NVIDIA Tesla T4 vs NVIDIA Tesla K20s
We compared two Professional market GPUs: 16GB VRAM Tesla T4 and 5GB VRAM Tesla K20s to see which GPU has better performance in key specifications, benchmark tests, power consumption, etc.
Main Differences
NVIDIA Tesla T4 's Advantages
Released 5 years and 7 months late
Boost Clock1590MHz
More VRAM (16GB vs 5GB)
Larger VRAM bandwidth (320.0GB/s vs 208.0GB/s)
64 additional rendering cores
Lower TDP (70W vs 225W)
Score
Benchmark
FP32 (float)
Tesla T4
+183%
8.141 TFLOPS
Tesla K20s
2.87 TFLOPS
Graphics Card
Sep 2018
Release Date
Feb 2013
Tesla
Generation
Tesla
Professional
Type
Professional
PCIe 3.0 x16
Bus Interface
PCIe 2.0 x16
Clock Speeds
585 MHz
Base Clock
-
1590 MHz
Boost Clock
-
1250 MHz
Memory Clock
1300 MHz
Memory
16GB
Memory Size
5GB
GDDR6
Memory Type
GDDR5
256bit
Memory Bus
320bit
320.0GB/s
Bandwidth
208.0GB/s
Render Config
-
-
-
40
SM Count
-
2560
Shading Units
2496
160
TMUs
208
64
ROPs
40
320
Tensor Cores
-
40
RT Cores
-
64 KB (per SM)
L1 Cache
16 KB (per SMX)
4 MB
L2 Cache
1280 KB
-
-
-
Theoretical Performance
101.8 GPixel/s
Pixel Rate
29.90 GPixel/s
254.4 GTexel/s
Texture Rate
119.6 GTexel/s
65.13 TFLOPS
FP16 (half)
-
8.141 TFLOPS
FP32 (float)
2.870 TFLOPS
254.4 GFLOPS
FP64 (double)
956.8 GFLOPS
Board Design
70W
TDP
225W
250 W
Suggested PSU
550 W
No outputs
Outputs
No outputs
None
Power Connectors
-
Graphics Processor
TU104
GPU Name
GK110
TU104-895-A1
GPU Variant
-
Turing
Architecture
Kepler
TSMC
Foundry
TSMC
12 nm
Process Size
28 nm
13.6 billion
Transistors
7.08 billion
545 mm²
Die Size
561 mm²
Graphics Features
12 Ultimate (12_2)
DirectX
12 (11_0)
4.6
OpenGL
4.6
3.0
OpenCL
3.0
1.3
Vulkan
1.1
7.5
CUDA
3.5
6.6
Shader Model
5.1
