Tesla T4 GPU 16GB GDDR3 Accelerator Card Graphic Card Game Enterprise GPUs
Description
T 4 data center GPU is the perfect general purpose accelerator for distributed computing environments. Revolutionary multi-precision performance accelerates deep learning as well as machine learning training and reasoning, video transcoding, and virtual desktops. The T 4 supports all AI frameworks and network types, and is robust and efficient to maximize the utility of large-scale deployments.
Turing Tensor core technology has the characteristic of multi-precision computing, which realizes the breakthrough AI performance from FP32, FP16 to INT8 and Int4 precision. Compared with CPU, its training performance is as high as 9.3 times, reasoning performance is as high as 36 times.
T4 GPU
GPU architecture Nvidia Turing
nvidia Turing Tensor cores 320
NVIDIA CUDA cores 2560
single precision 8.1 Tflops
hybrid precision (FP16/FP32) 65 tflops/ int8 130 Tops /int4 260 TOPS
GPU memory 16gb gddr6 300gb/S
ECC support
interconnect bandwidth 32gb/S
SEC system interface X16 PCIe GEN3
profile size PCIe semi-high card
cooling solution, passive
Computing API Cudanvidia TensorRTTMONNX
Product Spectification
| Turing Tensor Cores |
320 |
| CUDA® Cores |
2560 |
| Single-Precision |
8.1 TFLOPS |
| Mixed-Precision (FP16/FP32) |
65 TFLOPS |
| System Interface |
x16 PCIe Gen3 |
| Interconnect Bandwidth |
32 GB/sec |
| Thermal Solution |
Passive |
Q1: Can I get some samples?
A: Yes, sample order is available for quality check and market test. But you have to pay the sample cost and express cost.
Q2: Do you receive customized order?
A: Yes, ODM & OEM are welcomed.
Q3: What's the lead time.
A: According to the order quantity, small order usually need 3-5 days, big order need negotiation.
Q4: What's your warranty terms?
A: We supply 12 months warranty time.
Q5: What's your payment terms?
A: We receive Escrow, T/T, West Union, Cash and etc
