Lenovo NVIDIA RTX 2000 Ada Generation 16GB — low-profile workstation GPU for compact ThinkStation builds
16GB GDDR6 with ECC70W · slot powered4x mini-DisplayPort 1.4aLow profile · dual slot
The Lenovo NVIDIA RTX 2000 Ada Generation (Lenovo option 4X61P69962) is a low-profile, dual-slot professional graphics card with 16GB of ECC GDDR6 memory, built for CAD, BIM, design visualisation and light AI work in small form factor workstations. It draws a maximum of 70W, so it runs entirely from the PCIe slot with no auxiliary power cable.
It is the card to choose when the chassis is small but the software list is serious: certified professional drivers, 4th-generation Tensor cores, 3rd-generation RT cores and AV1 encode/decode in a card that fits where a full-height GPU will not. Lenovo lists the RTX 2000 Ada as a configurable graphics option for ThinkStation P3 Ultra SFF, P3 Ultra SFF Gen 2, P3 Tower Gen 2 and P2 Tower Gen 2.
Before you order: This is a dual-slot card. A low-profile chassis also needs two free low-profile slot openings next to each other, and the PCIe slot must be at least x8 electrically (x16 physical slots are fine). If your workstation is not on the list above, send us the model and we will check slot, airflow and BIOS support before you order.
What's in the Box
- NVIDIA RTX 2000 Ada Generation 16GB graphics card, supplied as Lenovo option 4X61P69962
- No power cable is needed — the card is powered from the PCIe slot (70W max)
- Bracket configuration (low-profile vs full-height) is not published by Lenovo for this option — ask us to confirm before ordering if your chassis needs a specific bracket
- Display outputs are mini-DisplayPort: adapters or mini-DP cables for your monitors are not listed as included
What It Means in Practice
- Four 4K monitors at 120Hz, or two 8K displays at 60Hz, from a card no taller than a low-profile bracket
- 16GB with ECC means large assemblies, BIM models and texture-heavy scenes stay in GPU memory and errors are corrected rather than crashing a long render
- 70W board power: no PSU upgrade and no extra cable in small chassis where the power supply is often 300W or less
- AV1 encode (8th-gen NVENC) is useful for screen recording, streaming design reviews and video exports with smaller files
What Will Not Work
- It will not fit single-slot-only chassis — it occupies two slot openings
- It has no HDMI or full-size DisplayPort; monitors need mini-DP cables or mini-DP adapters
- It is not a gaming card: GeForce drivers and game-ready features are not the target, and gaming performance is well below a similarly priced GeForce card
- It does not replace a high-end workstation GPU for large-scale AI training — 16GB on a 128-bit bus is a design and inference card
- Not the same card as the RTX 2000E Ada (single-slot server version) or the older RTX A2000 6GB/12GB — check the part number when replacing a card
Key Features
Ada Lovelace architecture
2,816 CUDA cores, 88 4th-generation Tensor cores and 22 3rd-generation RT cores deliver 12.0 TFLOPS FP32 and hardware ray tracing in professional viewports.
16GB GDDR6 with ECC
Error-correcting memory on a 128-bit bus at 224 GB/s — the reliability feature that separates RTX professional cards from consumer GeForce models.
Compact and efficient
Low-profile, dual-slot, 70W maximum and actively cooled. No auxiliary power connector, so it suits SFF towers and ultra-small workstations.
Professional software stack
Certified NVIDIA RTX Enterprise drivers, DirectX 12, OpenGL 4.6, Vulkan 1.3, CUDA and OpenCL 3.0, plus VR Ready support.
Technical Specifications
| Identification |
| Part Number |
4X61P69962 |
| Brand |
Lenovo (NVIDIA RTX 2000 Ada Generation) |
| Product type |
Professional graphics card (option) |
| GPU |
| Architecture |
NVIDIA Ada Lovelace |
| CUDA cores |
2,816 |
| Tensor cores |
88 (4th generation) |
| RT cores |
22 (3rd generation) |
| FP32 performance |
12.0 TFLOPS |
| RT core performance |
27.7 TFLOPS |
| Tensor performance |
191.9 TFLOPS (FP8, with sparsity) |
| Memory |
| Capacity |
16 GB GDDR6 with ECC |
| Memory interface |
128-bit |
| Bandwidth |
224 GB/s |
| Connectivity |
| Bus interface |
PCIe 4.0 x8 |
| Display outputs |
4x mini-DisplayPort 1.4a |
| Max displays |
4x 4096x2160 @120Hz, 4x 5120x2880 @60Hz or 2x 7680x4320 @60Hz |
| Video engines |
1x NVENC (8th gen, AV1 encode), 1x NVDEC (AV1 decode) |
| Physical & Power |
| Form factor |
Low profile, dual slot |
| Dimensions |
2.7 in H x 6.6 in L (approx. 69 x 168 mm) |
| Cooling |
Active |
| Max power consumption |
70 W |
| Auxiliary power |
None (slot powered) |
| Software |
| APIs |
DirectX 12, OpenGL 4.6, Vulkan 1.3, CUDA, OpenCL 3.0 |
| Lenovo platforms (PSREF) |
ThinkStation P3 Ultra SFF, P3 Ultra SFF Gen 2, P3 Tower Gen 2, P2 Tower Gen 2 |
| Warranty |
| Warranty |
1-Year Warranty (Singular) |
| Manufacturer Term |
Not published by Lenovo for this option — claims handled through Singular |
Frequently Asked Questions
Does it need a PCIe power cable from the PSU?
No. Maximum board power is 70W and the card has no auxiliary power connector, so it is powered from the PCIe slot.
Will it work in a non-Lenovo PC?
It is a standard PCIe 4.0 card, so in principle yes, provided the chassis has two adjacent slot openings of the right bracket height and adequate airflow. Lenovo positions this option for its ThinkStation systems; send us your PC model and we will check.
Can I connect HDMI monitors?
Yes, through mini-DisplayPort to HDMI adapters or cables. For 4K at high refresh rates use active adapters rated for the resolution you need.
Is this the same as the RTX A2000?
No. The RTX A2000 is the previous Ampere generation with 6GB or 12GB. The RTX 2000 Ada has 16GB, newer Tensor and RT cores and AV1 encoding.
Singular Expert Insight · Buyer's Guide
The RTX 2000 Ada is the sweet spot for small workstations: enough memory for real CAD and BIM datasets, ECC for long renders and a 70W envelope that avoids PSU and cooling upgrades. If you have a full-size tower and a stronger power supply, a larger RTX card gives more headroom — browse our professional
graphics cards to compare.
✓ Buy it if you run SolidWorks, Revit, AutoCAD, Adobe or light AI inference on a ThinkStation P3 Ultra / P3 / P2 class system and need four displays from a slot-powered card.
✕ Skip it if you need gaming performance, HDMI directly on the card, or more than 16GB of GPU memory for large AI models.
→ Also consider a higher RTX Ada / RTX PRO card for full-height towers, or the previous RTX A2000 only if you must match existing hardware.
Not sure it fits your setup? Send us the model of your rack, switch or device and we will check compatibility before you order.