BANANDRE

Tagged with

3 articles found

Tencent’s DSpine Just Made 27B Model Inference 3x Faster, But There’s a Catch
LLM Inference
Featured

Tencent’s DSpine Just Made 27B Model Inference 3x Faster, But There’s a Catch

Tencent’s new speculative decoding method DSpine delivers 3x faster inference for Qwen3.8-27B. Here’s how it works and where it falls short.

#LLM Inference#performance optimization#qwen...
Read More
Uber Eats Cut Search Latency in Half. The “Secret” Will Annoy You.
agentic coding

Uber Eats Cut Search Latency in Half. The “Secret” Will Annoy You.

Uber rebuilt its Eats search pipeline and slashed latency by 50%. The kicker? The biggest win was simply stopping work on results nobody sees.

#agentic coding#latency#performance optimization...
Read More
Apple’s Neural Engine Was Never Going to Save Us
hardware architecture

Apple’s Neural Engine Was Never Going to Save Us

A deep dive into the reverse-engineered ANE reveals why Apple’s beloved NPU was architected for a world that no longer exists.

#hardware architecture#neural engine#performance optimization
Read More