Skip to content
Update packet (01 02)

Update packet (01 02)

The update packet carries all values that a segment LCD of this family can display. Which of them a particular model actually shows depends on its segment layout.

Byte layout

Offsets refer to the 64-byte packet without report ID (see Transport).

OffsetHexContentEncodingRange
00x00reservedalways 0x00–
10x01Command0x01–
20x02Subcommand0x02–
30x03CPU temperatureinteger part0–255
40x04CPU temperaturehundredths ×1000–99
50x05CPU temperature unit0 = °C, 1 = °F ³0–1
60x06CPU usagepercent, integer part (hundredths: offset 48)0–100
70x07CPU powerwatts mod 100 (tens + ones)0–99
80x08CPU powerhundredths ×1000–99
90x09CPU clockMHz ÷ 100 (integer)0–99
100x0ACPU clockMHz mod 1000–99
110x0BCPU voltageinteger part (V)0–255
120x0CCPU voltagehundredths ×100 ¹0–99
130x0DGPU temperatureinteger part0–255
140x0EGPU temperaturehundredths ×1000–99
150x0FGPU temperature unit0 = °C, 1 = °F ³0–1
160x10GPU usagepercent, integer0–100
170x11GPU powerwatts mod 1000–99
180x12GPU powerhundredths ×1000–99
190x13GPU clockMHz ÷ 1000–99
200x14GPU clockMHz mod 1000–99
210x15Fan speedRPM ÷ 1000–99
220x16Fan speedRPM mod 1000–99
230x17Pump speedRPM ÷ 1000–99
240x18Pump speedRPM mod 1000–99
250x19Yearcentury (20 for 2026)0–99
260x1AYearyear mod 100 (26 for 2026)0–99
270x1BMonth1–121–12
280x1CDay1–311–31
290x1DHour0–230–23
300x1EMinute0–590–59
310x1FSecond0–590–59
320x20Day of week0 = Sunday … 6 = Saturday0–6
330x21RAM usagepercent, integer part (hundredths: offset 50)0–100
340x22CPU powerwatts ÷ 100 (hundreds)0–99
350x23GPU powerwatts ÷ 1000–99
360x24unused0x00–
370x25unknown ²original software (LB360-S): always 5?
38–47unused0x00–
480x30CPU usage ²hundredths ×100 (always 0 at 100 %)0–99
490x31unused0x00–
500x32RAM usage ²hundredths ×1000–99
51–63unused0x00–

¹ The original software (LA series, version 2025-03) sends floor(V) × 100 mod 256 here, i.e. the value 100 (0x64) for every 1.x V core voltage – presumably a bug. Since no known display shows the voltage, this has no consequences. The reference implementation sends the real two decimals.

² These bytes are not known from the LA analysis; they come from the captures of the TT SEGMENT LCD Software for the LB360-S. 48 and 50 are clearly identified: at 100 % CPU usage, byte 48 is 0 in all packets, and byte 33 + byte 50/100 yields a continuous RAM curve (e.g. 27.98 → 28.03). Byte 37 is always 5, meaning unknown; the reference implementation sends 0 there – the display works with that.

³ The LB360-S does not convert and always shows the °C symbol; with 1, values ≥ 100 are only shown with two digits (mod 100) instead of being capped at 99. For this model always send 0 – details in Verification.

Encoding rules

All values are single unsigned bytes. Multi-digit quantities are split into base-100 digits (not binary!): 1234 RPM → 12, 34 → 0x0C 0x22, not 0x04 0xD2.

QuantityFormulaExample
Fixed point (temperature, voltage)c = round(v × 100); whole = c div 100; frac = c mod 10042.5 °C → 42, 50
Base 100 (clock, speed)v = min(floor(v), 9999); hi = v div 100; lo = v mod 1004567 MHz → 45, 67
Power (3 bytes)c = round(v × 100); W = c div 100; [34/35] = W div 100; [7/17] = W mod 100; [8/18] = c mod 100123.45 W → [34]=1, [7]=23, [8]=45
Percent (CPU, RAM)fixed point like temperature: [6]/[48] or [33]/[50]37.25 % → 37, 25
Percent (GPU)min(floor(v), 255), no known decimals37.8 % → 37
Fahrenheithost converts: F = C × 9/5 + 32, unit byte 150 °C → 122, flag 1

Recommended handling of edge cases (as implemented in the reference library):

InputBehavior
Value unknown / not availablesend 0
NaN, ±∞, negativesend 0
Overflowclamp to the field’s maximum (temperature 255.99, base 100 9999, power 9999.99)
Rounding for fixed pointfirst round to 1/100, then split – otherwise 42.999 becomes 42, 100 instead of 43, 00
The LB360-S shows temperatures with two digits: values ≥ 100 appear as 99 (tested up to 255). Whether other models meaningfully display values above 99 in the “÷ 100” bytes or decimals above 99 is unknown – only send values within the specified range.

