AMD Radeon Instinct MI25 vs NVIDIA GRID A100A
We compared two Professional market GPUs: 16GB VRAM Radeon Instinct MI25 and 48GB VRAM GRID A100A to see which GPU has better performance in key specifications, benchmark tests, power consumption, etc.
Main Differences
AMD Radeon Instinct MI25 's Advantages
Boost Clock has increased by 49% (1500MHz vs 1005MHz)
Lower TDP (300W vs 400W)
NVIDIA GRID A100A 's Advantages
Released 2 years and 11 months late
More VRAM (48GB vs 16GB)
Larger VRAM bandwidth (1866GB/s vs 436.2GB/s)
2816 additional rendering cores
Score
Benchmark
FP32 (float)
Radeon Instinct MI25
12.29 TFLOPS
GRID A100A
+13%
13.89 TFLOPS
Graphics Card
Jun 2017
Release Date
May 2020
Radeon Instinct
Generation
GRID
Professional
Type
Professional
PCIe 3.0 x16
Bus Interface
PCIe 4.0 x16
Clock Speeds
1400 MHz
Base Clock
900 MHz
1500 MHz
Boost Clock
1005 MHz
852 MHz
Memory Clock
1215 MHz
Memory
16GB
Memory Size
48GB
HBM2
Memory Type
HBM2e
2048bit
Memory Bus
6144bit
436.2GB/s
Bandwidth
1866GB/s
Render Config
64
Compute Units
-
-
SM Count
108
4096
Shading Units
6912
256
TMUs
432
64
ROPs
192
-
Tensor Cores
432
-
-
-
16 KB (per CU)
L1 Cache
192 KB (per SM)
4 MB
L2 Cache
48 MB
-
-
-
Theoretical Performance
96.00 GPixel/s
Pixel Rate
193.0 GPixel/s
384.0 GTexel/s
Texture Rate
434.2 GTexel/s
24.58 TFLOPS
FP16 (half)
55.57 TFLOPS
12.29 TFLOPS
FP32 (float)
13.89 TFLOPS
768.0 GFLOPS
FP64 (double)
6.947 TFLOPS
Board Design
300W
TDP
400W
700 W
Suggested PSU
800 W
No outputs
Outputs
No outputs
2x 8-pin
Power Connectors
None
Graphics Processor
Vega 10
GPU Name
GA100
Vega 10 XT GL
(215-0894124)
GPU Variant
-
GCN 5.0
Architecture
Ampere
GlobalFoundries
Foundry
TSMC
14 nm
Process Size
7 nm
12.5 billion
Transistors
54.2 billion
495 mm²
Die Size
826 mm²
Graphics Features
12 (12_1)
DirectX
N/A
4.6
OpenGL
N/A
2.1
OpenCL
3.0
1.2
Vulkan
N/A
-
CUDA
8.0
6.4
Shader Model
N/A
