Compare · GPUs & AI chips
TPU v6e (Trillium) vs Trainium2
RANKED BY DENSE FP16/BF16 TENSOR TFLOPS PER CHIP
#11 VS #12
CHECKED 06 SEPT 2026
Side by side
Verdict
TPU v6e (Trillium) ranks #11 of 20 on Dense FP16/BF16 tensor TFLOPS per chip (918 TFLOPS); Trainium2 ranks #12 (667 TFLOPS).
Rank 11 of 20 · leads
Dense FP16/BF16 tensor TFLOPS per chip
918 TFLOPS
MakerGoogle
Date2024
Verified04 Sept 2026
Evidence
Google Cloud TPU v6e documentation, per-chip specification table: "Peak compute per chip (bf16) 918 TFLOPs". Google does not publish a separate dense FP16 cell for v6e; this row uses the official per-chip BF16 figure. 32 GB HBM per chip.Rank 12 of 20
Dense FP16/BF16 tensor TFLOPS per chip
667 TFLOPS
MakerAWS
Date2024
Verified04 Sept 2026
Evidence
AWS Neuron Trainium2 Architecture page, compute table: each Trainium2 chip delivers "667 BF16/FP16/TF32 TFLOPS" dense; sparse counterpart is 2,563 TFLOPS and was not used. 96 GiB device memory. Trn2 instances are generally available (AWS announcement 3 Dec 2024).Try another pair
More TPU v6e (Trillium) matchups
- TPU v6e (Trillium) vs Instinct MI355X
- TPU v6e (Trillium) vs Instinct MI350X
- TPU v6e (Trillium) vs Blackwell B200 SXM
- TPU v6e (Trillium) vs Blackwell Ultra B300 SXM
- TPU v6e (Trillium) vs Gaudi 3
- TPU v6e (Trillium) vs Instinct MI325X
- TPU v6e (Trillium) vs Instinct MI300X
- TPU v6e (Trillium) vs H200 SXM
- TPU v6e (Trillium) vs H100 SXM
- TPU v6e (Trillium) vs Instinct MI300A
- TPU v6e (Trillium) vs TPU v5p
- TPU v6e (Trillium) vs Gaudi 2
Source Vendor product pages and official spec documents (AMD Instinct MI355X / MI350X / MI325X / MI300X / MI300A / MI250X / MI250 / MI210; NVIDIA…
Last checked 06 Sept 2026