Jake Tech Lab

NVIDIA H100 vs NVIDIA L40S

NVIDIA H100 comes out ahead of NVIDIA L40S on 5 of the 7 things that decide this, going by what each maker publishes. The one that matters for you may be one the leader loses.

NVIDIA H100Winner
NVIDIA H100
6.0
Fair
5 of 7 rounds
vs
NVIDIA L40S
NVIDIA L40S
6.5
Fair
2 of 7 rounds

The verdict

NVIDIA H100 takes it, 5–2 on rounds.

On the published figures, NVIDIA H100 takes 5 of 7 rounds. Treat that as a starting point and check the current figures on both pages, because prices and plans move.

Choose NVIDIA H100 if…

  • • You want NVIDIA H100's edge on how much memory it has and how fast its memory is

Choose NVIDIA L40S if…

  • • You want NVIDIA L40S's edge on how much power it draws and how new it is

Round by round

  1. Round 1How much memory it hasNVIDIA H100 wins
    NVIDIA H1008/10

    80 GB HBM3

    gpu memory

    Source: nvidia.com ↗

    NVIDIA L40S5/10

    48 GB GDDR6

    gpu memory

    • 48 GB GDDR6, against 80 GB HBM3

    Source: nvidia.com ↗

    NVIDIA H100 wins this round on what the company itself publishes.

  2. Round 2How fast its memory isNVIDIA H100 wins
    NVIDIA H1008/10

    3.35 TB/s

    memory bandwidth

    Source: nvidia.com ↗

    NVIDIA L40S5/10

    864 GB/s

    memory bandwidth

    • 864 GB/s, against 3.35 TB/s

    Source: nvidia.com ↗

    NVIDIA H100 wins this round on what the company itself publishes.

  3. Round 3How fast it runs FP8 AI mathNVIDIA H100 wins
    NVIDIA H1008/10

    1,979 TFLOPS

    fp8 tensor core

    Source: dam-cdn.nvd.orangelogic.com ↗

    NVIDIA L40S5/10

    733 TFLOPS

    fp8 tensor core

    • 733 TFLOPS, against 1,979 TFLOPS

    Source: nvidia.com ↗

    NVIDIA H100 wins this round on what the company itself publishes.

  4. Round 4How fast it runs FP16 AI mathNVIDIA H100 wins
    NVIDIA H1008/10

    989 TFLOPS

    fp16 tensor core

    Source: dam-cdn.nvd.orangelogic.com ↗

    NVIDIA L40S5/10

    362 TFLOPS

    fp16 tensor core

    • 362 TFLOPS, against 989 TFLOPS

    Source: nvidia.com ↗

    NVIDIA H100 wins this round on what the company itself publishes.

  5. Round 5How much power it drawsNVIDIA L40S wins
    NVIDIA H1005/10

    700 W

    max tdp

    • 700 W, against 350 W

    Source: nvidia.com ↗

    NVIDIA L40S8/10

    350 W

    max power consumption

    Source: nvidia.com ↗

    NVIDIA L40S wins this round on what the company itself publishes.

  6. Round 6Which memory generation it usesNVIDIA H100 wins
    NVIDIA H1008/10

    HBM3

    memory type

    Source: dam-cdn.nvd.orangelogic.com ↗

    NVIDIA L40S5/10

    GDDR6

    memory type

    • GDDR6, against HBM3

    Source: nvidia.com ↗

    NVIDIA H100 wins this round on what the company itself publishes.

  7. Round 7How new it isNVIDIA L40S wins
    NVIDIA H1005/10

    2022

    year announced

    • 2022, against 2023

    Source: nvidianews.nvidia.com ↗

    NVIDIA L40S8/10

    2023

    year announced

    Source: nvidianews.nvidia.com ↗

    NVIDIA L40S wins this round on what the company itself publishes.

NVIDIA H100NVIDIA H100
NVIDIA L40SNVIDIA L40S

Side by side

Overall score
6.0
6.5
Best
Price
See price
See price
How fast its memory is
5.7
6.0
How much memory it has
5.4
6.0
How much power it draws
7.1
7.0
How fast it runs FP16 AI math
5.9
7.0
How new it is
5.4
8.0
Head-to-head findings
How fast its memory is
3.35 TB/s
864 GB/s
How much memory it has
80 GB HBM3
48 GB GDDR6
How much power it draws
700 W
350 W
How fast it runs FP16 AI math
989 TFLOPS
733 TFLOPS
How new it is
2022
2023
Which memory generation it uses
HBM3
GDDR6
How fast it runs FP8 AI math
1,979 TFLOPS
1,466 TFLOPS
How it handles scientific FP64 math
67 TFLOPS
—
Record
Head-to-head record
2–7
1–1–1
Head-to-heads won
22 %
50 %
Rounds won
31 %
47 %
Pros
      Cons
      • 80 GB HBM3, against 288 GB HBM3E
      • 3.35 TB/s, against 8 TB/s
      • 1,979 TFLOPS, against 5 PFLOPs
      • 989 TFLOPS, against 2.5 PFLOPs
      • 350 W, against 72 W
      • 48 GB GDDR6, against 80 GB HBM3
      • 864 GB/s, against 3.35 TB/s
      • 733 TFLOPS, against 1,979 TFLOPS
      Add a third option to this comparison →

      More AI accelerators match-ups