Explore/agent app/Memory-Bound but Not Bandwidth-Limited: The Physical AI Inference Gap in Batch-1 LLM Decode
M

Josef Chen/Memory-Bound but Not Bandwidth-Limited: The Physical AI Inference Gap in Batch-1 LLM DecodeUnknown

Physical AI systems, including robots, autonomous vehicles, embodied agents and edge copilots, often run a different inference workload from cloud LLM serving: single-stream, batch-1 autoregressive decode, where one robot, camera feed or user session waits on the next token. This workload is usually described as memory-bandwidth-bound. Each decode step streams model weights and the active KV cache, so latency should scale with peak HBM bandwidth. We show that this account is true but incomplete. We measure batch-1 decode for three 7 to 8B-class GQA transformers across four NVIDIA GPUs: H100 SXM5, A100-80GB SXM4, L40S and L4. We evaluate context lengths from 2048 to 16384, producing 44 valid cells under a controlled bf16 SDPA setup. The achieved fraction of peak HBM bandwidth falls as peak bandwidth rises. On the headline Qwen-2.5-7B ctx=2048 cell, an L4 reaches roughly 81 percent of its analytic memory floor, while an H100 reaches only 27 percent. Physical-AI decode is memory-dominated, but faster memory does not translate into proportional latency gains. We test the missing term with a CUDA Graphs A/B experiment. On H100 at ctx=2048, CUDA Graphs improves decode latency by 1.259x across N=10 fresh sessions, with a 95 percent bootstrap confidence interval of 1.253 to 1.267. On L4, the same intervention gives only 1.028x. This isolates a launch-side overhead that becomes visible on fast GPUs but remains mostly hidden on slower, bandwidth-bound GPUs. The deployment implication is that memory savings matter only when the runtime realises them. On L4, bf16 decode sits close to the memory floor, but common quantised paths do not recover the expected 4x weight-traffic reduction: bnb-nf4 reaches 59.36 ms/step and AutoAWQ+Marlin reaches 45.24 ms/step from a 62.32 ms bf16 baseline. GPTQ+ExLlamaV2, with Ada-tuned int4 kernels, reaches 17.36 ms/step.

agent app
GitHubCompare
Refreshed 23h ago
OverviewActivity52wAlternativesDocs
Stars0
Forks0
HF Downloads—30d
Last commit—
Refreshed23h ago
Project healthUnknownNo activity data.
Production readinessResearch / EarlyBest for exploration and prototyping.
Risk notesUnknown licenseVerify license before production use.
AgentHub Score
48 / 100
Composite score from 6 signals. How we score →
Active project
48Score
Growth
40C
Activity
30C
Documentation
70C+
Maturity
45C
Community
42C
Production
58C
GitHub stars · 90 days0 +0.0%
30d90d1y
latest release
Commit activity · 52 weeksActive contributor activity
LowHigh
JunSepDecMarNow
Practical assessment
Should you use it?

✓ Best for

  • Research and experimentation
  • Prototype development
  • Learning agentic patterns

◎ Strengths

  • Active community
  • Open source
  • Well-documented API

✕ Not ideal for

  • Untested at scale without validation
  • Teams without AI/ML expertise

⚠ Watch-outs

  • Review changelog before updating
  • Verify license for commercial use
Technical details
What's inside
Language—
License—
Sourcearxiv
Open source✗ No
Commercial use—
Docs—
Demo—

AgentHub Score

48
Score 48/100
Below average

Alternatives

C
crewai
26.1k · Multi-Agent
87
A
autogen
42.7k · Multi-Agent
71
S
smolagents
11.2k · Coding
84
O
openai-agents-python
9.4k · Multi-Agent
81
Compare all →

Recent activity

Latest commit ——
Indexed by AgentHub crawler23h ago
Monitor for new releasesongoing