How Important is RISK V to the Semi Ecosystem?

[Arthur Hanson]

Not being in chip design, I would appreciate any thoughts or comments on the impact of RISK V on the semi ecosystem and the impact it may or may not have on the semi ecosystem, THANKS

 

[Tanj]

It is a clean design. Not revolutionary, but it comes without baggage. ARM and X64 have baggage. It is extensible: at the low end you might compare it to Tensilica, a simple core which you can extend with opcodes of your own (look at Codasip). But it is a business model without baggage too. Anyone can design and prove a core, then sell it. You still need to buy IP - for a proven core, bus, IO devices, memory channels, etc. but the choice of folks is not those limited to a franchise with some architecture owner standing in the shadows.

The ecosystem is a business. The real question is does that business model have a decisive, multiplicative advantage, or is it just a slight step? Compare the CISC vs. RISC battle of 30 years ago. RISC had a step advantage, not a multiplicative advantage. That advantage was never eliminated, but it also was never enough to win the business model.

So what do you think? Is the business model a big enough difference for RV to win against ARM or X64? The ecosystem is hundreds of players. The RV architecture is a starting point, not a finished ecosystem.

Well I overlooked something. ARM (back then, a true RISC) did win mobile, a whole new market with no real incumbent, and they were cheap licenses too. RV might do that, quite likely in IoT. But then ARM took decades to build out from mobile to challenge in any other space. Will RV simply become IoT but in a similar way have a very slow influence wherever there are incumbents?

 

[Fred Stein]

Thanks Tanj. I'd love to hear more. And at a higher level. It appears we again have a renaissance in silicon design with so many companies designing their own chips and with VCs funding startups. And the ecosystem of EDA and foundries augmenting this, along with many open fora. SoC adds to this movement. Still Apple has two advantages the multiplicative advantage and control of both the chip design and the software. Will Apple, with TSMC's help, maintain their advantage? Or rather for how long?

 

[horace]

Whether RISC-V can unseat the ARM and x64's stronghold in mobile and PC/server will be a long up hill battle.
However RISC-V pretty much killed the market for all other smaller ISA. MIPS is certified dead already.
Cadence and Synopsys have their own ISA cores, which still have momentum from the past design wins.
It would be a lot harder to sell them to a brand new design.

 

[Daniel Nenni]

Personally I think it is critical. The more we go to open systems the more design starts we will see and the healthier the semiconductor ecosystem will be, absolutely.

 

[semiman]

It may become important, if it finds workloads which 'only RISC-V' can achieve. It will not be replacing ARM from the mobile market, or x86 from the server market. Designing new compilers to frameworks just to work with new ISA is pretty much overkill(and it will not work) for most customers.

ARM won mobile because Intel wasn't able to provide a low-power + small-size chip which could allow use of Netscape(don't need excel, powerpoints...etc) in smartphones. The advantage of 'Netscape-in-my-hand' was so massive that people decided to create their own ecosystems based on a new ARM land(And there were legacy phone vendors and fablesses with small ARM chips as well).
So RV will become important in IP, thus semi business if it finds a new continent. But will CPU uarch and ISA be really important in the future? not sure. Real competition of RV is not ARM, but it's IT tech giants who can rely on their in-house designed accelerators. Whatever ISA, or designs used, CPU cannot really win against accelerators when it comes to efficiency.

Of course, RV is winning in some embedded markets but that's because embedded software eco is separated. For example, Western digital's HDD controller firmware doesn't really need tons of software ecosystem to make it work. But in the mobile market, you need JIT compilers, optimized languages, codes from other people in github, operating systems who works well with AP, drivers which works well with that AP...etc etc.

 

[Fred Stein]

It may become important, if it finds workloads which 'only RISC-V' can achieve. It will not be replacing ARM from the mobile market, or x86 from the server market. Designing new compilers to frameworks just to work with new ISA is pretty much overkill(and it will not work) for most customers.

ARM won mobile because Intel wasn't able to provide a low-power + small-size chip which could allow use of Netscape(don't need excel, powerpoints...etc) in smartphones. The advantage of 'Netscape-in-my-hand' was so massive that people decided to create their own ecosystems based on a new ARM land(And there were legacy phone vendors and fablesses with small ARM chips as well).
So RV will become important in IP, thus semi business if it finds a new continent. But will CPU uarch and ISA be really important in the future? not sure. Real competition of RV is not ARM, but it's IT tech giants who can rely on their in-house designed accelerators. Whatever ISA, or designs used, CPU cannot really win against accelerators when it comes to efficiency.

Of course, RV is winning in some embedded markets but that's because embedded software eco is separated. For example, Western digital's HDD controller firmware doesn't really need tons of software ecosystem to make it work. But in the mobile market, you need JIT compilers, optimized languages, codes from other people in github, operating systems who works well with AP, drivers which works well with that AP...etc etc.

Thanks. I'm curious RV and Apple silicon. Apple has gone up, down and sideways and in M-series, wearables, and wearables - all in volume. Might Apple use RV as a component on SoCs? May RV play a role for Apple in data access acceleration? Oops I meant as in M-series, wearables and smart phones

 

[Tanj]

ARM won mobile because Intel wasn't able to provide a low-power + small-size chip which could allow use of Netscape

ARM won mobile long before browsing on a phone became important. They won in the 1990s and it was because ARM was IP - it could be integrated into customer chips. Customers like Nokia who leaned in agressively to make their own SOCs for lowest cost with high functionality. Which in those days meant talking to people and running a personal organizer (a Palm Pilot was like a Nokia phone without the phone, which is why it never amounted to much). Browsing on a phone was a niche activity until the iPhone.

The ARM IP was accessible to engineering teams around the world, with Linux the default lab tool for new chip bringup. Almost all feature-phones not made by Nokia were Linux by 2006, but that was almost never visible to a user. Android took off so fast because it leveraged the ubiquitous base and slapped a good, branded UX on top. Windows phones always were on the back foot because they had delays in chip bring-up, which only a few partners would even invest in doing. Apple was OK because their kernel has a lot of similarity to Linux, the bring up in 2006 was easy, and then soon they did their own SOC anyway.