The name of this CPU is bordering on securities fraud. When people see the term "AGI" now, they are assuming "Artificial General Intelligence", not "Agentic AI Infrastructure".
Of course people don't realize that, and people will buy ARM stock thinking they've cracked AGI. The people running Arm absolutely know this, so this name is what we in the industry call a "lie".
The marketers did this for 5G also, calling their product 5G before it was actually deployed, only because theirs came after 4G but wanted to ride the upcoming 5G buzz.
It seems marketing /depends/ on conflating terms and misleading consumers. Shakespeare might have gotten it wrong with his quip about lawyers.
There was soooo much intentional disinformation around 5G. Everyone who wanted to sell anything intentionally confused the >1Gbps millimeter wave line-of-sight kind of 5G with the "4G but with some changes to handle more devices connected to one tower" kind of 5G. I wonder how many bought a "5G phone" expecting millimeter wave but only got the slightly improved 4G.
If rich people are this stupid then they deserve to be parted with their cash.
If you invest money so mindlessly that you don’t even check what you buy, then no legislation in the world will manage to protect you from your own mind
It’s not just rich people though. Most people (at least in the US) have their retirements and the like in things like 401ks, tied to some kind of index like the S&P 500. A company doing bullshit to manipulate the stock affects pretty much anyone who uses an index fund or ETF, which is pretty much everyone in the US.
This smells like the beginning of entshitification at ARM. I'm not saying AMD or Intel are a whole lot better, but the move to compete with licencees of ARM tech and to cheekily use AGI in the name is not going to ensure confidence in the short or long term.
ARM have given enough time for their licenses, even with special terms and discount just hoping they could crack the server market. But six years in and looking at the next 3-4 years. They are no where near competing with x86 on server other than on hypersalers. I don't see how entering this market now is enshitification . If anything I wish they had enter the market two years earlier.
This sort of thing really bugs me! Marketing departments appropriate an existing term and use it in some new, often deceptive way. This goes all the way back to when IBM released “The IBM Personal Computer”, at a time when “personal computer” was a category name. Then Microsoft released Windows, when “windows” was a generic term for windowing systems. Intel did it with their “core” architecture. The list goes on.
Yes, that's how fraud works a lot of the time. It removes you from the market but not until after it's removed your money. And there's an endless supply of new people ready to make the same mistake after you've learned your lesson.
Does an iced tea company changing their name to Long Blockchain make any sense? No, not really, it's pretty stupid actually, but it managed to bump the stock by apparently 380%.
The stock market can be pretty dumb sometimes. Let's not forget the weird GME bubble.
Marketing is marketing, nothing about it was ever about being factual when there is a total addressable market to go after and dollars to be made! This is inline with much of the other marketing that exists in the AI space as it stands now, not mention the use of AGI within the space as it stands currently.
I'm not saying anything is going to happen, ARM holdings has a lot more money and lawyers than Long Blockchain did, but I'm just saying that it's not weird to think that a deceptive name could be considered false advertising.
An unappreciated aspect of Arm is they really were the Robin Saxby show. https://en.wikipedia.org/wiki/Robin_Saxby Whichever ISA had him selling it was going to win.
While AArch64 represents the technical revolution they needed their business compass has just gone ever since he stepped down. This grimy stuff, and as others noted competing with your own customers, were no goes in the earlier era.
Do you think that we should live in a world where investors who buy on a comical misinterpretation of an acronym are protected from their naivety?
Why isn't there a minority shareholder lawsuit on the news because someone bought MSFT not realizing that Copilot isn't actually certified to fly an airliner? A certain type of people would likely just buy MSFT on a massive lever and then if the bet fails to work out sue pretending that they did not understand.
People have been hearing for the last three years about how a specific acronym, "AGI", is the final frontier of artificial intelligence and how it's going to change the entire economy around it. They've been hearing about this quasi-theoretical, very specific thing, and a lot of them don't even know what the "G" stands for.
People haven't been hearing for years about a mythical "copilot", and as such I think people are much more likely to think it's not anything more than a cute nickname.
Are you suggesting that this is just a coincidence? The acronym AGI doesn't even make sense for Agentic AI Infrastructure, which should be AAII; they're clearly calling it AGI to mislead people. I refuse to think that the people running Arm are so stupid that they didn't even Google the acronym before releasing the chip.
