Nvidia has approached Softbank and is considering buying ARM Holdings

[uzzi38]

Will update with articles as they come out (they haven't yet but news is public). For now just a couple of Tweets:

https://twitter.com/x/status/1286003982079533056

https://twitter.com/x/status/1286004990327132161

This is a real significant move for Nvidia, having a huge role in ARM core designs could have major implementations down the line. Most importantly, it secures them a platform and ecosystem down the line.

Oh and I guess Nvidia becomes the defacto standard for GPU IP for mobile instead of Mali. That too.

EDIT: Bloomberg article here: https://www.bloomberg.com/news/arti...ompany-arm-is-said-to-attract-nvidia-interest

 

[Doug S]

If another company comes up with compelling chip designs that are competitive, will ARM still be worth $40 billion?

It takes years to build the sort of ecosystem ARM has, so it would take a lot more than someone doing a competitive RISC-V design and releasing it. If doesn't include a modem, that makes it a lot less attractive to the Android market which has never bought a modem separately (theoretically they could, but Qualcomm would make sure the price they charge for that option eats up any money an OEM might save over buying Qualcomm's SoCs)

 

[DrMrLordX]

It takes years to build the sort of ecosystem ARM has, so it would take a lot more than someone doing a competitive RISC-V design and releasing it. If doesn't include a modem, that makes it a lot less attractive to the Android market which has never bought a modem separately (theoretically they could, but Qualcomm would make sure the price they charge for that option eats up any money an OEM might save over buying Qualcomm's SoCs)

Or Qualcomm will release their own RISC-V SoCs once they realize that Softbank is using them as a piggy bank. Or trying to anyway.

 

[eek2121]

It takes years to build the sort of ecosystem ARM has, so it would take a lot more than someone doing a competitive RISC-V design and releasing it. If doesn't include a modem, that makes it a lot less attractive to the Android market which has never bought a modem separately (theoretically they could, but Qualcomm would make sure the price they charge for that option eats up any money an OEM might save over buying Qualcomm's SoCs)

Google is making RISC-V a first class citizen. Will find the source when I can. Android on RISC-V is being worked on by Google engineers, fyi.

 

[scannall]

Google is making RISC-V a first class citizen. Will find the source when I can. Android on RISC-V is being worked on by Google engineers, fyi.

You know, until they cancel it 2 or 3 years out.

 

[LightningZ71]

Who would license a modem cheap enough to make the money work?
(Editor went crazy)

 

[Doug S]

You know, until they cancel it 2 or 3 years out.

Porting Android to RISC-V is just a tiny portion of the problem. There's a ton of infrastructure required for OEMs to do the bringup of new designs, and all the stuff Google doesn't provide like the firmware support like uboot, drivers and so forth. Then you have to convince developers it is worth the trouble to port and support popular apps like Facebook or the latest games to RISC-V (since pretty much all Android apps use native code these days)

So even if Google doesn't abandon this project like they do everything else, that's just a first step that might make OEMs begin to consider producing a RISC-V phone. They certainly wouldn't risk that on a high end device, this would be on some sort of $30 very low end Android phone sold by some no name Indian company or Chinese company for the African market. Which makes convincing developers to port even harder, because why should they bother porting something for customers they can't monetize enough to pay for it?

 

[moinmoin]

Google is making RISC-V a first class citizen. Will find the source when I can. Android on RISC-V is being worked on by Google engineers, fyi.

This was posted before:

Google announces official Android support for RISC-V

Google's keynote at the RISC-V Summit promises official, polished support.

arstechnica.com

Porting Android to RISC-V is just a tiny portion of the problem.

While it's a big project I think you overstate Google's position somewhat. Kernel level support needs to exist in the still used Linux kernel first, that's essentially independent from Android. Hardware SoC support including firmware is something manufacturers need to come up, that's essentially independent from Android. Android is already prepared to run on x86 so supporting another ISA would extend that area of support. Google gave its OK to port and extend Android and make RISC-V a first tier citizen, so this will lead to more activity on respective AOSP efforts regardless of Google degree of involvement like already happening at:

aosp-riscv

This project is initiated by PLCT lab, the purpose is to port Android to RISC-V platform. - aosp-riscv

github.com

So while a lot of parts need to come together and match before we see a RISC-V Android phone on the market, to me you seem to focus too much on the commercial angle.

 

[Nothingness]

Wasn't MIPS a first class Android citizen? And yes there even was real Android MIPS devices. Where are the RISC-V ones?

 

[soresu]

Wasn't MIPS a first class Android citizen? And yes there even was real Android MIPS devices. Where are the RISC-V ones?

It was once, but back then MIPS was still pretty big and ARM hadn't yet covered the world in its dominance of market sectors previously owned by MIPS and other players.

RISC-V is slowly creeping onto the scene, but it is definitely coming.

