Arm is Developing Its Own Advanced Chip (ft.com) 31
"Arm is developing its own chip," reports the Financial Times, "to showcase the capabilities of its designs, as the SoftBank-owned group seeks to attract new customers and fuel growth following a blockbuster IPO later this year."
The company will team up with manufacturing partners to develop the new chip, according to people briefed on the move who describe it as the most advanced chipmaking effort the Cambridge-headquartered group has ever embarked upon. The effort comes just as SoftBank seeks to drive up Arm's profits and attract investors to a planned listing on New York's Nasdaq exchange... The hope is that the prototype will allow it to demonstrate the power and capabilities of its designs to the wider market.
Arm has previously built some test chips with partners including Samsung and Taiwan Semiconductor Manufacturing Co, largely aimed at enabling software developers to gain familiarity with new products. However, multiple industry executives told the FT that its newest chip — on which it started work in the past six months — is "more advanced" than ever before. Arm has also formed a bigger team that will execute the effort and is targeting the product at chip manufacturers more than software developers, they said...
Rumblings about Arm's chipmaking moves have stoked fears in the semiconductor industry that if it makes a good enough chip, it could seek to sell it in the future and thereby become a competitor to some of its biggest customers, such as MediaTek or Qualcomm. People close to Arm insist there are no plans to sell or license the product and that it is only working on a prototype. Arm declined to comment.
The article cites "people briefed on the move" as saying that Arm plans to build prototype chips for laptops, mobile devices, and other electronics. "The team will also expand on Arm's existing efforts to enhance the performance and security of designs, as well as bolster developer access to its products. "
The article points out that the head of Arm's engineering team previously oversaw the development of Qualcomm's Snapdragon chip.
Arm has previously built some test chips with partners including Samsung and Taiwan Semiconductor Manufacturing Co, largely aimed at enabling software developers to gain familiarity with new products. However, multiple industry executives told the FT that its newest chip — on which it started work in the past six months — is "more advanced" than ever before. Arm has also formed a bigger team that will execute the effort and is targeting the product at chip manufacturers more than software developers, they said...
Rumblings about Arm's chipmaking moves have stoked fears in the semiconductor industry that if it makes a good enough chip, it could seek to sell it in the future and thereby become a competitor to some of its biggest customers, such as MediaTek or Qualcomm. People close to Arm insist there are no plans to sell or license the product and that it is only working on a prototype. Arm declined to comment.
The article cites "people briefed on the move" as saying that Arm plans to build prototype chips for laptops, mobile devices, and other electronics. "The team will also expand on Arm's existing efforts to enhance the performance and security of designs, as well as bolster developer access to its products. "
The article points out that the head of Arm's engineering team previously oversaw the development of Qualcomm's Snapdragon chip.
Can It Play Crysis? (Score:2)
If not then go home.
Re: (Score:2)
wrong (Score:1, Interesting)
Good news: 0.3 Ghz faster (Score:3)
why is ARM competing with its own Licensees? (Score:2)
why is ARM trying to compete with its own Licensees? surely those are its main business? it couldn't happen to have something to do with Softbank taking out a USD 11 billion mortgage against ARM's value when Softbank first bought it, putting a massive black hole debt against the company that nobody but really large companies like NVIDIA were prepared to take on, or would have if they hadn't run into Anti-Trust, and now ARM is desperate to make the company look good before it IPOs, would it?
Re: (Score:2)
why is ARM trying to compete with its own Licensees? surely those are its main business?
It's the Microsoft model.
Re: (Score:2)
RISC-V is a huge problem for ARM.
And Softbank? Oh, boy.
Re: (Score:2)
Re:why is ARM competing with its own Licensees? (Score:5, Interesting)
It is? An open source but still mostly experimental CPU based upon CPU technology from the 1980s? With virtually no implementations that run at efficiencies close to what's in the average cellphone, let alone what's in a desktop or server?
RISC-V is gaining a lot of traction lately, with for example Android announcing support for it. Also Chinese hardware makers are looking at it as a way to be more sanctions-proof.
ARM is based on the same 1980's RISC base, while the other big player (x86-64) is based on technology from the 1970's. Not all old ideas are outdated: many programming languages are based on fundamental ideas from the 1960's.
For a long time, ARM itself was only used for low-power devices, but Apple's M1 showed that the architecture is suitable for high performance as well. If enough resources are thrown at desiging RISC-V chips, I don't see why it wouldn't be able to perform.
How? And given RISC-V's main virtue right now is in being a free-for-all design that can easily be embedded in a larger ASIC/PAL/etc, how is bankrolling the fabbing of its own CPUs going to help fight that? That's like Microsoft deciding Linux is a huge threat so deciding to sell more XBoxes.
It ensures that there will be at least one customer for ARM's designs: their own chips division. If that is their reasoning, they must be pretty desperate. But the only other explanation I can think of is that ARM doesn't realize that competing with their own customers will make it more difficult to sell licenses, which seems even less likely.
Re: (Score:2)
RISC-V needs some big players to throw serious money at it, but if that happens it's not certain that they will share the fruits of their work. With ARM if ARM Ltd. develops some new tech (like big.LITTLE or NEON) then all their licensees can make use of it, as well as their efforts to improve GCC and LLVM.