You think it's a "comical misinterpretation", but I don't think it is. When I saw the article, I thought "shit; did they manage to crack AGI?", and I clicked the article and was disappointed. I suspect a lot of people aren't even going to read the press release.
I mean we can all meme on investors, but I don't thing many people can submit a buy order whilst assuming they missed the AGI news headline because of a product name.
It's just going the way of "Smartphone" and "Smart Car" they'll market it as such to get people riled up about it. Consumers will eat it up. I'm sure Scam Altman is ready to show us "AGI" next too. If ARM is making AGI's meaning shift to a CPU descriptor, anyone can call their tech "AGI" by just using their chips.
> The name of this CPU is bordering on securities fraud.
No. For it to be securities fraud, Arm would need to make a materially false statement of fact that misleads investors. Naming the CPU in this way doesn't clear the bar because:
a) the name is clearly product brand, similar to how macOS Lion, or Microsoft Windows, or Ford Mustang, or Yves Saint Laurent Black Opium don't mean literally what they say)
b) Arm explicitly defines it as silicon "designed to power the next generation of AI infrastructure", with the technical specs fully disclosed
c) sophisticated investors, the relevant standard for securities fraud, can read a spec sheet
d) Arms' EVP said "We think that the CPU is going to be fundamental to ultimately achieving AGI", framing it as contribution towards AGI, not AGI itself
I was on board with A through C, but then with D it's either clearly a lie or stupidity. I guess it's not a lie technically if they believe it though, so the latter then. But I also don't want to assume someone in their position to be stupid, so then I'm back to the former.
Honestly: The people who buy stock because a product says "AGI" in the name deserve to lose their shirt.
And no, it's not "a lie", because only an utter idiot would consider a product name an actual fact. It's a name. The Hopper GPUs also didn't ship with a lifesize cutout of Grace Hopper.
No, it's actually a lie, and it's different than the Hopper GPU you mentioned.
People have been seeing every big AI company talk about how AGI is the holy grail of AI, and how they're all trying to reach it. Arm naming a chip AGI is clearly meant to make casual observers think they cracked AGI.
The Hopper GPU isn't the same, because Nvidia isn't actively trying to make people think that it includes a lifesize cutout of Grace Hopper. Not a dig on her, but most people don't know who Grace Hopper is, people haven't been hearing on the news for the last several years about how having a Grace Hopper is going to make every job irrelevant.
If you showed someone what our computers can do with the latest LLMs now to someone 5 years ago they would probably say it sure looks a lot like AGI.
We have to keep defining AGI upwards or nitpick it to show that we haven't achieved it.
I would argue that LLMs are actually smarter than the majority of humans right now. LLMs do not have quite the agency that humans have, but their intelligence is pretty decent.
We don't have clear ASI yet, but we definitely are in a AGI-era.
I think we are missing an ego/motiviations in the AGI and them having self-sufficiency independent of us, but that is just a bit of engineering that would actually make them more dangerous, it isn't really a significant scientific hurdle.
Ok, but it's not AGI. People five years ago would have been wrong. People who don't have all the information are often wrong about things.
ETA:
You updated your comment, which is fine but I wanted to reply to your points.
> I would argue that LLMs are actually smarter than the majority of humans right now. LLMs do not have quite the agency that humans have, but their intelligence is pretty decent.
I would actually argue that they are decidedly not smarter than even dumb humans right now. They're useful but they are glorified text predictors. Yes, they have more individual facts memorized than the average person but that's not the same thing; Wikipedia, even before LLMs also had many more facts than the average person but you wouldn't say that Wikipedia is "smarter" than a human because that doesn't make sense.
Intelligence isn't just about memorizing facts, it's about reasoning. The recent Esolang benchmarks indicate that these LLMs are actually pretty bad at that.
> We don't have clear ASI yet, but we definitely are in a AGI-era.
My definition of AGI hasn't changed - it's something that can perform, or learn to perform, any intellectual task that a human can.
5 years ago we thought that language is the be-all and end-all of intelligence and treated it as the most impressive thing humans do. We were wrong. We now have these models that are very good at language, but still very bad at tasks that we wrongly considered prerequisites for language.
> If you showed someone what our computers can do with the latest LLMs now to someone 5 years ago they would probably say it sure looks a lot like AGI.
Would they? Perhaps if you only showed them glossy demos that obscure all the ways in which LLMs fail catastrophically and are very obviously nowhere even close to AGI.
