- AWS becomes the first cloud provider to offer rentable PCIe 6.0 processors
- Graviton5 combines 192 arm cores with 96 PCIe lanes
- Memory bandwidth exceeds 800 GB/s across AWS’s latest server platform
AWS has quietly achieved a milestone that neither AMD nor Intel reached first in commercially available cloud infrastructure by deploying a PCIe 6.0-capable processor.
The company’s Graviton5 CPU is now generally available via Amazon EC2 M9g and M9gd instances, allowing customers to rent PCIe 6.0 hardware on an hourly basis.
While this development sounds significant on paper, practical benefits are still difficult to identify for most users at this stage of implementation.
PCIe 6.0 arrives in the cloud before it reaches most hardware
Developed by Annapurna Labs, the Graviton5 adopts a chiplet design built on TSMC’s 3nm manufacturing process technology.
The processor combines four compute dies, each containing 48 Arm v3 cores, bringing the total number of cores to 192.
AWS says each core carries 1MB of dedicated cache, while the platform integrates 12 DDR5 memory channels operating at speeds up to DDR5-8800.
According to company figures, the memory subsystem can deliver more than 800 GB/s of aggregate bandwidth across demanding workloads.
The processor also includes 96 PCIe 6.0 lanes, making it the first cloud CPU customers can actively access with PCIe 6.0 connectivity.
Communication between chiplets relies on a coherent interconnect capable of transferring data at 420 GB/s while maintaining consistent operation.
AWS claims Graviton5 can deliver performance improvements reaching 25% compared to previous generations deployed across its infrastructure.
Additional figures suggest that application workloads can run 35% faster, while database operations improve by 30% under the right conditions.
Network bandwidth reportedly increases by as much as 15%, while storage bandwidth increases by approximately 20% across instance categories.
For larger deployments, AWS says network throughput can double compared to previous offerings available through its cloud platform.
Why PCIe 6.0 might not matter much yet
The challenge is that PCIe 6.0 alone does not automatically transform application performance unless the surrounding hardware can take advantage of the additional bandwidth.
This limitation becomes more apparent when examining storage devices that are able to take advantage of the newer interface standard today.
Micron’s 9650 NVMe SSD is among the first PCIe 6.0 drives to reach commercial availability, although its audience remains hyperscale operators.
The SSD can reportedly achieve sequential read speeds of 28GB/s, nearly double the throughput normally associated with PCIe 5.0 storage.
Still, these drives are largely intended for AI inference environments rather than conventional enterprise or cloud computing workloads.
The same pattern appears in Teamgroup’s recently announced PCIe 6.0 SSD, which reaches 28GB/s but is still far from mainstream implementation.
For many AWS customers, processor architecture, memory bandwidth, cache capacity, and software optimization are likely to matter much more.
The M9gd instances also include local SSD storage that reaches 11.4 TB capacity and delivers 30% higher IOPS than predecessors.
Although PCIe 6.0 gives AWS early technology distinction, meaningful gains will depend heavily on broader ecosystem adoption.
At present, the performance appears more important as an infrastructure milestone than as a feature that instantly changes everyday cloud workloads.
Via The Guru of 3D / Wccftech
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews and opinions in your feeds.
