NVIDIA L40 vs AMD Radeon Instinct MI8
We compared two Professional market GPUs: 48GB VRAM L40 and 4GB VRAM Radeon Instinct MI8 to see which GPU has better performance in key specifications, benchmark tests, power consumption, etc.
Main Differences
NVIDIA L40 's Advantages
Released 5 years and 10 months late
Boost Clock2490MHz
More VRAM (48GB vs 4GB)
Larger VRAM bandwidth (864.0GB/s vs 512.0GB/s)
14080 additional rendering cores
AMD Radeon Instinct MI8 's Advantages
Lower TDP (175W vs 300W)
Score
Benchmark
FP32 (float)
L40
+1004%
90.52 TFLOPS
Radeon Instinct MI8
8.192 TFLOPS
Graphics Card
Oct 2022
Release Date
Dec 2016
Tesla Ada
Generation
Radeon Instinct
Professional
Type
Professional
PCIe 4.0 x16
Bus Interface
PCIe 3.0 x16
Clock Speeds
735 MHz
Base Clock
-
2490 MHz
Boost Clock
-
2250 MHz
Memory Clock
500 MHz
Memory
48GB
Memory Size
4GB
GDDR6
Memory Type
HBM
384bit
Memory Bus
4096bit
864.0GB/s
Bandwidth
512.0GB/s
Render Config
-
Compute Units
64
142
SM Count
-
18176
Shading Units
4096
568
TMUs
256
192
ROPs
64
568
Tensor Cores
-
142
RT Cores
-
128 KB (per SM)
L1 Cache
16 KB (per CU)
96 MB
L2 Cache
2 MB
-
-
-
Theoretical Performance
478.1 GPixel/s
Pixel Rate
64.00 GPixel/s
1414 GTexel/s
Texture Rate
256.0 GTexel/s
90.52 TFLOPS
FP16 (half)
8.192 TFLOPS
90.52 TFLOPS
FP32 (float)
8.192 TFLOPS
1414 GFLOPS
FP64 (double)
512.0 GFLOPS
Board Design
300W
TDP
175W
700 W
Suggested PSU
450 W
4x DisplayPort 1.4a
Outputs
No outputs
1x 16-pin
Power Connectors
1x 8-pin
Graphics Processor
AD102
GPU Name
Fiji
AD102-895-A1
GPU Variant
Fiji XT CA
(215-0862120)
Ada Lovelace
Architecture
GCN 3.0
TSMC
Foundry
TSMC
4 nm
Process Size
28 nm
76.3 billion
Transistors
8.9 billion
608 mm²
Die Size
596 mm²
Graphics Features
12 Ultimate (12_2)
DirectX
12 (12_0)
4.6
OpenGL
4.6
3.0
OpenCL
2.1
1.3
Vulkan
1.2.170
8.9
CUDA
-
6.6
Shader Model
6.5
