Or trying to catch up to Apple?
Ascalon is a RISC-V microarchitecture expected to be released in 2024, with similar performance to projected AMD Zen5 (also 2024), but lower power consumption.
If ARM is still trying to catch up to Apple, they'll be leapfrogged by RISC-V, as early as that.
ARM has very successful architectures and cores already in the market. The current iteration of the Snapdragon 8cx does a pretty good job in general purpose computing scenarios [0].
There are no RISC-V cores in the market to compete with it.
Vendors can and will promise breakthroughs every now and then, that is just how marketing works. It happened before with things like Graviton, and of course Tenstorrent is doing it now.
[0] https://www.notebookcheck.net/SD-8cx-Gen-3-vs-M1-vs-M2_14600...
Depends on what "it" is.
For the whole Cortex-M line and A53/A55, SiFive has had designs with better power-performance-area for years now.
For the higher performance A70 line, SiFive has better designs competing up to A76.
There's several other vendors, with their own competitive RISC-V cores.
>This is pure speculation.
Not when it's based in announcements and talks by the relevant companies.
>It happened before with things like Graviton, and of course Tenstorrent is doing it now
Wait, why do you think Tenstorrent will not deliver? Do you actually have anything?
I have no reason to believe that strong teams that have successfully delivered competitive microarchitectures in a range of companies won't do it with RISC-V now that they're with Tenstorrent, MIPS, Rivos, Ventana, Qualcomm or any others.
Ventana's Veyron and Tenstorrent Ascalon are imminent. The other companies are less public about what they're doing.
Any one of these efforts could be delayed or underdeliver. All of them? That seems highly unlikely.
RISC-V is inevitable.
Are RISC-V “better designed” than x86 processors, even when the latter are overwhelmingly more popular and faster than the former?
> Not when it's based in announcements and talks by the relevant companies.
Again, this is exactly what marketing is. I cannot begin to count how many companies formed by wildly successful leaders of the industry, have failed to deliver.
> Any one of these efforts could be delayed or underdeliver. All of them? That seems highly unlikely.
Core design and implementation takes years to complete.
The supposition is that a handful of relatively new companies will take the processor market by storm in a year or two, with less than a decade of work behind them, and a tenth of the budget, compared to companies like AMD or Intel.
That is speculation if I ever seen any.
I mentioned it: power-performance-area.
In the case of SiFive, they seem to like comparing each of their designs with a specific one from ARM. They typically manage to perform better, use less area and consume less power, at the same time, relative to the ARM design they picked.
>Core design and implementation takes years to complete.
These years already happened, these are several years underway. What follows is the products that we are about to see.
>and a tenth of the budget
There's several billions already invested. Source: RISC-V Foundation's main talks at RISC-V Summit last December.
In 2021, the theme was acceleration. In 2022, it was momentum, because it is already there. RISC-V is huge.
RISC-V is inevitable.
>The supposition is that a handful of relatively new companies will take the processor market by storm in a year or two, with less than a decade of work behind them, and a tenth of the budget, compared to companies like AMD or Intel.
Exciting times to live in.
I expect AMD and Intel to survive, and find a way to thrive. They can e.g. provide RISC-V CPUs with excellent x86 acceleration, making them an easy choice in markets that depend on x86 software still.
I do not have faith ARM will be able to adapt, due to their management's recent behavior. They are dinosaurs. Won't know what hit them.
It’s like saying that the better architecture is the one with available CPUs manufactured on the smaller node and with the higher CoreMark score.
> Won’t know what hit them
This is still a extremely naive take.
Qualcomm and Apple are not going to ditch ARM any time soon, and there is no market where RISC-V currently has a meaningful share. Hell, there are more x86 SBC out there than RISC-V by a double digit factor.
What I’m saying is that RISC-V is a hypothetical contender until manufacturers adopt it.
Of course, I will eat my hat if we get competitive RISC-V cores across the market in less than five years, but I’m confident that will not be the case.
Just two notes:
- Qualcomm is a top-tier RISC-V member and has an ugly lawsuit going on with ARM about Nuvia's IP. They did multiple presentations in RISC-V Summit where they were very clear about their intent to move to RISC-V.
- Apple has been hiring RISC-V experts for a while (now years). Keywords included performance and optimization.
So do not count on either to remain faithful to ARM.
>What I’m saying is that RISC-V is a hypothetical contender until manufacturers adopt it.
As per RISC-V's foundation presentation in RISC-V Summit December 2022, RISC-V had at that point already shipped 10 billion cores.
That's some serious manufacturing, and we're talking about what already happened. It's safe to assume much more of that is ongoing.
It’s any company’s duty to explore alternatives. Intel used to make ARM CPUs, Mediatek makes MIPS chips, etc.
It does not mean that it is possible to completely reverse course in less than a decade.
Qualcomm is definitely not going to dramatically reverse its core business direction to go full RISC-V in less than a decade. It makes no sense, especially after dumping $1.5B in Nuvia.
ARM’s case against Qualcomm is a licensing one. So you suggest that Qualcomm may be ditching ARM because of an argument about licensing ARM, which would make no sense whatsoever. It would be akin to Intel ditching AMD64 and adopting some exotic architecture, because of a disagreement about cross-licensing.
Apple just switched to ARM, of which they have a perpetual license. They are also getting into display production and modem design. Their aim is to control the production line to the last bit, they do not need to be compatible with anything else, so why would they ditch ARM any time soon? It would make no financial sense.
> As per RISC-V's foundation presentation in RISC-V Summit December 2022, RISC-V had at that point already shipped 10 billion cores.
There were 7 billion ARM cores shipped in the last quarter of 2020 alone, many of them inside actual consumer products, not SBCs or niche platforms. Virtually all portable, IoT, and in-car devices run on ARM. ~14% of laptops run ARM and the share is growing, same goes for datacenters. And most importantly, software support for RISC-V is still in its infancy, with no large scale consumer products currently available.
This is to say that RISC-V has a very long road ahead, and it is not going to topple ARM dominance any time soon, simply because the economics of it make no sense.
Explain the RISC-V cores that already moved.
It's inexpensive. What a feat.
Can you point at your source for that? I am not interested in speculation.
Anyway, there is a good number of ESP32 microcontrollers using RISC-V cores, replacing the aging ARM, e.g. ESP32-C3, ESP32-P4.
Get off your bubble. SBCs are a tiny fraction of the CPU market.
But custom/semicustom is huge, and is where I expect most already shipped cores would be. Most of those cores would not be user visible.
Western Digital probably accounts for a lot of shipped cores, as they were early adopters and moved their HDDs controllers to RISC-V.
I’m pretty sure nothing is “inevitable”.