> Markdown version of [/events/conference-india-2026/sessions/2003-how-8-gpus-became-16](https://www.wearedevelopers.com/events/conference-india-2026/sessions/2003-how-8-gpus-became-16). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# How 8 GPUs Became 16: Rethinking LLM Inference Through Parallelism, Memory, and Runtime Design


- **Event:** Conference India 2026

## Description

Most teams scale LLM inference by adding more GPUs. This talk shows why that approach is fundamentally flawed—and how rethinking parallelism strategy, memory layout, and runtime configuration can deliver up to 2× production throughput improvements without changing hardware.

I will break down a real-world high-concurrency inference system serving millions of users, where the key breakthrough came not from bigger clusters but from architectural decisions: tensor vs data parallelism trade-offs, KV cache optimization, and Mixture-of-Experts routing efficiency.

The session dives deep into vLLM internals, FP8 quantization impacts, KV cache memory tuning, and GPU communication overheads, showing how reducing synchronization and improving locality dramatically changes system performance.


## Speaker

### [Purnima Kumari](https://www.wearedevelopers.com/@purnima-kumari)

Senior Solutions Architect @ DigitalOcean | Where cloud meets storytelling 🌊📖


## Related talks at this congress

- [Architecting Trusted AI: Coupling Multi-Agent Systems with dbt Semantic Store for Zero-Hallucination](https://www.wearedevelopers.com/events/conference-india-2026/sessions/1976-architecting-trusted) — Pushkar Mishra
- [How We Cut Our API's p99 Latency from Minutes to Under a Second](https://www.wearedevelopers.com/events/conference-india-2026/sessions/1987-how-we-cut-our-api-s) — Deepak Agrawal
- [React Under Fire: Rendering Performance for Streaming Agent UIs](https://www.wearedevelopers.com/events/conference-india-2026/sessions/1992-react-under-fire) — Abhirupa Mitra
- [AI Is the Easy Part: Why Faster Engineers Don't Make Faster Businesses](https://www.wearedevelopers.com/events/conference-india-2026/sessions/1975-ai-is-the-easy-part) — Bharat Sharma

## Watch remotely

Can’t make it to Bengaluru? Watch this session live with Pro. You also get:

- All full videos, bookmarks, and playlists
- World Congress livestreams

[See pricing](https://www.wearedevelopers.com/pricing)

## Links

- [Get tickets](https://www.wearedevelopers.com/conference-india/tickets)