RK3576 PVT vs EVT: Hardware Modules and Interface Test
2026-10-22 (UTC+7) · Wilber Zheng · read live over SSH (sysfs, /proc/device-tree, i2c names); read-only, no reboot
Modules
| Component | PVT Shenzhen | evt-01 | |
|---|
| eMMC | TY2964, 58 GiB, manfid 0x70, oemid 0x0100, fw 0x5b, made 10/2025 | MMC64G, 58 GiB, manfid 0x32, oemid 0x0101, fw 0x03, made 06/2025 | different part |
| SoC / CPU | Rockchip RK3576, 8 cores (4 × 2016 MHz + 4 × 2208 MHz) | same |
| RAM | 4 GB, DDR max 2112 MHz | same |
| PMIC | Rockchip RK806 (i2c 1-0023) | same |
| RTC | HYM8563 (i2c 2-0051) | same |
| Audio codec | Everest ES8389 (i2c 3-0010, es8390 driver) | same |
| HDMI-in bridge | Rockchip RK628 (i2c 6-0051) | same |
| Wi-Fi / Bluetooth | AIC 8800D80 combo on USB (a69c:8d81), 2.4 + 5 GHz | same |
| USB hubs | Genesys 05e3:0610 (USB 2.0) + 05e3:0620 (USB 3.2) | same |
| Ethernet | RK3576 GMAC (st_gmac), 1000 Mb/s capable | same |
| GPU / NPU / VPU | Mali (panfrost) 800 MHz · NPU 950 MHz · rkvdec2 / rkvenc2 / RGA | same |
| Display outputs | HDMI-A-1, DP-1 | same |
| Kernel | 6.1.99-rockchip-standard (build of 2026-04-17) | same |
| OS image | dbg455 | dbg464 | software only |
Device tree
| PVT Shenzhen | evt-01 | |
|---|
| Model / compatible | "OptiSigns ProPlayer Gen5 (EVT)" · optisigns,proplayer-gen5-evt | same |
| Nodes / enabled | 407 nodes, 241 enabled | same |
| Node-by-node diff | No node, compatible string or status differs | identical |
Interface test (S8, read-only)
One SSH call per board that only reads sysfs/proc: no reboot, no player stop, no network change. Link gate and restore check before and after; the player kept playing.
| Area | PVT Shenzhen (dbg455) | evt-01 (dbg464) |
|---|
| Result | 23 pass · 0 fail · 0 skip | 23 pass · 0 fail · 2 skip (SD socket faulty on EVT; no monitor) |
| USB | 2 root hubs, xHCI bound; Wi-Fi/BT module enumerated |
| Ethernet | Not cabled (runs on Wi-Fi) | 1000 Mb/s full, 0 errors |
| Wi-Fi | 2.4 + 5 GHz; on 5745 MHz, -57 dBm | 2.4 + 5 GHz; on 5180 MHz, -64 dBm |
| Bluetooth | Controller up, not rfkill-blocked |
| RTC | About 176 days behind (reads 2026-04-29); NTP not synced on the site network | 0 s drift, NTP synced |
| eMMC health | Life used 0-10% (0x01), pre-EOL normal. Bus mode (HS400) is not visible without root. |
| SD slot | Host OK, empty (an SDR104 SDXC card was seen yesterday) | Not judged (faulty socket) |
| HDMI | Xiaomi Mi Monitor, EDID OK, 1920×1080 @ 100 Hz | No monitor (headless) |
| GPU / VPU / NPU | panfrost, vop2, rkvdec2, rkvenc2, RKNPU and RGA bound; /dev nodes present |
| Thermal (idle) | 45–49 °C | 53–57 °C |
| Memory / PCIe | About 3.9 GB usable; no PCIe devices |
RTC: the PVT system clock is correct now, but the correct time was never written back to the RTC. After a power loss with no network, the board would boot with an April date. Worth checking the RTC write-back path and NTP reachability on that network.
Takeaway: both boards pass every interface check. The only hardware difference is the eMMC part, which is a second source. Both boards run the same EVT device tree, so the PVT has no device tree of its own yet. The second eMMC part is worth a storage benchmark before mass production.