AWS EC2 trn2.48xlarge
The trn2.48xlarge is AWS's single-node Trainium2 instance: sixteen Trainium2 chips joined by NeuronLink-v3 at 1,024 GB/s per chip, with 1,536 GiB of device memory at 46.4 TB/s, 192 vCPUs, 2 TB of host memory and 3,200 Gbps of EFAv3 networking. AWS rates it at 20.8 PFLOPS of dense FP8, 10.7 PFLOPS across dense FP16, BF16 and TF32, and 2.9 PFLOPS of dense FP32, with a separate sparse figure of 41 PFLOPS covering FP8, FP16, BF16 and TF32 but not FP32. Note that the sparse figure is larger than the dense FP8 one, so the two must not be confused: dense FP16 is 10.7 PFLOPS, not 41. AWS's sparse mode runs at the chip's low-precision rate whatever format is fed in, which is why 41 is 1.97x the dense FP8 rate but 3.84x the dense BF16 rate; it is not 2:4 structured sparsity. AWS publishes no power figure for this instance or for the Trainium2 chip, so total system power is left blank rather than estimated. The trn2u.48xlarge variant carries the same specification and is the instance from which UltraServers are composed.
We will point you at suppliers who have it. Free, and no signup.
System Details
GPU Configuration
System Specifications
Precision Performance Breakdown
| Precision | System Performance | Per GPU | Efficiency |
|---|---|---|---|
| 2,900 TFLOPS | 181 TFLOPS | — | |
| 10,700 TFLOPS | 669 TFLOPS | — | |
| 10,700 TFLOPS | 669 TFLOPS | — | |
| 10,700 TFLOPS | 669 TFLOPS | — | |
| 20,800 TFLOPS | 1,300 TFLOPS | — |
Where a vendor states a basis, the figure shown is the dense one, and any figure it publishes is listed separately. Vendors commonly headline the sparse number instead: for NVIDIA's 2:4 structured sparsity that is exactly twice the dense figure, though other vendors' sparse modes do not all follow that ratio.
Powered by AWS Trainium2
This system utilizes 16 × AWS Trainium2 GPUs, each delivering exceptional performance for AI and HPC workloads.
Per GPU TDP
—
Per GPU Memory
96 GB
Process Node
—
Architecture
NeuronCore-v3
Related Systems
Documentation & Resources
Flopper Spec Sheet
AWS EC2 trn2.48xlarge specifications, generated from our database. Printable.
Vendor product page
AWS EC2 trn2.48xlarge documentation
AWS announces Amazon EC2 Inf2 instances (Preview)
AWS • 2022-11-29
Typical Use Cases
The AWS EC2 trn2.48xlarge runs 16× Trainium2 GPUs over NeuronLink-v3 at 1,024 GB/s per chip within the instance; EFAv3 networking at 3,200 Gbps, delivering 20,800 TFLOPS FP8 dense.