Certainly, they wouldn't expect that an AI able to score 150 on an IQ test is unable to play a casual game of chess because it isn't coherent enough to play without making illegal moves.
It doesn't look anything like AGI and no one who knows what that means would be confused in any era.
Is it useful? Yes. Is it as smart as a person? Not even remotely. It can't even remember things it already was told 5 minutes ago. Sometimes even if they are still in the context window un compacted!
> If you showed someone what our computers can do with the latest LLMs now to someone 5 years ago they would probably say it sure looks a lot like AGI.
But this is a CPU! It's not a GPU / TPU. Even if you think we've achieved AGI, this is not where the matrix multiplication magic happens. It's pure marketing hype.
I did AI back before it was cool and I think we have agi. Imo the whole distinction was from extremely narrow AI to general intelligence. A classifier for engine failure can only do that - a route planner can only do that…
Now we have things I can ask a pretty arbitrary question and they can answer it. Translate, understand nuance (the multitude of ways of parsing sentences, getting sarcasm was an unsolved problem), write code, go and read and find answers elsewhere, use tools… these aren’t one trick ponies.
There are finer points to this where the level of autonomy or learning over time may be important parts to you but to me it was the generality that was the important part. And I think we’re clearly there.
Agi doesn’t have to be human level, and it doesn’t have to be equal to experts in every field all at once.
> LLMs are actually smarter than the majority of humans right now
I consider myself a bit of a misanthrope but this makes me an optimist by comparison.
Even stupid people are waaaaaay smarter than any LLM.
The problem is the continued habit humans have of anthropomorphizing computers that spit out pretty words. It’s like Eliza only prettier. More useful for sure. Still just a computer.
A human can think logically with reason, not to say they are smart or smarter. But LLMs cannot. You can convince a LLM anything is correct and it will believe you. You can't convince a human anything is correct.
I can't argue that LLMs do not know an absolute insane amount of information about everything. But you can't just say LLMs are smarter then most humans. We've already decided that smartness is not about how much data you know, but thinking about that data with logical reasoning. Including the fact it may or may not be true.
I can run a LLM through absolutely incorrect data, and tell it that data is 100% true. Then ask it questions about that data and get those incorrect results as answers. That's not easy to do with humans.
The problem with definitions is that they are all wrong when you try to apply them outside mathematical models. Descriptive terms are more useful than normative ones when you are dealing with the real world. Their meaning naturally evolves when people understand the topic better.
General intelligence, as a description, covers many aspects of intelligence. I would say that the current AIs are almost but not quite generally intelligent. They still have severe deficiencies in learning and long-term memory. As a consequence, they tend to get worse rather than better with experience. To work around those deficiencies, people routinely discard the context and start over with a fresh instance.
This is just a Neoverse CPU that Arm will manufacture themselves at TSMC and then sell directly to customers.
It isn't an "AI" CPU. There is nothing AI about it. There is nothing about it that makes it more AI than Graviton, Epyc, Xeon, etc.
This was already revealed in the Qualcomm vs Arm lawsuit a few years ago. Qualcomm accused Arm of planning to sell their CPUs directly instead of just licensing. Arm's CEO at the time denied it. Qualcomm ends up being right.
Ha, I wasn't old or into it enough at the time to remember that, but it is consistent with just about every IC datasheet ever with their list of possible applications. (Like: logic gate; applications include Walkman, Rocket ship, Fuzzy Logic Washing Machine, mobile phone, AGI co-processor, ...)
The Dell marketing machine in particular is bludgeoning everyone that will listen about Dell AI PCs. The implication that folks will miss the boat on AI by not having a piddly NPU in their laptop is silly.
This was exactly my first thought when I saw the title. And after reading the contents of the blog, it's pretty clear that ARM is laser focused on getting a piece of their customer's cake by competing with them. This is likely why they are riding the AI hype train hard with their ill-suited name (AGI).
Unfortunately for them, I think hardware vendors will see past the hype. They'll only buy the platform if it is very competitively priced (i.e., much cheaper) since fortune favours long-lived platforms and organizations like Apple and Qualcomm.
I can imagine a lot of ARM engineers being frustrated at seeing their cores being used in stupid ways for decades to finally flex what they can do (outside of Apple).
Agreed, it will be _very_ interesting to see what waves this causes. It would be like TSMC deciding to make and sell their own CPUs, now ARM is directly competing with some of their clients.
