What are you talking about? Samsung was making ARM chips way before Apple.
If it wasn't for Samsung, there wouldn't even be an iPhone as the earlier iPhone's were Samsung phones with an Apple logo on it.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Samsung to Rival Apple Silicon by Developing Custom PC and Mobile Processors
- Thread starter MacRumors
- Start date
- Sort by reaction score
What are you talking about? Samsung was making ARM chips way before Apple.
If it wasn't for Samsung, there wouldn't even be an iPhone as the earlier iPhone's were Samsung phones with an Apple logo on it.
This can be done and is common. As an example, I have both Raspberry RP2040 chips that are ARM and some ESP32 chips that are Risk V. I am working on a motor controller and I test using both chips using pretty much the same software.
Yes of course the binary executables can not possibly run on the different architectures, but the source code is the same
What Samsung would need to is the source code of the OS. Will Microsoft Allow this? I think so with a few trusted partners.
Taking a customer provided design and producing it is very different than creating that design in the first place. That design step is what this article is about. And those old pre-A-series chips were basically just ordinary ARM designs until Apple started making their own.
Most people marry to the brand and the marketing, no to the “superior” tech. That’s why I still keep buying IPhones year after year of no-innovation…
Android is distributed to Samsung as source code. Samsung can modify it as they wish. The already likely do
Apple were involved in ARM development way before that.
ARM (the company) started out in 1990 as a joint venture between Acorn (who had already developed the original ARM chip), VLSI (who had manufactured the original ARM chips) and Apple (who wanted an ARM chip for the Apple Newton).
Samsung didn't design those, they were "stock" ARM designs. They just manufactured them. Also, 15 years later, the chips are an order of magnitude more complex, and now Samsung is planning on DESIGNING their own chip. (Also likely with an ARM core, just like Apple.)
I'm not saying they can't or anything, but it is an incredible undertaking. That's why their last effort failed and they shuttered it in 2019, as the article states.
This isn't about manufacturing, they are quite good at that. The design is the hard part.
Sounds like a brain dead executive without a single technical bone in his / her body thinks it will just magically happen even though they've already failed at it. Who knows, maybe this time they'll do better?
And if they were found lacking in "power efficiency, heat generation, and multi-core efficiency" then what exactly were they good at? Draining your battery? Keeping you warm?
Not as bad as the monopoly abuse and lack of competition under Intel. AMD only exist as a historical accident because IBM demanded a second source for their chips back in the 1980s.
Any Samsung chip will most likely be based on the ARM instruction set (unless ARM ltd. self destruct - not impossible - but until then companies can just license the instruction set and develop their own cores). RISC-V is a faint possibility.
Anyway - I think the days of the chip instruction set being so significant are coming to an end
Modern Windows software developed using the common language runtime is distributed as processor-independent microcode, likewise for Android, and a "Just in time" translated into native code. "All" the developer of a new chip has to do is implement the appropriate runtime for their operating system and the appropriate compiler backends. Even Apple has a system whereby Apps can be uploaded as bytecode and automatically translated "on delivery" - although AFAIK they're only using this to cope with different generations of A-series chips at the moment. Then, look at how relatively painless Apple's Intel-to-ARM has been - sure, there are problems and bad luck if you're affected, but it's amazing how much "just works" and gives reasonable performance with Rosetta 2.
Once you're talking about modern software that uses abstracted libraries and frameworks rather than directly accessing processor features like SIMD instructions and neural engines there's really no need or incentive to use lovingly hand-crafted assembly language and the CPU instruction set becomes increasingly irrelevant. The problem is with legacy code - which is a far bigger issue for Microsoft (Windows has only recently dropped support for 16-bit era code) than Apple (who have been pretty ruthless in dropping support for old software).
I interpreted the article that Samsung was going to be designing its own CPU cores for the ARM ISA, which is exactly what Apple does. (People get ISA and microarchitecture confused but they’re two very distinct things).
Early versions of the Exynos and A-series used off-the-shelf ARM Cortex CPU designs. Apple eventually started designing its own CPU cores that ran the ARM ISA (similar to what AMD does with x86-64). It sounds like Samsung will be doing the same thing, there’s nothing indicating they’re trying to create their own ISA on top of their own CPU microarchitecture designs. There’s just no benefit to that when they likely already have an ARM license.
Everything is moving to mobile. Desktop PCs will not exist as we know them today in 20 years. Why would I waste all that money on something I can't take with me? Also, Console has surpassed PC gaming for the first time this generation. My PS5 at $499 plays a game with the same basic graphic fidelity as a PC with a high end card. For $499.
