Jake Tech Lab

Google TPU v6e Trillium vs Google TPU v7 Ironwood

Google TPU v7 Ironwood comes out ahead of Google TPU v6e Trillium on 6 of the 6 things that decide this, going by what each maker publishes. The one that matters for you may be one the leader loses.

Google TPU v6e Trillium
Google TPU v6e Trillium
5.0
Weak
0 of 6 rounds
vs
Google TPU v7 IronwoodWinner
Google TPU v7 Ironwood
7.4
Good
6 of 6 rounds

The verdict

Google TPU v7 Ironwood takes it, 6–0 on rounds.

On the published figures, Google TPU v7 Ironwood takes 6 of 6 rounds. Treat that as a starting point and check the current figures on both pages, because prices and plans move.

Choose Google TPU v6e Trillium if…

  • • You have a reason to pick Google TPU v6e Trillium that is not on this list

Choose Google TPU v7 Ironwood if…

  • • You want Google TPU v7 Ironwood's edge on how much memory it has and how fast its memory is

Round by round

  1. Round 1How much memory it hasGoogle TPU v7 Ironwood wins
    Google TPU v6e Trillium5/10

    32 GB HBM

    hbm capacity per chip

    • 32 GB HBM, against 192 GB HBM

    Source: docs.cloud.google.com ↗

    Google TPU v7 Ironwood8/10

    192 GB HBM

    hbm capacity per chip

    Source: docs.cloud.google.com ↗

    Google TPU v7 Ironwood wins this round on what the company itself publishes.

  2. Round 2How fast its memory isGoogle TPU v7 Ironwood wins
    Google TPU v6e Trillium5/10

    1638 GBps

    hbm bandwidth per chip

    • 1638 GBps, against 7.37 TB/s

    Source: docs.cloud.google.com ↗

    Google TPU v7 Ironwood8/10

    7.37 TB/s

    hbm bandwidth per chip

    Source: blog.google ↗

    Google TPU v7 Ironwood wins this round on what the company itself publishes.

  3. Round 3How fast it runs FP8 AI mathGoogle TPU v7 Ironwood wins
    Google TPU v6e Trillium5/10

    918 TFLOPs

    peak fp8 compute per chip (tpu7x comparison table)

    • 918 TFLOPs, against 4,614 TFLOPs

    Source: docs.cloud.google.com ↗

    Google TPU v7 Ironwood8/10

    4,614 TFLOPs

    peak fp8 compute per chip

    Source: blog.google ↗

    Google TPU v7 Ironwood wins this round on what the company itself publishes.

  4. Round 4How fast it runs FP16 AI mathGoogle TPU v7 Ironwood wins
    Google TPU v6e Trillium5/10

    918 TFLOPs

    BF16 peak per chip (Google publishes no FP16 figure)

    • 918 TFLOPs, against 2,307 TFLOPs

    Source: docs.cloud.google.com ↗

    Google TPU v7 Ironwood8/10

    2,307 TFLOPs

    BF16 peak per chip (Google publishes no FP16 figure)

    • peak bf16 compute per chip

    Source: docs.cloud.google.com ↗

    Google TPU v7 Ironwood wins this round on what the company itself publishes.

  5. Round 5How fast chips talk to each otherGoogle TPU v7 Ironwood wins
    Google TPU v6e Trillium5/10

    800 GBps

    ici bandwidth per chip, bidirectional

    • 800 GBps, against 1.2 TBps

    Source: docs.cloud.google.com ↗

    Google TPU v7 Ironwood8/10

    1.2 TBps

    ici bandwidth per chip, bidirectional

    Source: blog.google ↗

    Google TPU v7 Ironwood wins this round on what the company itself publishes.

  6. Round 6How new it isGoogle TPU v7 Ironwood wins
    Google TPU v6e Trillium5/10

    May 15, 2024

    announcement blog date

    • May 15, 2024, against Apr 09, 2025

    Source: cloud.google.com ↗

    Google TPU v7 Ironwood8/10

    Apr 09, 2025

    announcement date

    Source: blog.google ↗

    Google TPU v7 Ironwood wins this round on what the company itself publishes.

Google TPU v6e TrilliumGoogle TPU v6e Trillium
Google TPU v7 IronwoodGoogle TPU v7 Ironwood

Side by side

Overall score
5.0
7.4
Best
Price
See price
See price
How fast its memory is
5.0
7.3
How much memory it has
5.0
7.3
How fast it runs FP16 AI math
5.0
7.3
How new it is
5.0
7.8
Head-to-head findings
How fast its memory is
1638 GBps
7.37 TB/s
How much memory it has
32 GB HBM
192 GB HBM
How much power it draws
—
—
How fast it runs FP16 AI math
918 TFLOPs
2,307 TFLOPs
How new it is
May 15, 2024
Apr 09, 2025
Which memory generation it uses
—
HBM3E
How fast it runs FP8 AI math
918 TFLOPs
4,614 TFLOPs
How it handles scientific FP64 math
—
—
Record
Head-to-head record
0–1
3–1
Head-to-heads won
0 %
75 %
Rounds won
0 %
75 %
Pros
    • peak bf16 compute per chip
    Cons
    • 32 GB HBM, against 192 GB HBM
    • 1638 GBps, against 7.37 TB/s
    • 918 TFLOPs, against 4,614 TFLOPs
    • 918 TFLOPs, against 2,307 TFLOPs
    • 192 GB HBM, against 288 GB HBM3E
    • 7.37 TB/s, against 8 TB/s
    • 4,614 TFLOPs, against 5 PFLOPs
    • 2,307 TFLOPs, against 2.5 PFLOPs
    Add a third option to this comparison →

    More AI accelerators match-ups