Eh, I'm not so sure it'll be that big a deal. The whole supply chain is so twisted and tangled all the way up and down. Shuffling out one piece doesn't seem like it will, on its own, be so major. Samsung made the chips for the iPhone, then made their own phone, then Apple designed their own chips made by TSMC, now Apple is exploring the possibility of having Samsung make those chips again.
Also, it takes a willful ignorance of history for ARM to claim this is the first time they've manufactured hardware. I mean, maaaaybe, teeeeechnically that's true, but ARM was the Acorn RISC Machine, and Acorn was in the hardware business...at least as much as Apple was for the first iPhone.
Well technically the ARM1 was a Acorn product (made by VLSI). ARM as a company was only incorporated in 1990 (as a joint venture between Acorn, VLSI and drumroll Apple), I guess that's where the mentioned 35 years and "first time in our history" come from.
Can this be read as finally the financial incentives to join the AI silicone race has become too tempting. Finally the incentives to sell chips are definitely stronger than the cost of competing with your own licensees?
> I work at ARM, we're launching a new CPU optimized for LLM usage. We're thinking of calling it "Arm Agentic AI Infrastructure CPU", or "Arm AGI CPU" for short. Do you think this is a good idea?
> No. I would not use it as the product name. “AGI CPU” will be read as artificial general intelligence, not “agentic AI infrastructure,” so it invites confusion and sounds hypey.
To bad these executives seemingly don't have access to ChatGPT.
They did ask AI if AGI what a great name.
It said that it was the greatest name possible. It's bold, aspirational, and ... polarizing?!
Oh god! Mistral tell me it's highly polarizing, will make the buzz and it's risky but anyway people will know that ARM is doing CPU again now (maybe I did put too many context).
> Studies 1-5 showed that people are disproportionately likely to live in places whose names resemble their own first or last names (e.g., people named Louis are disproportionately likely to live in St. Louis).
When I lived in Austin, it seemed like a third of boys born were being named Austin. I presume many of them will end up living there as adults but not because of this particular bias, because they were raised there and have family’s there seems to be a more likely driver.
Seems more likely this falls under the replication crisis umbrella. My wife's favorite numbers are my birthday (mm-dd), which is a small reason she fell in love with me. Neither of those numbers are related to her birthday. My favorite number(s) do not overlap with my birthday. Maybe my mm-dd values just aren't low enough, like 02-02?
Meta are heavily invested in building their own chips with ARM to reduce their reliance on Nvidia as everyone is going after their (Nvidia) data center revenues.
This is why Meta acquired a chip startup for this reason [0] months ago.
Call this an “AGI CPU” just feels like the most out of touch, terrible marketing possible. Maybe this is unfair but it makes me think ARM as a whole is incompetent just because it is so tasteless.
> Arm has additionally partnered with Supermicro on a liquid-cooled 200kW design capable of housing 336 Arm AGI CPUs for over 45,000 cores.
Also just bad timing on trying to brag about a partnership with Supermicro, after a founder was just indicted on charges of smuggling Nvidia GPUs. Just bizarre to mention them at all.
If you read past the marketing talk, this is basically a massively multicore system (136) with significantly reduced power usage (300W).
Where does Agentic come into this? ARMs explanation is that future Agentic workloads will be both CPU and GPU bound thus the need for significant CPU efficiency.
How fun would it be if due to improved chips handling more model state RAM needs are reduced and Sama cannot make all those RAM purchases he booked?
VC without a degree who has no grasp of hardware engineering failed up when all he had to do was noodle numbers in an Excel sheet.
He is so far behind the hardware scene he thinks its sitting still and RAM requirements will be a nice linear path to AGI. Not if new chips optimized for model streaming crater RAM needs.
Hilarious how last decades software geniuses are being revealed as incompetent finance engineers whose success was all due to ZIRP offering endless runway.
The thing they are good at is bullshitting and selling hype. Which we see here doesn't mean they are actually going to be good at running a business. Smart leaders understand they are not omnipotent and omniscient so they surround themselves who know how to get things done. Weak, narcissist leaders think they're the smartest one in the room and fail.
Unfortunately failing upwards is still somehow common, probably because the skill of parting fools from their money is still valuable.
No, he is also good at networking. When OpenAI was mission-driven and Sam was more respected, he could convince the most talented people to work for him.
Now the talent is going to other places for a variety of reasons, not all due to Sam (one of which is little room for options to grow). However it’s hard to believe his tanking reputation is not badly hurting the company. Other than Jakub and Greg, I believe there are not many top tier people left, those in top positions are there because they are yes-men to Sam.
