The Tesla T4 is a professional graphics card by NVIDIA, launched in September 2018. Built on the 12 nm process, and based on the TU104 graphics processor, in its TU104-895-A1 variant, the card supports DirectX 12.0. The TU104 graphics processor is a large chip with a die area of 545 mm² and 13,600 million transistors. Unlike the fully unlocked GeForce RTX 2080, which uses the same GPU but has all 2944 shaders enabled, NVIDIA has disabled some shading units on the Tesla T4 to reach the product's target shader count. It features 2560 shading units, 160 texture mapping units and 64 ROPs. Also included are 320 tensor cores which help improve the speed of machine learning applications. The card also has 40 raytracing acceleration cores. NVIDIA has placed 16,384 MB GDDR6 memory on the card, which are connected using a 256-bit memory interface. The GPU is operating at a frequency of 585 MHz, which can be boosted up to 1590 MHz, memory is running at 1250 MHz.
Being a single-slot card, the NVIDIA Tesla T4 does not require any additional power connector, its power draw is rated at 70 W maximum. This device has no display connectivity, as it is not designed to have monitors connected to it. Tesla T4 is connected to the rest of the system using a PCI-Express 3.0 x16 interface. The card measures 168 mm in length, and features a single-slot cooling solution.
Graphics Processor
- GPU Name
- TU104
- GPU Variant
- TU104-895-A1
- Architecture
- Turing
- Foundry
- TSMC
- Process Size
- 12 nm
- Transistors
- 13,600 million
- Die Size
- 545 mm²
Graphics Card
- Release Date
- Sep 13th, 2018
- Generation
- Tesla
(Txx)
- Production
- Active
- Bus Interface
- PCIe 3.0 x16
Clock Speeds
- GPU Clock
- 585 MHz
- Boost Clock
- 1590 MHz
- Memory Clock
- 1250 MHz
10000 MHz effective
Memory
- Memory Size
- 16 GB
- Memory Type
- GDDR6
- Memory Bus
- 256 bit
- Bandwidth
- 320.0 GB/s
Render Config
- Shading Units
- 2560
- TMUs
- 160
- ROPs
- 64
- SM Count
- 40
- Tensor Cores
- 320
- RT Cores
- 40
- L1 Cache
- 64 KB (per SM)
- L2 Cache
- 4 MB
Theoretical Performance
- Pixel Rate
- 101.8 GPixel/s
- Texture Rate
- 254.4 GTexel/s
- FP16 (half) performance
- 65.13 TFLOPS (8:1)
- FP32 (float) performance
- 8.141 TFLOPS
- FP64 (double) performance
- 254.4 GFLOPS (1:32)
Board Design
- Slot Width
- Single-slot
- Length
- 6.6 inches
168 mm
- TDP
- 70 W
- Outputs
- No outputs
- Power Connectors
- None
- Board Number
- PG183 SKU 200
Graphics Features
- DirectX
- 12.0 (12_1)
- OpenGL
- 4.6
- OpenCL
- 1.2
- Vulkan
- 1.1.103
- CUDA
- 7.5
- Shader Model
- 6.4