is an add-in card for a PCIe slot, and a benchmark or a listing for one of the 8 GB products says nothing about it.
Where the card sits among the half-height cards
| CARD | MEMORY | SINGLE PRECISION | POWER, SLOTS |
|---|---|---|---|
| RTX A1000 | 8 GB GDDR6, 192 GB/s | 6.7 TFLOPS | 50 W, single |
| RTX 2000E Ada | 16 GB GDDR6 with ECC, 224 GB/s | 8.9 TFLOPS | 50 W, single |
| RTX 2000 Ada | 16 GB GDDR6 with ECC, 224 GB/s | 12.0 TFLOPS | 70 W, dual |
| RTX PRO 2000 Blackwell | 16 GB GDDR7 with ECC, 288 GB/s | 17 TFLOPS | 70 W, dual |
NVIDIA datasheets: RTX A1000 (April 2024), RTX 2000E Ada Generation (4328932, October 2025), RTX 2000 Ada Generation (2999417, January 2024), RTX PRO 2000 Blackwell (5322351, June 2026).
In one slot at 50 W the step below is the NVIDIA RTX A1000 with half the memory: RTX A1000 specs, buy the RTX A1000. With two slot widths and 70 W available, the RTX 2000 Ada and the NVIDIA RTX PRO 2000 Blackwell are the 16 GB cards with the higher single-precision rates; the Blackwell card is described in its specs and on the page to buy the RTX PRO 2000 Blackwell, and our existing comparison RTX PRO 2000 vs RTX 2000 Ada covers the generation step. What the 16 GB of the 2000E hold in a compact inference system is the subject of our article on the RTX 2000E Ada for edge AI.
NVIDIA names no successor for the RTX 2000E Ada, and the R620 driver branch of 4 November 2026 lists the card. The steps of an order are on our page to buy the RTX 2000E Ada. If you are fitting a series of compact systems, write to us with the chassis and the free slot; the quote then names the 2000E or, if the slot allows more, one of the cards beside it.
FAQ
What are the specs of the NVIDIA RTX 2000E Ada?
Is the RTX 2000E Ada a single-slot, low-profile card?
Does the RTX 2000E Ada need a power cable, and which power supply?
When was the RTX 2000E Ada released?
Which driver does the RTX 2000E Ada need?
What is the difference between the RTX 2000E Ada and the RTX 2000 Ada?
Is the RTX 2000E Ada good for gaming?
We supply the NVIDIA RTX 2000E Ada Generation and the other half-height RTX cards across the EU, as cards or in systems built to order. Describe your task in a few sentences and our team will help. We reply within one business day.
Check price and availabilityWe reply within one business day