What RAM? OpenAI booked the silicon wafers, they can print anything they want on them. I wouldn't call them "far behind" on hardware when OpenAI are actively buying Cerebras chips.
Yes exactly; he is behind in that he has to buy others chips with little say on how they work.
Apple and Google control their own designs.
Sama is 100% an outsider, merely a customer. The chip insiders are onto his effort to pivot out of meme-stock hyping, into owning a chunk of their fiefdom. They laughed off his claims a couple years ago as insane VC gibberish (third hand paraphrase from social network in chip and hardware land).
No way he can pivot and print whatever. Relative to hardware industry he is one of those programmers who can say just enough to get an interview but whiffs the code challenge.
He has no idea where the bleeding edge is so he will just release dated designs. Chip IP is a moat.
Plus a bunch of RAM companies would be left hanging; no orders, no wafers. Sama risks being Jimmy Hoffa'd imploding the asset values of other billionaires.
For those wanting to know more about software stack,
> Arm is actively collaborating with leading Linux distributions from Canonical, Red Hat, and SUSE to ensure certified support for the production systems.
It isn't obvious to me that they intended to give this as the maximum single-core performance, or just the proportional share of 844GB/s across 136 cores. Implementations of Neoverse V2 by Nvidia and Amazon hit 20-30GB/s in single-threaded work.
If I try to cut through the hype, it seems the main features of this processor, or rather processor + memory controller + system architecture, is < 100 ns for accessing anything in system memory and 6 GB/sec for each of a large-ish number of cores, so a (much?) higher overall bandwidth than what we would see in a comparable Intel x86_64 machine.
I also miss the all-capitals ARM spelling. I think they've never been the same since they've changed that, since around the same time their business strategy went from sensible to nonsense.
Poor TSMC (and ASML)! They were already struggling with capacity to fulfill orders from their established customers. With ARM now joining the party, I don't know how they're going to cope.
Edit: The new CPU will be built with the soon-to-be-former leading edge process of 3nm lithography.
TSMC has multiple fabs being constructed, they'll be okay. The biggest losers here are AMD, Intel and Apple who will be forced to pay AI-hype prices to mass-produce boring consumer hardware.
It only took a quarter century, but I'm glad that somebody is finally adding a little multicore competition since Moore's law began failing in the mid-2000s.
I looked around a bit, and the going rate appears to be about $10,000 per 64 cores, or around $150 per core. Here is an Intel Xeon Platinum 8592+ 64 Core Processor with 61 billion transistors:
So that's about 500 million transistors per dollar, or 1 billion transistors for $2.
It looks like Arm's 136 core Neoverse V3 has between 150 and 200 billion transistors, so it should cost around $400. Each blade has 2 of those chips, so should be around $800-1000 for compute. It doesn't say how much memory the blades come with, but that's a secondary concern.
Note that this is way too many cores for 1 bus, since by Amdahl's law, more than about 4-8 cores per bus typically results in the remaining cores getting wasted. Real-world performance will be bandwidth-limited, so I would expect a blade to perform about the same as a 16-64 core computer. But that depends on mesh topology, so maybe I'm wrong (AI thinks I might be):
Intel Xeon Scalable: Switched from a Ring to a Mesh Architecture starting with Skylake-SP to handle higher core counts.
Arm Neoverse V3 / AGI: Uses the Arm CMN-700 (Coherent Mesh Network), which is a high-bandwidth 2D mesh designed specifically to link over 100 cores and multiple memory controllers.
I find all of this to be somewhat exhausting. We're long overdue for modular transputers. I'm envisioning small boards with 4-16 cores between 1-4 GHz and 1-16 GB of memory approaching $100 or less with economies of scale. They would be stackable horizontally and vertically, to easily create clusters with as many cores as one desires. The cluster could appear to the user as an array of separate computers, a single multicore computer running in a unified address space, or various custom configurations. Then libraries could provide APIs to run existing 3D, AI, tensor and similar SIMD code, since it's trivial to run SIMD on MIMD but very challenging to run MIMD on SIMD. This is similar to how we often see Lisp runtimes written in C/C++, but never C/C++ runtimes written in Lisp.