I wouldn't be surprised to find AMD replacing the smol Cortex A cores in their CPUs used for trusted environment with a RISC V core in the future.

 

[Doug S]

So while a lot of parts need to come together and match before we see a RISC-V Android phone on the market, to me you seem to focus too much on the commercial angle.

The commercial angle is what matters because until OEMs sell RISC-V Androids so there are actual phones reaching the market, no one is going to work on AOSP Android for RISC-V.

 

[moinmoin]

no one is going to work on AOSP Android for RISC-V.

As I wrote and linked that's already happening regardless.

 

[Mopetar]

Wasn't MIPS a first class Android citizen? And yes there even was real Android MIPS devices. Where are the RISC-V ones?

I honestly can't recall an Android phone with a MIPS-based SoC. Maybe there was one in some Asian country, but I can't recall anything that would have come to the US.

MIPS was (and still is) a lot more popular in the embedded space than it ever was in mobile devices. Though MIPS has been in some standout consumer products like the Nintendo 64 which will always give it a special place in my heart.

Frankly anyone can port Android to their hardware if they wanted to since it's open enough to allow it. I doubt that if it did happen that it was ever for mobile phones but probably for other embedded devices that needed an OS and were designed to replace WINCE (Windows CE) hardware.

 

[gdansk]

The commercial angle is what matters because until OEMs sell RISC-V Androids so there are actual phones reaching the market, no one is going to work on AOSP Android for RISC-V.

Doug I think it is less about phones and more about Android-powered TVs, automobile infotainment systems, single board computers, etc. The embedded space, basically.
These are absolutely margin-driven areas and RISC V cores are a one time cost for really big manufacturers who can design their own core unlike ARM with its continual licensing fees.

 

[Mopetar]

Most companies probably won't design their own RISC-V cores, but will license a design from someone else unless they already have a team of engineers doing some kind of IC design work.

The type of people who can develop a chip, even just a simple implementation of RISC-V are rare and therefore expensive. It's a big investment just to create a team like that since it won't yield any results for several years.

It's generally cheaper to buy some COTS parts from whatever vendor can supply them at the lowest cost. This is especially true for smaller companies. But it's not really a problem since the lack of licensing fees for RISC-V mean that there's always going to be an opportunity for companies to break into the market and sell those commodity chips.

 

[itsmydamnation]

its funny, I feel the demise of x86 by ARM is well overstated and I feel the demise of ARM by RISC-V is well overstated ......lol

 

[gdansk]

its funny, I feel the demise of x86 by ARM is well overstated and I feel the demise of ARM by RISC-V is well overstated ......lol

It took a 1000 years for Rome to truly fall. Demise is a long process. But I'm sold on the long term economic advantages of RISC-V outweighing its slight inferiority to ARM. At least in the markets where lower bill of materials > software compatibility.

 

[gdansk]

Most companies probably won't design their own RISC-V cores, but will license a design from someone else unless they already have a team of engineers doing some kind of IC design work.

The type of people who can develop a chip, even just a simple implementation of RISC-V are rare and therefore expensive. It's a big investment just to create a team like that since it won't yield any results for several years.

It's generally cheaper to buy some COTS parts from whatever vendor can supply them at the lowest cost. This is especially true for smaller companies. But it's not really a problem since the lack of licensing fees for RISC-V mean that there's always going to be an opportunity for companies to break into the market and sell those commodity chips.

That's the neat bit. Already there are open source cores. Some are derived from commercial products. These are good starting points and in some cases all you'd need. ARM can't compete with free especially when you'll have to do the integration work in either case. For companies like Nvidia and Western Digital this is appealing. Yes, the company trying to buy ARM had already chose RISC-V for the tiny cores embedded on their GPUs.

If, 30 years from now, ARM is still more common than RISC-V in the embedded space it'll be the result of ARM somehow changing their business model.
But I don't think RISC-V will be compelling in phones or PCs.

 

[eek2121]

This was posted before:

Google announces official Android support for RISC-V

Google's keynote at the RISC-V Summit promises official, polished support.

arstechnica.com

While it's a big project I think you overstate Google's position somewhat. Kernel level support needs to exist in the still used Linux kernel first, that's essentially independent from Android. Hardware SoC support including firmware is something manufacturers need to come up, that's essentially independent from Android. Android is already prepared to run on x86 so supporting another ISA would extend that area of support. Google gave its OK to port and extend Android and make RISC-V a first tier citizen, so this will lead to more activity on respective AOSP efforts regardless of Google degree of involvement like already happening at:

aosp-riscv

This project is initiated by PLCT lab, the purpose is to port Android to RISC-V platform. - aosp-riscv

github.com

So while a lot of parts need to come together and match before we see a RISC-V Android phone on the market, to me you seem to focus too much on the commercial angle.

