Home GPU Comparison NVIDIA H100 PCIe vs NVIDIA L40

NVIDIA H100 PCIe vs NVIDIA L40

AI GPU We compared a GPU: 80GB VRAM H100 PCIe and a Professional market GPU: 48GB VRAM L40 to see which GPU has better performance in key specifications, benchmark tests, power consumption, etc.

Main Differences

NVIDIA H100 PCIe 's Advantages
More VRAM (80GB vs 48GB)
Larger VRAM bandwidth (2039GB/s vs 864.0GB/s)
NVIDIA L40 's Advantages
Released 7 months late
Boost Clock has increased by 42% (2490MHz vs 1755MHz)
3584 additional rendering cores
Lower TDP (300W vs 350W)

Score

Benchmark

FP32 (float)
H100 PCIe
51.22 TFLOPS
L40 +76%
90.52 TFLOPS
Blender
H100 PCIe
4845
L40 +95%
9486
VS
L40

Graphics Card

Mar 2022
Release Date
Oct 2022
Tesla Hopper
Generation
Tesla Ada
AI GPU
Type
Professional
PCIe 5.0 x16
Bus Interface
PCIe 4.0 x16

Clock Speeds

1095 MHz
Base Clock
735 MHz
1755 MHz
Boost Clock
2490 MHz
1593 MHz
Memory Clock
2250 MHz

Memory

80GB
Memory Size
48GB
HBM2e
Memory Type
GDDR6
5120bit
Memory Bus
384bit
2039GB/s
Bandwidth
864.0GB/s

Render Config

114
SM Count
142
-
Compute Units
-
14592
Shading Units
18176
456
TMUs
568
24
ROPs
192
456
Tensor Cores
568
-
RT Cores
142
256 KB (per SM)
L1 Cache
128 KB (per SM)
50 MB
L2 Cache
96 MB

Theoretical Performance

42.12 GPixel/s
Pixel Rate
478.1 GPixel/s
800.3 GTexel/s
Texture Rate
1414 GTexel/s
204.9 TFLOPS
FP16 (half)
90.52 TFLOPS
51.22 TFLOPS
FP32 (float)
90.52 TFLOPS
25.61 TFLOPS
FP64 (double)
1414 GFLOPS

Graphics Processor

GH100
GPU Name
AD102
-
GPU Variant
AD102-895-A1
Hopper
Architecture
Ada Lovelace
TSMC
Foundry
TSMC
4 nm
Process Size
4 nm
80 billion
Transistors
76.3 billion
814 mm²
Die Size
608 mm²

Board Design

350W
TDP
300W
750 W
Suggested PSU
700 W
No outputs
Outputs
4x DisplayPort 1.4a
1x 16-pin
Power Connectors
1x 16-pin

Graphics Features

N/A
DirectX
12 Ultimate (12_2)
N/A
OpenGL
4.6
3.0
OpenCL
3.0
N/A
Vulkan
1.3
9.0
CUDA
8.9
N/A
Shader Model
6.6

Related GPU Comparisons

© 2024 - TopCPU.net   Contact Us Privacy Policy