It would have been unthinkable to design such a thing even a year ago, but with the arrival of AI, that seems straightforward, even pedestrian. If this design ever manifests, I do wonder how hard it would be to get into a fab. It's a chicken and egg problem, because people can't imagine a world that isn't compute-bound, just like they couldn't imagine a world after the arrival of AI.
Edit: https://news.ycombinator.com/item?id=47506641 has Arm AGI specs. Looks like it has DDR5-8800 (12x DDR5 channels) so that's just under 12 cores per bus, which actually aligns well with Amdahl's law. Maybe Arm is building the transputer I always wanted. I just wish prices were an order of magnitude lower so that we could actually play around with this stuff.
Many of these words are unexplained. "Memory and I/O on the same die". Oh? What does this mean? All of the DRAM in the photo/render is still on sticks. Do they mean the memory controller? Or is there an embedded DRAM component?
What is “agentic AI cloud era” referring to? I honestly don’t know what this buzz-speak is targeting. Running models locally on the server, for cloud workloads? Agentic, that is just a LLM pattern.
Hmm all my experience with using AI has been mostly VRAM. I haven't experienced any bottleneck on the CPU side. What does this chip offer over Intel or Apple Silicon? Anyone expert here know whatit is?
The arm family of chips (apple A series, m series, and qcom snapdragon) are better on energy usage (thus battery life) and performance and design compared to many x86 style chips (intel, amd).
Time will tell if ARMs owncpu is on par or better than Apple’s ARM based chips
Huh, many companies use TSMC, in fact, probably all of them use TSMC, including Intel, yet there are only a few who dominates in performance. There are much more in designing chips than what you just listed.
There's a big difference between just providing IP and actually doing the physical design, manufacturing and packaging. You can't just send your RTL to TSMC and magically get packaged chips back.
I haven't ever ordered an ARM SoC but I also wouldn't be surprised if there were significant parts that they left up to integrators before - PLLs, pads, SRAM etc.
is this a cpu that's meant for AI training or is it more for serving inference? I don't quite get why I would want to buy an arm CPU over a nvidia GPU for ai applications.
I found this article extremely frustrating to read. Maybe I lack some required prior knowledge and I am not the target audience for this.
> built on the Arm Neoverse platform
What the heck is "Arm Neoverse"? No explanation given, link leads to website in Chinese. Using Firefox translating tool doesn't help much:
> Arm Neoverse delivers the best performance from the cloud to the edge
What? This is just a pile of buzzwords, it doesn't mean anything.
The article doesn't seem to contain any information on how much it costs or any performance benchmarks to compare it with other CPUs. It's all just marketing slop, basically.
> The ARM Neoverse is a group of 64-bit ARM processor cores licensed by Arm Holdings. The cores are intended for datacenter, edge computing, and high-performance computing use. The group consists of ARM Neoverse V-Series, ARM Neoverse N-Series, and ARM Neoverse E-Series.
More precisely, this Neoverse V3 core is the server version of the Cortex-X4 core from smartphones. The actual core is pretty much identical, but the cache memories and the interfaces between cores are different.
Neoverse V3 is also used in AWS Graviton5 and in several NVIDIA products.
You should look at the benchmarks of the Cortex-X4 cores used in many smartphones from 2 years ago, because it is the same core as Neoverse V3.
AWS Graviton5 uses the same cores, but it has 192 cores per socket.
So Graviton5 has more cores per socket, but I think that it does not support dual socket boards.
This Arm AGI supports dual socket boards, so it provides 272 cores per board, more than Graviton5 MBs.
However, this is puny in comparison with Intel Clearwater Forest, which provides 576 cores per board, and the Intel Darkmont cores are almost exactly equivalent for all characteristics with Arm Neoverse V3.
The name of this CPU is bordering on securities fraud. When people see the term "AGI" now, they are assuming "Artificial General Intelligence", not "Agentic AI Infrastructure".
Of course people don't realize that, and people will buy ARM stock thinking they've cracked AGI. The people running Arm absolutely know this, so this name is what we in the industry call a "lie".
Considering AGI has been degraded into a generic feelgood marketing word, I can't wait to get my AGI-scented deodorant.
Long Blockchain Corp. remembers [1].
[1] https://en.wikipedia.org/wiki/Long_Blockchain_Corp.
You can already drink AGI! Oh sorry, AG1. The resemblance must be a complete coincidence.
7 replies →
Artificial Gut Incense?
Buy it in combo with the good ol' Blockchain perfume!
1 reply →
> I can't wait to get my AGI-scented deodorant.
