RISC-V promises eightfold performance, but the catch is in the numbers

RISC-V processors have taken a theoretical leap in performance: the new SpacemiT K3 (2026) is eight times faster than the SiFive from 2021, according to their manufacturers. The figure sounds like a revolution, but it hides a marketing trick. These chips aim to rival the Raspberry Pi, although in practice, the ARM ecosystem remains faster, cheaper, and more accessible for the average user.

photorealistic technical illustration of two processor dies side by side on a workbench, left chip labelled SiFive 2021 with a single small glowing core, right chip labelled SpacemiT K3 2026 with eight stacked cores glowing brightly, a magnifying glass hovering over the right chip revealing hidden architectural shortcuts and simplified cache hierarchy, a performance graph display in background showing artificial benchmark scaling with exaggerated multiplier, an engineer shaking head while pointing at misleading 8x arrow, cinematic workshop lighting with blue and amber tones, ultra-detailed silicon wafer textures, macro photography style, engineering visualization

The trick of benchmark tests in open architecture 🧐

Companies selling RISC-V design the tests to pit their new chip against the slowest model from five years ago. Thus, an 8x jump seems epic, but it hides that the real evolution is modest. In standard benchmarks like SPEC, the SpacemiT K3 barely matches a Cortex-A72 from 2015, the same core as the Raspberry Pi 4. The improvement exists, but it is not the revolution they announce.

The Raspberry Pi laughs while RISC-V counts its cycles 😂

While RISC-V engineers celebrate their performance multiplication, the humble Raspberry Pi 5 continues to sell for 60 euros and run retro games without breaking a sweat. The SpacemiT K3, if it reaches the market, will cost more and have less software. It's like bragging about losing 8 kilos... but comparing yourself to your photo when you weighed 120. The merit exists, but the context deflates it.