AWS Graviton5 beats Intel in encoding but trails AMD EPYC

For streaming platforms, these benchmarks highlight that ARM-based custom silicon like Graviton5 is now a viable, and often superior, alternative to legacy x86 architectures for high-efficiency AV1 encoding. While AMD's Turin remains the performance king for raw horsepower, AWS's vertical integration provides a competitive edge in common utility tasks like 7-Zip compression. This shift pressures Intel to recapture its lead in an encoding standard it helped found. Strategists should monitor if Graviton's persistent performance gains translate to lower encoding egress costs compared to EPYC-backed instances, as high-density encoding efficiency becomes the primary lever for reducing per-stream infrastructure spend.

The benchmark results come as AWS recently moved Graviton5 into general availability in June 2026. According to Amazon, the Graviton5 chip features 192 Arm Neoverse V3 cores and provides up to 25% better compute performance than its predecessor, Graviton4. This latest generation was specifically optimized for memory-intensive and parallel workloads, incorporating PCIe Gen 6 and DDR5-8800 memory support, which AWS claims is the fastest memory available in the cloud (per Amazon and SiliconAngle, June 2026). This architectural push is part of a broader trend where hyperscale providers are shifting away from merchant silicon to custom designs to maximize VM density and energy efficiency.

Competitive pressure is mounting as AMD’s 5th-Gen EPYC Turin line, launched in late 2024, continues to dominate multi-threaded throughput. Per Phoronix and TechPowerUp (July 2025), a 192-core EPYC 9965 can deliver nearly triple the performance of legacy Intel Xeon flagships in specific AI and video transcoding tasks. Conversely, Intel has focused its Xeon 6 Granite Rapids strategy on AI inference acceleration via its AMX instructions to counter AMD's raw core-count advantage. Despite these efforts, industry data from December 2025 indicated that AMD’s server market share reached 27.8%, while ARM-based processors climbed to 13.2% of total server sales (per Introl, February 2026). For the video streaming industry, this means the choice of instance type now depends heavily on whether the priority is single-threaded latency, where Intel remains competitive, or massive parallel throughput, where AMD and AWS's Graviton are currently outdistancing the field.