NVIDIA T400
Dem still dey call am nvidia t400, t400
NVIDIA T400 na desktop workstation GPU. NVIDIA dey list 384 CUDA cores and 2 GB of GDDR6 memory. The base T400 get 2 GB; the T400 wey dem mention separately, 4GB, get 4 GB.
Wetin you need know
NVIDIA T400 na desktop workstation GPU. NVIDIA dey list 384 CUDA cores and 2 GB of GDDR6 memory. The base T400 get 2 GB; the T400 wey dem mention separately, 4GB, get 4 GB.
Reference specifications wey dem record no be tested performance or operating-system compatibility verdict. Verify di exact model and platform wit di manufacturer.
Model identity and limitations
The base T400 get 2 GB; the T400 wey dem mention separately, 4GB, get 4 GB.
Specifications
Manufacturer specification reference
The specifications wey dem select below come from the exact desktop manufacturer datasheet. Maximum card power and peak throughput na reference ratings, no be measured whole-system draw or application performance. Superscript reference markers for the PDF na footnotes, no be extra digits for API versions.
| CUDA cores | 384 |
|---|---|
| GPU memory | 2 GB |
| Memory interface | 64 bits |
| Peak memory bandwidth | Up to 80 GB/s |
| System interface | PCI Express 3.0 x16 |
| Maximum card power | 30 W |
| Cooling | Active |
| Card dimensions | 2.713 inches high x 6.137 inches long, single slot |
| Display connectors | 3 x Mini DisplayPort 1.4 with latching mechanism |
| Peak single-precision throughput | Up to 1.09 TFLOPs; manufacturer peak based on GPU Boost clock |
Independent scores
Dis ones na third-party benchmark numbers wey dem store for inside Feex Hardware Database. Dem no be Feex measurements.
| PassMark G3D | 3,619 PassMark figure via PassMark Videocard Benchmark. No be Feex measurement. |
|---|---|
| PassMark G2D | 583 PassMark figure via PassMark Videocard Benchmark. No be Feex measurement. |
Identify the exact card and memory variant
The base T400 get 2 GB; the T400 wey dem mention separately, 4GB, get 4 GB.
Read the complete name on the card label or system inventory and match am to the manufacturer table. A similar numeric name no dey enough: desktop workstation cards, laptop GPUs and other vendors' products fit use overlapping numbers. No transfer the specifications of a 8 GB variant to a 4 GB board, or desktop board dimensions to a laptop implementation.
Memory capacity, bandwidth and practical workloads
This model get 2 GB of dedicated GDDR6 memory on a 64-bit interface. Capacity dey determine how much data fit reside for that memory; the rated bandwidth of up to 80 GB/s describe another limit. A bigger capacity alone no mean more shader resources or higher bandwidth.
For a CAD scene, image batch or compute workload, include textures, geometry, temporary buffers and application overhead when you dey estimate memory demand. A program wey pass local memory fit slow down or fail based on im own memory-management behavior. Compare the actual project and software version instead of using a memory-size label as a performance score.
Power, cooling and physical fit
NVIDIA dey list maximum card power of 30 W, active cooling and a single-slot board wey measure 2.713 inches high by 6.137 inches long. Maximum card power no be idle draw, a whole-PC measurement or a power-supply recommendation.
Check the exact workstation's supported graphics options, available slot and bracket, clearance around adjacent components and required airflow. A low-profile board still need the appropriate mounting bracket for im enclosure. A mechanically suitable PCIe slot no dey establish firmware, power or cooling compatibility by itself. Follow the system manufacturer's installation procedure.
Display connections and adapters
The reference card provide 3 Mini DisplayPort 1.4 connectors. Mini DisplayPort na connector type; e no mean say a full-size DisplayPort cable fit fit directly. The datasheet describe adapter support for VGA, DVI and HDMI, but the correct adapter depend on the required signal, resolution and refresh rate.
The manufacturer dey list 3 simultaneous 3840 x 2160 displays at 120 Hz or 3 simultaneous 5120 x 2880 displays at 60 Hz. These na stated configurations, no be promise for every combination of adapters, cables, color settings and monitors. Check the intended topology and each component's supported link mode.
Peak specifications and application support
The datasheet's up-to-1.09-TFLOPs single-precision figure na theoretical peak based on GPU Boost clock. E no be measured frame rate, rendering score or sustained application result. Workloads also depend on memory access, precision, CPU performance, drivers and application-specific implementation.
The Turing architecture name no establish the presence of every feature of a higher-tier RTX product. Verify the application's exact graphics/compute API, hardware feature level, CUDA runtime and driver requirements. A manufacturer certification claim for a product family no be proof say every version of a specific application and operating system dey certified for this card.
If the card dey detected but the application or display fail
Confirm the exact card identity and memory size, then install the system or GPU vendor's supported driver for the OS wey you dey run. Check say the monitor connect to the intended graphics device and say the selected input, cable and adapter support the requested mode. Test one display and a documented mode before you add a complex multi-monitor arrangement.
For application failures, preserve the error message and a small reproducible workload. Check the application's supported GPU/driver matrix and actual rendering or compute backend. Reduce project memory demand as a diagnostic when the application report an allocation failure. Avoid firmware changes, clock overrides or random driver packages as an initial response; use supported recovery procedures if a driver update need to be reversed.
Manufacturer-checked specifications
NVIDIA T400
These fields dem check against NVIDIA's model-specific reference on 2026-10-31. Power ratings na manufacturer-defined metrics, no be measured system consumption. Other imported fields no don all been independently reverified. The base T400 get 2 GB; the T400 wey dem mention separately, 4GB, get 4 GB.
| Form factor | Desktop workstation GPU |
|---|---|
| GPU architecture | Turing |
| CUDA cores | 384 |
| GPU memory | 2 GB |
| Memory type | GDDR6 |
| Memory interface | 64 bits |
| Peak memory bandwidth | Up to 80 GB/s |
| PCIe generation | PCI Express 3.0 x16 |
Sources and verification
Imported reference. Source-specific checks dey beside di supported fields and citations. How we dey do research.
- Feex Hardware Database · Specification source wey dem use for dis sheet. CC BY 4.0.
- PassMark Videocard Benchmark · Independent benchmark. No be Feex measurement.
- NVIDIA: NVIDIA T400 specifications · Supports the fields for the manufacturer-reference section; check 2026-10-31. Source evidence SHA-256 1852e5f0bf06. This check no verify every imported field.
Dis article make sense to you?
Yes or no dey enough. Choose reason if you want, or add short note.
Thank you. Dat one dey help us make dis page better.