Apple's M1 and M2 are still quite mediocre compared to x86 for performance. Or more accurately, compared to AMD64. While x86 might go back to the 70s, AMD64 is much newer and has some RISC-like features.
U
Re: (Score:2)
An open source but still mostly experimental CPU
A lot is happening with RISC-V. New boards are announced every week. Companies are working on high-end designs.
based upon CPU technology from the 1980s?
Are you serious? RISC-V is way newer tech than ARM.
Re: (Score:2)
There are a metric buttload of 48-60 MHz ARM-based MCU's out there, orders of magnitude more than there are multi-GHz ARMs. There are probably 3 or 4 of them in your phone complementing the multi-GHz main CPU, there are probably 50 of them in your car. This is a market that RISC-V is primed to take over today, while it matures and reaches into the high-end over the next 10 years. Despite these chips being vastly cheaper than the latest Qualcomm chip, they are likely a larger source of revenue to ARM, a r
Re: (Score:2)
RISC-V is a huge problem for ARM.
It is? An open source but still mostly experimental CPU based upon CPU technology from the 1980s?
That used to describe ARM, and to some extent, it still does.
given RISC-V's main virtue right now is in being a free-for-all design that can easily be embedded in a larger ASIC/PAL/etc
That's no different from ARM, except that ARM has licensing costs. Nobody else has Apple's ARM cores, so they're not an advantage to ARM, only to Apple.
It's pretty easy to see why no licensing costs is an advantage. It worked for Linux! Hardware is not software, but many of the same rules apply any time you're selling something.
Re: (Score:2)
RISC-V is a huge problem for ARM.
On the one hand, selling their own chips could be a way to survive if RISC-V succeeds in eroding ARM's market share. On the other hand, ARM selling their own chips significantly increases the chances of existing ARM licensees adopting RISC-V.
Re: (Score:2)
On the other hand, ARM selling their own chips significantly increases the chances of existing ARM licensees adopting RISC-V.
Why would one who is making his own ARM chips (one who has an ARM license), switch to RISC-V just because the guy who sold him the license is now making chips, too?
Your reasoning makes no sense.
Re:why is ARM competing with its own Licensees? (Score:4, Interesting)
We built custom chips with a licensed processor in them (not ARM, but similar). Every time we created a new chip (once or twice a year), there was a license fee to be paid. Our chips sold for about $1, so every license fee was a significant hit to profits. We would have switched in an instant.
Now, imagine an embedded MCU supplier like NXP - they build the 50 or so MCU's that you might find in your car running everything from the antilock brakes to the personalized seating. Perhaps they charge $4 for their MCU; but ARM comes along and starts selling a similar device for $3.50 because they don't have to pay a license fee to themselves. If NXP wants to stay in business, they'll switch to RISC-V in a heartbeat just to stay price-competitive. And it's not like absolute performance is much of an issue in moving six motors in the drivers seat to defined position....
AI will accelerate code optimisation chunks (Score:2)
Re: (Score:3)
The only ARM licensee actually designing their own CPU cores is Apple. Qualcomm used to, but with the move to AArch64 they switched to using slightly modified ARM Cortex cores. Samsung always licensed ARM cores from others. DEC designed the StrongARM, which was the basis for Intel's Xscale, but they sold that to Marvell and it just petered out. TI had the OMAP ARM implementation, but that was years ago now.
Re: (Score:2)
The only ARM licensee actually designing their own CPU cores is Apple.
The only person who used the word "cores" is you, you're talking to yourself there. ARM isn't competing by designing cores, they're competing by designing a complete CPU. Allegedly, anyway, according to the middle paragraph of TFS.
DEC designed the StrongARM, which was the basis for Intel's Xscale, but they sold that to Marvell and it just petered out.
The design would scale up, but it wouldn't scale down (in power) — Intel couldn't manage it either. Intel was able to sell it into stationary applications where power didn't matter, and an x86 processor was overkill, but the design wasn't worth updating so they let it die.
Re: (Score:3)
That is incorrect. There are several ARM licensees designing their own cores. Not just Apple. One example is the Fujitsu A64FX processor used in the Fugaku supercomputer which had a microarchitecture fully designed by Fujutsu. Qualcomm also designs their own cores for the Kryo SoCs using semi-custom design with some components from ARM. Huawei's HiSilicon designed the cores in the Kunpeng 920 processor.
No, Probably Not (Score:4, Insightful)
People are jumping the gun here. These appear to be little more than test chips with manufacturing partners. This is something Arm does on the regular; Arm's customers want IP that's already been verified on the process node said customer tends to use, which means Arm has to get test chips built to do that verification.
Re: (Score:1)
I have no doubts that RISC-V will be on one of my boards in the future. But it will be a while. The silicon side needs quite a bit of maturity as does the softwa
Re: (Score:2)
I would assume we're a couple of years away from the software maturing. Namely the vector 1.0 and virtualization extensions aren't yet in current products such as VisionFive 2, leaving plenty of room to optimize with respect to, say, a web browser or an Android runtime.
Pine64 are already pushing a RISC-V tablet even though they acknowledge the performance will lag but perhaps the next iteration will be adequate - nevertheless I can't see them catching up to Snapdragon/M2/Exynos any time soon.