Radeon Instinct MI300X vs GRID M40
We compared two Professional market GPUs: 192GB VRAM Radeon Instinct MI300X and 8GB VRAM GRID M40 to see which GPU has better performance in key specifications, benchmark tests, power consumption, etc.
Main Differences
Radeon Instinct MI300X 's Advantages
Released 7 years and 7 months late
Boost Clock2100MHz
More VRAM (192GB vs 8GB)
Larger VRAM bandwidth (5171GB/s vs 83.20GB/s)
19072 additional rendering cores
GRID M40 's Advantages
Lower TDP (50W vs 750W)
Score
Benchmark
FP32 (float)
Radeon Instinct MI300X
+10205%
81720
GRID M40
793
Graphics Card
Dec 2023
Release Date
May 2016
Radeon Instinct
Generation
GRID
Professional
Type
Professional
PCIe 5.0 x16
Bus Interface
PCIe 3.0 x16
Clock Speeds
1000 MHz
Base Clock
-
2100 MHz
Boost Clock
-
2525 MHz
Memory Clock
1300 MHz
Memory
192GB
Memory Size
8GB
HBM3
Memory Type
GDDR5
8192bit
Memory Bus
128bit
5171GB/s
Bandwidth
83.20GB/s
Render Config
304
Compute Units
-
19456
Shading Units
384
1216
TMUs
32
0
ROPs
16
1216
Tensor Cores
-
16 KB (per CU)
L1 Cache
64 KB (per SMM)
16 MB
L2 Cache
2 MB
256 MB
L3 Cache
-
Theoretical Performance
0 MPixel/s
Pixel Rate
16.53 GPixel/s
2554 GTexel/s
Texture Rate
33.06 GTexel/s
653.7 TFLOPS
FP16 (half)
-
81.72 TFLOPS
FP32 (float)
793.3 GFLOPS
81.72 TFLOPS
FP64 (double)
24.79 GFLOPS
Board Design
750W
TDP
50W
1150 W
Suggested PSU
250 W
No outputs
Outputs
No outputs
None
Power Connectors
-
Graphics Processor
Aqua Vanjaram
GPU Name
GM107
CDNA 3.0
Architecture
Maxwell
TSMC
Foundry
TSMC
5 nm
Process Size
28 nm
153 billion
Transistors
1.87 billion
1017 mm²
Die Size
148 mm²
Graphics Features
N/A
DirectX
12 (11_0)
N/A
OpenGL
4.6
3.0
OpenCL
3.0
N/A
Vulkan
1.3
-
CUDA
5.0
N/A
Shader Model
5.1