Old spice for me, thanks!
1 reply →
The marketers did this for 5G also, calling their product 5G before it was actually deployed, only because theirs came after 4G but wanted to ride the upcoming 5G buzz.
It seems marketing /depends/ on conflating terms and misleading consumers. Shakespeare might have gotten it wrong with his quip about lawyers.
https://www.pbs.org/newshour/economy/att-to-drop-misleading-...
There was soooo much intentional disinformation around 5G. Everyone who wanted to sell anything intentionally confused the >1Gbps millimeter wave line-of-sight kind of 5G with the "4G but with some changes to handle more devices connected to one tower" kind of 5G. I wonder how many bought a "5G phone" expecting millimeter wave but only got the slightly improved 4G.
4 replies →
Bill Hicks had some thoughts, too:
https://youtube.com/watch?v=GaD8y-CGhMw
1 reply →
Yes, my wireless router has "5G WiFi" but only does 4G. I didn't have a choice about using it since it comes from the provider, but still stupid.
3 replies →
If rich people are this stupid then they deserve to be parted with their cash.
If you invest money so mindlessly that you don’t even check what you buy, then no legislation in the world will manage to protect you from your own mind
It’s not just rich people though. Most people (at least in the US) have their retirements and the like in things like 401ks, tied to some kind of index like the S&P 500. A company doing bullshit to manipulate the stock affects pretty much anyone who uses an index fund or ETF, which is pretty much everyone in the US.
3 replies →
This smells like the beginning of entshitification at ARM. I'm not saying AMD or Intel are a whole lot better, but the move to compete with licencees of ARM tech and to cheekily use AGI in the name is not going to ensure confidence in the short or long term.
ARM have given enough time for their licenses, even with special terms and discount just hoping they could crack the server market. But six years in and looking at the next 3-4 years. They are no where near competing with x86 on server other than on hypersalers. I don't see how entering this market now is enshitification . If anything I wish they had enter the market two years earlier.
1 reply →
AGI is a poorly-defined concept anyway. It’s just vibes, nothing descriptive.
AGI is the automation of self-regulation of language
source: 100% personal certainty
idk, but when i see the abbreviation "AGI" i associate it with "ads generated income" ...
every other interpretation is hollowed out / reduced to "marketing speak" by now ;)
This sort of thing really bugs me! Marketing departments appropriate an existing term and use it in some new, often deceptive way. This goes all the way back to when IBM released “The IBM Personal Computer”, at a time when “personal computer” was a category name. Then Microsoft released Windows, when “windows” was a generic term for windowing systems. Intel did it with their “core” architecture. The list goes on.
(Disclosure: I am a casual investor in ARM.)
> Of course people don't realize that, and people will buy ARM stock thinking they've cracked AGI.
Doesn't seem like a very credible assertion. Picking stocks in this way would remove you from the market pretty quickly.
Didn't random companies add block chain to their names only just a few years ago and get 30+% jumps in stock price immediately?
6 replies →
Yes, that's how fraud works a lot of the time. It removes you from the market but not until after it's removed your money. And there's an endless supply of new people ready to make the same mistake after you've learned your lesson.
I didn't say it would be a wise decision to pick stocks that way, but this has already happened: https://en.wikipedia.org/wiki/Long_Blockchain_Corp.
Does an iced tea company changing their name to Long Blockchain make any sense? No, not really, it's pretty stupid actually, but it managed to bump the stock by apparently 380%.
The stock market can be pretty dumb sometimes. Let's not forget the weird GME bubble.
5 replies →
People ruined the word AGI before this to be fair, we will need another word to describe "real AGI" when it comes
Marketing is marketing, nothing about it was ever about being factual when there is a total addressable market to go after and dollars to be made! This is inline with much of the other marketing that exists in the AI space as it stands now, not mention the use of AGI within the space as it stands currently.
Sure, but there are plenty of cases where a deceptive name has been considered enough to at least warrant an investigation: https://en.wikipedia.org/wiki/Long_Blockchain_Corp.
I'm not saying anything is going to happen, ARM holdings has a lot more money and lawyers than Long Blockchain did, but I'm just saying that it's not weird to think that a deceptive name could be considered false advertising.
5 replies →
An unappreciated aspect of Arm is they really were the Robin Saxby show. https://en.wikipedia.org/wiki/Robin_Saxby Whichever ISA had him selling it was going to win.