But as Apple is showing, you can build an extremely enticing RISC ARMx64 architecture that sure, may not have the clock or RAM of an Intel machine, but RAM and Clock comparisons are for morons. Bertrand and Jon wrote the book on how clock and ram are a gimmick. My Intel i9 MacBook Pro 2016 with 16 GB of RAM is completely outmatched by my 13.3" MacBook Pro M1 8GB.
From battery life to task time, the i9 just held no advantage in lead. For a 10 minute task on the 8GB M1, it was 8 and some change on the i9. Saved a whole two minutes. For the hour long task on an M1 8GB, it was 50 minutes on i9. Again, begs the question of why bother with Intel for a marginal increase in performance.
And that is just the FIRST GENERATION of M-Series. Apple aint gonna stop improving it. By M3, I think Apple will have proven that CISC X86-64 is a dead end with no innovation left and that Arm64 its the future.
Same with gaming PCs. Eventually, I got tired of always keeping up with graphics cards, and getting a chance to buy one, and the constant upgrading and drivers and support and I got to a point where I asked, "Can I just get a Jupiter damned box that stays the same and plays games at the same quality?"
Why anyone would want to stick around for an industry with no real gameplan or coordination. Look at the disaster with S3 Sleep and Windows 10/11. Intel problem!
This will be strategically very difficult for Samsung to pull off successfully.
Not because I don't think Samsung couldn't make their own processors, but rather because creating their own chipsets and optimizing everything for them lengthens the product planning cycle, thereby compromising Samsung's ability to quickly follow and imitate whatever Apple or other competitor creates.
Doing so would require Samsung having their own vision for where phones, tablets, laptops, and watches are going, and there's scant evidence that Samsung even knows what a vision is
Not because I don't think Samsung couldn't make their own processors, but rather because creating their own chipsets and optimizing everything for them lengthens the product planning cycle, thereby compromising Samsung's ability to quickly follow and imitate whatever Apple or other competitor creates.
Doing so would require Samsung having their own vision for where phones, tablets, laptops, and watches are going, and there's scant evidence that Samsung even knows what a vision is
I agree that what you're saying is probably what's happening, but the way it's being written makes it all a bit unclear. Maybe because the authors don't fully understand the distinctions. The language being used seems to suggest that Samsung is breaking with Arm in a more serious way (the MR wording is vague, but I suspect because the source article also isn't clear):
"Samsung has historically relied on British company ARM for CPU cores that go into its Exynos APs. Samsung’s rival Qualcomm also makes APs based on ARM’s designs. If Samsung succeeds in developing its own CPU cores, it will be able to greatly improve the optimization of its smartphones."
"Moreover, potential risk is growing over Samsung Electronics’ alliance with ARM. This explains why Samsung Electronics has to expedite the development of its own CPUs."
As far as I can tell, there's nothing explaining what that "growing risk" is about.
The Mongoose cores mentioned were Arm ISA, so maybe this is just a repeat of that effort.
The story says “the first version of the Galaxy Chip is highly likely to have ARM CPUs.” Which is indicating that their following versions of the Galaxy chip is likely to NOT have ARM CPU’s. As they’re bringing on someone from AMD, maybe their focus will be on something x86 compatible.
I’m guessing in a few weeks they’ll need to clarify this as there are a lot of people that want to know if they’re “leaving ARM” what ISA do they plan to support?
Is this article implying that the Samsung’s SoC’s will NOT be based on ARM eventually? Because that’d be wild. Is anyone in the mobile space not using some variant of an ARM design?
Considering how much Samsung's Exynos chips get criticized, I wonder how these will do... their own S23 series dropped Exynos entirely too. 😅
They foundry-ed the first iPhone CPUs, Apple was already doing the design.
Edit: Apple was already at work designing the first A series chips when the first iPhone was released, but an Apple designed chip didn't appear in an iPhone till the A4, 3 years later. And yes, using Samsung as the foundry.
Last edited:
Everything you said applies to consumer pc. The state in the corporate world is completely different. Business is learning the hard way that you have to design and protect your data. That’s job number one. Corporate leaders are going to go back to the beginning. Data processed in data centers and sent out to employers using tiny little dumb terminals. Don’t worry the players are going to be the same people that they are now.
Good luck Samsung you’re going to need it. More competition is good. Let’s see if Samsung can actually bring it.