Selection guide7 min readReviewed 2026

SK6812 vs WS2812B vs HD108: SPI and DMX512

WS2812B and SK6812 use single-wire NZR data, while HD108 is a clocked SPI part. Compare frame rate, dimming, white channel and DMX512 architecture before choosing the pixel chip.

SK6812 vs WS2812B vs HD108: SPI and DMX512

Three pixel chips with similar 5050 footprints can behave very differently once the controller is chosen. For lighting engineers, the control interface usually decides whether a design works at scale, not the brightness figure on the first page of the datasheet.

What SPI Means for These Chips

WS2812B and SK6812 are not native SPI devices. Both use a single-wire NZR signal at roughly 800 kbps, five-volt logic and an 8-bit value per colour channel. A 24-bit frame drives RGB; SK6812 RGBW builds extend this to 32 bits for the dedicated white die. Their bit timing is fixed, with a data period near 1.25 µs and a reset low that the WS2812B datasheet puts above 50 µs, while SK6812 builds are commonly specified at 80 µs. Many controllers generate that waveform through an SPI peripheral, which is why the two are loosely described as SPI chips. That is a driver technique, not a clocked interface.

HD108 is a genuine two-wire SPI part. It carries a clock and a data line, 16 bits per channel for 65,536 levels, and a 64-bit frame per pixel. Its PWM runs at 27 to 28 kHz, against roughly 400 Hz for WS2812B and about 1.2 kHz for SK6812, both revision-dependent. The clock also lifts throughput: at 20 MHz, 1,024 HD108 pixels refresh near 305 fps. The same count runs at about 32 fps on WS2812B and 24 fps on SK6812 RGBW at 800 kbps. The cost is a second conductor, more EMI risk and no data redundancy.

DMX512 Sits at a Different Layer

DMX512 is not a chip protocol. It is an RS-485 differential link at 250 kbit/s, carrying 512 eight-bit channels per universe, with a full-universe refresh ceiling near 44 Hz and a practical limit of 32 unit loads per run. A universe addresses about 170 RGB pixels, so 1,024 pixels need six universes and a decoder architecture. The differential layer tolerates cable runs of hundreds of metres, which single-wire five-volt pixel data does not.

The naming trap is the single-wire DMX pixel chip, such as the TM512 or UCS512 families. These accept DMX-style serial data directly at the chip, at rates from 200 to 1,000 kbps, and each pixel carries its own address. Their physical layer is not RS-485, and they do not share the 512-channel universe budget in the same way. Treating them as standard DMX512 will mislead the wiring and controller plan.

Choosing on Project Constraints

Dense pixel mapping over short runs, with tight cost control, favours WS2812B; specify SK6812 when a real white die matters. HD108 earns its place where dimming curves, camera capture or high frame rates drive the brief. Long cable runs, console integration and zone-level control point to DMX512 with decoders, or to addressable DMX chips when per-pixel addressing is required.

Confirm the exact variant, reset timing and voltage-drop plan against the datasheet before the run is booked. We supply SK6812, WS2812B and HD108 chips and match strip, driver and controller in one specification; send a drawing for review.

Key takeaways

  • WS2812B and SK6812 are single-wire NZR parts; HD108 uses a clocked two-wire SPI interface.
  • DMX512 is an RS-485 link with 512 channels per universe, not a pixel-chip protocol.
  • Choose by run length, frame rate, white channel and console integration, then confirm the exact variant and timing against the datasheet.

Continue reading