LLM inference load shows 3 daily peaks, Friday down 20%
jwt0625 · x · 2026-09-02
Shared a chart of LLM inference workloads (coding and conversation) over a week. The data reveals three distinct peaks per day, with Friday's peak dropping approximately 20% compared to Tuesday.
Related event: LLM inference workloads and carbon emissions fluctuate weekly(2 posts)→
More from Infra
- Bought a $6,279 RTX 5090 Prebuilt: Was it the right choice for local AI? — Sexyvette07 · 2026-09-02
- DARPA's 1.8 Gigapixel ARGUS Camera Recorded 1M TB Daily in 2013 — aakashgupta · 2026-09-02
- Dev Petitions to Make Quantization Aware Training the Norm for Open Weights — MADxMORON · 2026-09-02
- 5 Ways to Run LLMs Locally for Privacy and Cost Savings — goyalshaliniuk · 2026-09-02
- Microsoft Paper Introduces Slasher for Cloud Datacenter Power Flexibility — jwt0625 · 2026-09-02
- MCP usage surged 98x after protocol went stateless — lee_stott · 2026-09-02