NVIDIA GRID A100B vs NVIDIA Tesla K20s
We compared two Professional market GPUs: 48GB VRAM GRID A100B and 5GB VRAM Tesla K20s to see which GPU has better performance in key specifications, benchmark tests, power consumption, etc.
Main Differences
NVIDIA GRID A100B 's Advantages
Released 7 years and 3 months late
Boost Clock1005MHz
More VRAM (48GB vs 5GB)
Larger VRAM bandwidth (1866GB/s vs 208.0GB/s)
4416 additional rendering cores
NVIDIA Tesla K20s 's Advantages
Lower TDP (225W vs 400W)
Score
Benchmark
FP32 (float)
GRID A100B
+383%
13.89 TFLOPS
Tesla K20s
2.87 TFLOPS
Graphics Card
May 2020
Release Date
Feb 2013
GRID
Generation
Tesla
Professional
Type
Professional
PCIe 4.0 x16
Bus Interface
PCIe 2.0 x16
Clock Speeds
900 MHz
Base Clock
-
1005 MHz
Boost Clock
-
1215 MHz
Memory Clock
1300 MHz
Memory
48GB
Memory Size
5GB
HBM2e
Memory Type
GDDR5
6144bit
Memory Bus
320bit
1866GB/s
Bandwidth
208.0GB/s
Render Config
-
-
-
108
SM Count
-
6912
Shading Units
2496
432
TMUs
208
192
ROPs
40
432
Tensor Cores
-
-
-
-
192 KB (per SM)
L1 Cache
16 KB (per SMX)
48 MB
L2 Cache
1280 KB
-
-
-
Theoretical Performance
193.0 GPixel/s
Pixel Rate
29.90 GPixel/s
434.2 GTexel/s
Texture Rate
119.6 GTexel/s
55.57 TFLOPS
FP16 (half)
-
13.89 TFLOPS
FP32 (float)
2.870 TFLOPS
6.947 TFLOPS
FP64 (double)
956.8 GFLOPS
Board Design
400W
TDP
225W
800 W
Suggested PSU
550 W
No outputs
Outputs
No outputs
None
Power Connectors
-
Graphics Processor
GA100
GPU Name
GK110
-
-
-
Ampere
Architecture
Kepler
TSMC
Foundry
TSMC
7 nm
Process Size
28 nm
54.2 billion
Transistors
7.08 billion
826 mm²
Die Size
561 mm²
Graphics Features
N/A
DirectX
12 (11_0)
N/A
OpenGL
4.6
3.0
OpenCL
3.0
N/A
Vulkan
1.1
8.0
CUDA
3.5
N/A
Shader Model
5.1
