# IntuVision PC — CPU / RAM load analysis (2026-07-21)
**Host:** `DESKTOP-1JL79ER` · **Specs:** ATX, Intel **i5-10400F** (6-core/12-thread),
**16 GB RAM**, **NVIDIA GTX 1650**, 1 TB HDD · **Software:** IntuVision VA **v17.06**,
**4 cameras**. Source: three Task Manager screenshots (`screenshots/`).
## What the screenshots show
Headline totals were consistent across all three captures: **CPU 100% · Memory 91% ·
Disk 15–20% · Network 1–2% · GPU 2–4%.**
**CPU — two IntuVision processes consume ~90% of the whole machine:**
| Process | CPU | Notes |
|---|---|---|
| intuVision – **Machine Controller** | **58.7–60.3%** | "Very high" / "High" power draw |
| intuVision – **Site Controller** | **30.8–31.8%** | "Very high"; ~12 Mbps network (camera ingest) |
| intuVision – Master Tracker ×several | 0.6–2.1% each | one per camera stream |
| intuVision – Digital Video Controller | ~1.4% | — |
| PostgreSQL Server ×several | <1.5% each | the event database |
**Memory — 91% of 16 GB (~14.5 GB used), almost no free headroom:**
| Process | RAM |
|---|---|
| intuVision – Digital Video Controller | ~933 MB |
| intuVision – Master Tracker (×3 visible) | ~933 MB, ~687 MB, ~286 MB |
| intuVision – Site Controller | ~91 MB |
| Multiple PostgreSQL Server instances + Windows overhead | remainder |
**GPU — the GTX 1650 is essentially idle: 0–4% utilization.** `NVIDIA Container` is
running (driver present), and an `AXIS Device Manager Service` is present (the cameras
appear to be **AXIS**). Minor background load from Google Chrome (~225 MB), Windows
"News and interests," and Search.
## Diagnosis — this is a configuration problem, not a hardware limit
**All the heavy video work is running on the CPU while a capable GPU sits unused.** The
Machine Controller + Site Controller are saturating the 6-core i5-10400F doing decode
and analytics that the **GTX 1650 should be accelerating** — but GPU utilization is
~0%. This matches IntuVision's president's own comment ("I don't know why he wasn't
turning on the GPU"). Version 17.06 is the *lighter*, pre-"model-based" release, so on
these specs 100% CPU "just doesn't make sense" for 4 cameras unless GPU offload is off
and/or per-camera settings are heavy.
Memory at 91% is a secondary problem: 16 GB leaves ~1.5 GB free, so the machine has no
slack — a contributor to instability (and consistent with the PSU/thermal stress from
running flat-out continuously).
## Recommendations (highest impact first)
1. **Enable GPU acceleration in IntuVision** — assign video decode/detection to the GTX
1650. This is the single biggest fix; it should pull CPU down from 100% dramatically.
(This is the core answer to "what config changes stabilize the machine.")
2. **Verify the NVIDIA driver / CUDA** is the version IntuVision v17.06 expects, so the
GPU is actually usable by the analytics pipeline.
3. **Tune per-camera analytics** — detection resolution, frame rate, and analytic zones —
to cut the per-stream (Master Tracker) CPU/RAM cost, and to reduce false alarms.
4. **Add RAM 16 GB → 32 GB** (~$40) for headroom; 91% is risky and cheap to fix.
5. **Trim Windows bloat** — disable "News and interests," background Search indexing, and
close Chrome on the box. Minor, but free.
6. After GPU-on + tuning, headroom should comfortably support the board's intended return
to **6 cameras** (IntuVision says these specs are good for many more).
## Ties to the IntuVision call
This is the technical backing for RF's Q1 to Sadiye ("what configuration changes are
needed to stabilize the machine?"). It also supports the redundancy plan (Q3): with a
correctly-configured node, a ~$600 refurbished second PC as a warm spare is realistic.
See `../vendor/intuvision-email-thread-2026-07.md` and `intuvision-call-2026-07-board-summary.md`.