Kernel level support is very close to landing. I have a RISC-V board and from what I understand it will happen within the next few months.

I am not trying to derail this thread, of course, however if NVIDIA is smart, they will be building a RISC-V chip, that way when the NVIDIA/ARM deal finally goes south, they can push out a performant chip.

 

[gdansk]

Kernel level support is very close to landing. I have a RISC-V board and from what I understand it will happen within the next few months.

I am not trying to derail this thread, of course, however if NVIDIA is smart, they will be building a RISC-V chip, that way when the NVIDIA/ARM deal finally goes south, they can push out a performant chip.

For high performance they'll stick to using ARM because end user code will be running on it. That's worth the licensing fees.

But it is interesting Nvidia already includes NV-RISCV cores (in the GSP) in their GPUs. No end user code, it's basically invisible.
You can verify this by inspecting the GSP firmware distributed with Linux drivers:

% file gsp.bin
gsp.bin: ELF 64-bit LSB executable, UCB RISC-V, soft-float ABI, version 1 (SYSV), statically linked, stripped

 

[Ajay]

As I wrote and linked that's already happening regardless.

Yeah, I but hardly anyone is working on that project. Looks like 5 people from China.

 

[gdansk]

Yeah, I but hardly anyone is working on that project. Looks like 5 people from China.

You'd be surprised how many projects are 5 people from China.
Or fewer. But AOSP is pretty big, so we'll see. The big hurdles will be the compiler backends for v8 and ART. But the work is underway by a group that ported a JDK and Java runtime to RISC-V previously so that is pretty promising. And some Huawei employees working on v8 RISC-V for some reason. Maybe a Chinese state initiative

 

[NostaSeronx]

Wasn't MIPS a first class Android citizen? And yes there even was real Android MIPS devices. Where are the RISC-V ones?

MIPS had a lower barrier of entry to Android devices back in Android 1~5.1.1 era. The RISC-V needs cores that are comparable to latest Cortex-Xx/Cortex-A7xx. As the premium to ultra-premium is what pushes the current mobility market. No one is going to buy a JH7110 in a Phone/Phablet or a TH1520 in a Phone/Phablet.

The other issue is the community driving RISC-V on Android. With the above doing it in their own garden via OpenHarmony/HarmonyOS [Huawei's branch of Android]. 80% of RISC-V foundation force has ditched standard Android and gone for OpenHarmony instead.

Dubhe via JH7110 successor by StarFive and C920 via TH1520 successor by T-Head. Will of course be exclusive to their domestic market.

I believe however Alibaba's T-head has already achieved monopoly status.
ROMA Laptop dev platform.
HH-SCDAYU800 Tablet dev platform.

It isn't far off to expect a smartphone dev platform sometime this year, since they already have two of three mobility rings. NetEase is already "producing" an education software based off T-head's RISC-V hardware. In turn, allowing T-head to dominate the domestic EDU/IND/GOV markets.

 

[Nothingness]

I am not trying to derail this thread, of course, however if NVIDIA is smart, they will be building a RISC-V chip, that way when the NVIDIA/ARM deal finally goes south, they can push out a performant chip.

You're just one year late about the deal: https://aibusiness.com/verticals/nv...nated-over-significant-regulatory-challenges-

 

[Nothingness]

But it is interesting Nvidia already includes NV-RISCV cores (in the GSP) in their GPUs. No end user code, it's basically invisible.

RISC-V definitely is successful in embedded world. Just as Arm was 20 years ago. And even then at that time they already owned the phone market. Looking at how long it took Arm to succeed the way they did, I don't expect R-V to displace Arm in the non embedded world before 10 years. But that doesn't mean I'm not looking into R-V architecture for a potential next job

 

[lobz]

Thanks for the much needed clarification, I edited my post accordingly.

Nah-ah... sorry, but that just won't do... mistakes like this really don't fly in this line of work, or more like calling, that is to be a part of the Anandtech Forums.

You have admittedly and quite publicly made an erroneous remark, so you shall be labelled and forever remembered as such, the one who can't even be always perfect, let alone immaculately spotless, which is basically the basic pre-requisite, kind of the lowest bar to clear in order to be eligible to participate in these debates here, which represent previously undocumented stages of ascension, and also levels of intellectual elevation previously unheard of.

If you think about it for a moment, probably even you could reach the necessary conclusions to make. You simply do not belong here. Our spiritual leader, juanrga tEh_GReaT, sends his utmost unhappy regards concerning your imminent departure.

Nevertheless, I sincerely hope you'll still have fun following our superior dialogues sometimes in the future, even if you'd obviously only understand the underlying sentiments and not the actual particularities - because, of course, you can't honestly expect us to tone down our heights of sophistication as we casually engage in our everyday 'tech banter for simpletons' style of conversation - not even for you 😔 sorry 😔