Time

Fields 25–32 contain the host’s local time. The LA display does not show it; whether the round LB360-S display has a clock is an open question. The original software always fills the fields; a library should do the same (it costs nothing, and devices with a clock will then work).

Test vectors

All vectors use the fixed time Friday, 2026-11-20 13:45:30 (bytes 25–32 = 14 1a 0b 14 0d 2d 1e 05). Generated with seglcd send … --time 2026-11-20T13:45:30 --dry-run, backed by the unit tests.

Vector 1 – example from the LA documentation

CPU 42.5 °C, 37 %, 123.45 W, 4567 MHz; fan 1234 RPM; pump 2345 RPM.

00 01 02 2a 32 00 25 17 2d 2d 43 00 00 00 00 00 00 00 00 00 00 0c 22 17 2d 14 1a 0b 14 0d 2d 1e
05 00 01 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

Vector 2 – all fields populated

CPU 65.37 °C, 12 %, 88.2 W, 4450 MHz, 1.2 V; GPU 47 °C, 3 %, 18.5 W, 1714 MHz; fan 1566 RPM; pump 2947 RPM; RAM 41 %.

00 01 02 41 25 00 0c 58 14 2c 32 01 14 2f 00 00 03 12 32 11 0e 0f 42 1d 2f 14 1a 0b 14 0d 2d 1e
05 29 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

Vector 3 – Fahrenheit

CPU 50 °C, GPU 100 °C, unit °F → 122 °F / 212 °F, flags at offsets 5 and 15.

00 01 02 7a 00 01 00 00 00 00 00 00 00 d4 00 01 00 00 00 00 00 00 00 00 00 14 1a 0b 14 0d 2d 1e
05 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

Vector 4 – rounding and clamping

CPU 42.999 °C (→ 43.00), CPU 1234.567 W (→ 12 | 34 | 57), fan 20000 RPM (→ 9999).

00 01 02 2b 00 00 00 22 39 00 00 00 00 00 00 00 00 00 00 00 00 63 63 00 00 14 1a 0b 14 0d 2d 1e
05 00 0c 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00

Vector 5 – packet from the original software

From the second capture of the original software on 2026-11-20 (t = 0.161 s): CPU usage 6.20 %, “voltage” 1.55 V, RAM 27.64 %, Friday 2026-11-20 03:14:53, byte 37 = 5. The reference implementation produces it byte-identically from these values (byte 37 set by hand) – see unit test Reproduces_packet_of_original_software_except_byte_37.

00 01 02 00 00 00 06 00 00 00 00 01 37 00 00 00 00 00 00 00 00 00 00 00 00 14 1a 0b 14 03 0e 35
05 1b 00 00 00 05 00 00 00 00 00 00 00 00 00 00 14 00 40 00 00 00 00 00 00 00 00 00 00 00 00 00

As hidraw write

Vector 1 as passed to write(fd, buf, 65) – with report ID 00 in front:

00 | 00 01 02 2a 32 00 25 17 2d 2d 43 00 00 … (65 bytes in total)

Reference pseudocode

function encode_update(f, now) -> byte[64]:
    p = zeros(64)
    p[1] = 0x01; p[2] = 0x02
    fixed2(p, 3, 4,   to_unit(f.cpu_temp, f.unit), max_whole=255)
    p[5]  = f.unit                      # 0 = °C, 1 = °F
    fixed2(p, 6, 48,  f.cpu_usage, max_whole=255)
    power(p, 34, 7, 8, f.cpu_power)
    base100(p, 9, 10, f.cpu_clock)
    fixed2(p, 11, 12, f.cpu_voltage, max_whole=255)
    fixed2(p, 13, 14, to_unit(f.gpu_temp, f.unit), max_whole=255)
    p[15] = f.unit
    p[16] = int_clamp(f.gpu_usage, 255)
    power(p, 35, 17, 18, f.gpu_power)
    base100(p, 19, 20, f.gpu_clock)
    base100(p, 21, 22, f.fan_rpm)
    base100(p, 23, 24, f.pump_rpm)
    p[25] = now.year div 100;  p[26] = now.year mod 100
    p[27] = now.month; p[28] = now.day
    p[29] = now.hour;  p[30] = now.minute; p[31] = now.second
    p[32] = now.weekday                 # Sunday = 0
    fixed2(p, 33, 50, f.ram_usage, max_whole=255)
    return p

sanitize(v)        = (v is finite and v > 0) ? v : 0
int_clamp(v, m)    = min(floor(sanitize(v)), m)
fixed2(p,a,b,v,m): c = min(round_half_away(sanitize(v)*100), m*100+99); p[a]=c div 100; p[b]=c mod 100
base100(p,a,b,v):  n = min(floor(sanitize(v)), 9999);                 p[a]=n div 100; p[b]=n mod 100
power(p,h,l,f,v):  c = min(round_half_away(sanitize(v)*100), 999999); w=c div 100
                   p[h]=w div 100; p[l]=w mod 100; p[f]=c mod 100