diff options
| author | Yuren Hao <yurenh2@illinois.edu> | 2026-07-17 16:01:23 -0500 |
|---|---|---|
| committer | Yuren Hao <yurenh2@illinois.edu> | 2026-07-17 16:01:23 -0500 |
| commit | 98935725be4e33bd4e0f6830bf56e180fd1c98ed (patch) | |
| tree | ddb3b5f6b0280b1a600c2bb0de792a62d025c8cd /hw_sim/NOTES.md | |
| parent | 7ae978808165d29019d8cb814dbbbb6246e82805 (diff) | |
Energy ledger: SPICE core 2.87pJ/MAC; boards lose ~1000x (trainability demos); coherent integrated projection 0.21-0.63 pJ/MAC = parity-to-5x vs digital, ADC-dominated — honest range, not CIM marketing
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
Diffstat (limited to 'hw_sim/NOTES.md')
| -rw-r--r-- | hw_sim/NOTES.md | 17 |
1 files changed, 17 insertions, 0 deletions
diff --git a/hw_sim/NOTES.md b/hw_sim/NOTES.md index 2bac023..eee68c6 100644 --- a/hw_sim/NOTES.md +++ b/hw_sim/NOTES.md @@ -67,3 +67,20 @@ differential resistive columns with: per-bit ladder mismatch (fixed device), per NEXT candidates (autonomy line): scale toy to 32x32 (overnight class); add settle-transient into the loop (replace DC solves at the found (R_f,C_f) point) to couple timing and training; port the ride/beta story onto the circuit noise floor (nudge amplitude sweep vs 139 uV). + +## Energy ledger (energy_ledger.py, 2026-07-17): SPICE core + datasheet periphery vs digital +SPICE-measured analog network core: 2.87 pJ/MAC (resistive burn over the 4.2 us discrete dwell). +| scenario | pJ/MAC (wiring 1-3x band) | vs digital INT8 system 0.3-1 pJ/MAC | +|---|---|---| +| MVP discrete parts | 487-1462 | loses ~1000x (op-amp quiescent x long dwell + discrete ADC) | +| T64 word-streaming | 587-1562 | loses ~1000x (+ reload/DRAM tax) | +| Integrated weight-stationary (coherent: C/100 -> 100 ns dwell) | 0.21-0.63 | **0.5x-4.8x: parity to ~5x win, ADC-dominated** | +HONEST CONCLUSIONS: +1. The boards (MVP/T64) are trainability demos, never efficiency demos — say it before referees do. +2. The integrated projection at 8-bit lands at PARITY-TO-5x, not the 10-100x of CIM marketing; + the residual is the ADC tax. Paths beyond: fewer/narrower reads, analog inter-layer + accumulation, low-precision contrast reads. +3. EP's energy contribution is CATEGORICAL, not per-MAC: it makes TRAINING possible on analog + fabric at all (inference-only CIM can't train; digital training is the displaced baseline). +4. Method: analog side SPICE-measured (+-2-3x wiring band), digital side literature constants + (Horowitz/H100 envelope) — the standard comparison protocol, uncertainty stated. |