L40 and H100 CNX - video card comparison
L40
H100 CNX
Architecture
Ada Lovelace
Hopper (2022)
GPU code name
AD102
GH100
Market segment
Desktop
Desktop
Release date
October 2022
March 2023
Interface
PCIe 4.0 x16
PCIe 5.0 x16
Supplementary power connectors
1x 16-pin
8-pin EPS
Boost clock speed
2490 MHz
1845 MHz
Manufacturing process technology
4 nm
4 nm
Number of transistors
76,300 million
80,000 million
Pipelines / CUDA cores
18176
14592
Power consumption (TDP)
300 Watt
350 Watt
Texture fill rate
1,414
841.3
Memory type
GDDR6
HBM2e
Maximum RAM amount
48 GB
80 GB
Memory bus width
384 Bit
5120 Bit
Memory clock speed
18 GB/s
3.2 GB/s
Memory bandwidth
864.0 GB/s
2,039 GB/s
Display Connectors
4x DisplayPort 1.4a
No outputs
DirectX
12 Ultimate (12_2)
N/A
CUDA
8.9
9.0
OpenCL
3.0
3.0
OpenGL
4.6
N/A
Shader Model
6.7
N/A
Vulkan
1.3
N/A