[pat64]

Couple of observations here;

- RISC-V has miles to go, I worry that with the current pace of ARM it may never catch up. - The assertion you can’t buy UMPCs in 2024 is patently false. GPD and a whole host of manufacturers make em. - A non mobile optimised blog in this day and age? That was a chore to read on the go.

[brucehoult]

> RISC-V has miles to go, I worry that with the current pace of ARM it may never catch up

That's a weird assertion. This machine is basically competitive with a Pi 4 (certainly miles ahead of Pi 3, both in CPU power and in having 8 or 16 GB RAM vs 0.5 or 1 GB).

The initial version of the RISC-V spec was ratified in July 2019, and the C910 core in this chip was announced in the same month. The TH1520 SoC hit the market in June 2023 (my board was delivered that month), four years after the CPU core was announced, and also four years after the Pi 4.

Note that the Arm A72 cores in the Pi 4 were announced in February 2015, so it was 4 1/2 years from core to Pi 4 using it, slightly longer than the THead C910 core announcement to Lichee Pi 4A shipping.

The Arm A53 also took about four years from announcement to the Pi 3 and Odroid C2. And the Arm A76 took four years from announcement to the RK3588 and Rock 5 shipping.

This is just the industry.

If the SG2380 machine(s) really come out this year then they will be faster than the A76 RK3588 boards and Pi 5 and around 2 to 2.5 years behind. But their P670 cores are equivalent to Arm A78, which is not available on an SBC -- and there are 16 of them, which is also not available in any cheap Arm-based SBC.

That's halving the gap.

SiFive's fastest core, the P870, is around Arm Cortex-X3 performance, and announced just 16 months later (October 2023 vs June 2022).

[G3rn0ti]

> RISC-V has miles to go, I worry that with the current pace of ARM it may never catch up.

Only assuming development progress happens linearly. In fact, technical progress goes exponentially. It took years to even get to the point to have RISC-V SoCs and now we are seeing the first laptop products. The pace will accelerate, especially with the promise of RISC-V being a cheaper platform to build hardware on than ARM and X86.

[sapiogram]

> In fact, technical progress goes exponentially.

Their point was that ARM is also progressing exponentially.

[G3rn0ti]

This line of arguing would imply that an established company selling established products would not need to fear any competition as long as it can keep the pace of its innovation cycles. While this line of reasoning is not wrong it completely ignores the possibility of "disruptive innovation" to use a buzz word here. I honestly think RISC-V could be a shooting star that will come to many as a surprise. And as long as RISC-V lags in performance behind ARM it could still compete successfully in price due to free licensing of RISC-V versus expensive licensing of ARM64. This has been AMD's strategy for many years (before they had better integrated GPUs compared to Intel). And think of how IBM completely lost its PC business to cheap competition.

[rwmj]

But only a few companies (two? Arm and Apple?) develop new Arm designs, whereas anyone is free to develop new RISC-V designs.

[bigfishrunning]

Most of those new Risc-V designs Are on the extreme low end; not a bad thing, but there's not a ton of high performance Risc-V stuff happening.

It's more likely to fill in the ESP32 niche then the ARM one

[brucehoult]

> but there's not a ton of high performance Risc-V stuff happening.

Uhhh ... yes, there is.

Several companies designing very high performance RISC-V were founded in 2021 or so. Most of them have top designers fresh from designing the latest Apple, Intel, AMD chips and they are aiming to match current x86 and Apple performance.

You'll see their products in the market around 2026.