It’s a bit of a shame that this machine won’t see a lot of use, but I anticipate that will change as RISC-V adoption continues to grow.
From my experience with these SoCs they really need a big leap in terms of quality and power to performance ratio. Even at rest these things tend to draw a lot of power.
The other issue is that they keep combining the RISC-V cores with these Imagination video chips which require a separate firmware blob for both GPU acceleration and video processing and that's not great for innovating on other operating systems other than Linux. And I need to mention that firmware is not very capable. I wish they could find some way to open up access to the graphics so everyone can take a stab at improving performance and capabilities of the hardware.
From my experience with these SoCs they really need a big leap in terms of quality and power to performance ratio.
100%, and I'm not convinced we'll see it.
Intel and AMD are as good as they are due to competitive pressure from each other; they've been competing on performance, and to some degree performance/watt, for decades. We saw what happened when Intel stopped executing during the 2010s; Apple jumped ship to their own IP and AMD ate what was left of their lunch. Having the best performance and perf/watt has been rewarded in the massive laptop, server, desktop and game console markets.
ARM chips are also amazing due to competition between Apple, Qualcomm, Rockchip and Samsung, partially in collaboration with and partially competing with ARM Holdings. Having the best ARM chip has been rewarded in the massive phone market.
Both x86 and ARM are entrenched in their respective markets due to inertia. And ARM only emerged as an alternative to x86 in the phone market because neither Intel nor AMD were making very good phone chips, nobody else could make x86 chips, and ARM was a viable alternative with a licensing model which allowed phone manufacturers to make their own chips.
I don't really see a path forwards for RISC-V to replace either of them. Its main advantage compared to ARM is that there's no license fee and restrictions for the architecture itself, but I don't see that making a difference for the relatively high-margin chips which can comfortably run Linux. It's a win in the low-margin microcontroller space, which is the space where RISC-V is already hugely successful.
I think RISC-V is a cool architecture and I've played around with it a bit, including making my own toy implementation in Logisim, and I would certainly like it if a permissively licensed non-proprietary ISA was used in more areas than just microcontrollers. But I don't understand how market forces are supposed to take us there.
I mean there's one potential path: China. If the West keeps being an unreliable partner and keeps using the ARM ISA license as a weapon in trade wars, China's domestic chip design industry will naturally migrate away from ARM and towards something they don't need the West's permission to use. This might be LoongArch, or it might be RISC-V. It's probably not a coincidence that Alibaba and Milk-V, both Chinese companies, are arguably leading in RISC-V.
I'll happily remind people that ultra fast CPU cores aren't always necessary. Inexpensive, easy to use cores that get the job done are fine for many, many tasks. For high performance, the options are there, and for everything else, RISC-V is worth considering.
Yes, I agree that highest performance is not always critical. But the market for medium performance is also very much contended, I think, with tons of previous-generation CPUs around, both in the x86 world and the ARM world, that fit the different market needs.
valpackett | 16 hours ago
Hilarious duckduckgo snippet from imaginationtech.com:
…I guess it at least advertises "64 FP32 FLOPs/Clock" which is double what they had in the GE8300 (featured in the Mediatek MT8167)
proctrap | 6 hours ago
Any reasons for this ?
manfred | 9 hours ago
From my experience with these SoCs they really need a big leap in terms of quality and power to performance ratio. Even at rest these things tend to draw a lot of power.
The other issue is that they keep combining the RISC-V cores with these Imagination video chips which require a separate firmware blob for both GPU acceleration and video processing and that's not great for innovating on other operating systems other than Linux. And I need to mention that firmware is not very capable. I wish they could find some way to open up access to the graphics so everyone can take a stab at improving performance and capabilities of the hardware.
mort | 5 hours ago
100%, and I'm not convinced we'll see it.
Intel and AMD are as good as they are due to competitive pressure from each other; they've been competing on performance, and to some degree performance/watt, for decades. We saw what happened when Intel stopped executing during the 2010s; Apple jumped ship to their own IP and AMD ate what was left of their lunch. Having the best performance and perf/watt has been rewarded in the massive laptop, server, desktop and game console markets.
ARM chips are also amazing due to competition between Apple, Qualcomm, Rockchip and Samsung, partially in collaboration with and partially competing with ARM Holdings. Having the best ARM chip has been rewarded in the massive phone market.
Both x86 and ARM are entrenched in their respective markets due to inertia. And ARM only emerged as an alternative to x86 in the phone market because neither Intel nor AMD were making very good phone chips, nobody else could make x86 chips, and ARM was a viable alternative with a licensing model which allowed phone manufacturers to make their own chips.
I don't really see a path forwards for RISC-V to replace either of them. Its main advantage compared to ARM is that there's no license fee and restrictions for the architecture itself, but I don't see that making a difference for the relatively high-margin chips which can comfortably run Linux. It's a win in the low-margin microcontroller space, which is the space where RISC-V is already hugely successful.
I think RISC-V is a cool architecture and I've played around with it a bit, including making my own toy implementation in Logisim, and I would certainly like it if a permissively licensed non-proprietary ISA was used in more areas than just microcontrollers. But I don't understand how market forces are supposed to take us there.
I mean there's one potential path: China. If the West keeps being an unreliable partner and keeps using the ARM ISA license as a weapon in trade wars, China's domestic chip design industry will naturally migrate away from ARM and towards something they don't need the West's permission to use. This might be LoongArch, or it might be RISC-V. It's probably not a coincidence that Alibaba and Milk-V, both Chinese companies, are arguably leading in RISC-V.
manfred | 3 hours ago
And the European Chips Act spending some peanuts out of FOMO.
johnklos | 2 hours ago
I'll happily remind people that ultra fast CPU cores aren't always necessary. Inexpensive, easy to use cores that get the job done are fine for many, many tasks. For high performance, the options are there, and for everything else, RISC-V is worth considering.
felixs | 2 hours ago
Yes, I agree that highest performance is not always critical. But the market for medium performance is also very much contended, I think, with tons of previous-generation CPUs around, both in the x86 world and the ARM world, that fit the different market needs.