While AArch64 represents the technical revolution they needed their business compass has just gone ever since he stepped down. This grimy stuff, and as others noted competing with your own customers, were no goes in the earlier era.
It's HD and ai and 5G and and that
Can you imagine being an engineer and working hard to create something new and cool and some jackass in marketing slaps the name “AGI CPU” on it?
Do you think that we should live in a world where investors who buy on a comical misinterpretation of an acronym are protected from their naivety?
Why isn't there a minority shareholder lawsuit on the news because someone bought MSFT not realizing that Copilot isn't actually certified to fly an airliner? A certain type of people would likely just buy MSFT on a massive lever and then if the bet fails to work out sue pretending that they did not understand.
You're being purposefully obtuse.
People have been hearing for the last three years about how a specific acronym, "AGI", is the final frontier of artificial intelligence and how it's going to change the entire economy around it. They've been hearing about this quasi-theoretical, very specific thing, and a lot of them don't even know what the "G" stands for.
People haven't been hearing for years about a mythical "copilot", and as such I think people are much more likely to think it's not anything more than a cute nickname.
Are you suggesting that this is just a coincidence? The acronym AGI doesn't even make sense for Agentic AI Infrastructure, which should be AAII; they're clearly calling it AGI to mislead people. I refuse to think that the people running Arm are so stupid that they didn't even Google the acronym before releasing the chip.
You think it's a "comical misinterpretation", but I don't think it is. When I saw the article, I thought "shit; did they manage to crack AGI?", and I clicked the article and was disappointed. I suspect a lot of people aren't even going to read the press release.
If sind can't do the most basic due diligence as in reading up on stuff you invest in using Wikipedia or a search engine, best of luck to them.
On the contrary, I love that companies are semantically overloading this stupid concept (purposefully or not) which is 100% hype marketing.
I don't understand why this label is still a thing in the current discourse, and I hope such moves will finally help people and the industry move on.
People buying these kinds of chips will know. AGI is barely a popular concept. Nobody in my family knows what it means.
I mean we can all meme on investors, but I don't thing many people can submit a buy order whilst assuming they missed the AGI news headline because of a product name.
Those in the industry don't call it a lie, they call it "marketing".
It's those out of the industry who call them lies.
Touché. I guess I should have said "I call it a lie".
the whole AI space is rife with much worse example of what could be considered securities fraud tbh
I'm "people" and AGI means nothing to me
In case you haven't noticed, this whole thing has been a grift since 2022. It's kind of amazing that nobody thought of making AGI processors before
It's just going the way of "Smartphone" and "Smart Car" they'll market it as such to get people riled up about it. Consumers will eat it up. I'm sure Scam Altman is ready to show us "AGI" next too. If ARM is making AGI's meaning shift to a CPU descriptor, anyone can call their tech "AGI" by just using their chips.
> The name of this CPU is bordering on securities fraud.
No. For it to be securities fraud, Arm would need to make a materially false statement of fact that misleads investors. Naming the CPU in this way doesn't clear the bar because:
a) the name is clearly product brand, similar to how macOS Lion, or Microsoft Windows, or Ford Mustang, or Yves Saint Laurent Black Opium don't mean literally what they say)
b) Arm explicitly defines it as silicon "designed to power the next generation of AI infrastructure", with the technical specs fully disclosed
c) sophisticated investors, the relevant standard for securities fraud, can read a spec sheet
d) Arms' EVP said "We think that the CPU is going to be fundamental to ultimately achieving AGI", framing it as contribution towards AGI, not AGI itself
I was on board with A through C, but then with D it's either clearly a lie or stupidity. I guess it's not a lie technically if they believe it though, so the latter then. But I also don't want to assume someone in their position to be stupid, so then I'm back to the former.
2 replies →
I thought they were adding support for AGI slots
If this headline lead you to believe that ARM has somehow cracked AGI, you deserve to lose your money..
ARM has cracked Agentic AI infrastructure. What are you on about? AGI is a solved problem. The next generation models will have AGI capabilities.
1 reply →
Honestly: The people who buy stock because a product says "AGI" in the name deserve to lose their shirt.
And no, it's not "a lie", because only an utter idiot would consider a product name an actual fact. It's a name. The Hopper GPUs also didn't ship with a lifesize cutout of Grace Hopper.
No, it's actually a lie, and it's different than the Hopper GPU you mentioned.
People have been seeing every big AI company talk about how AGI is the holy grail of AI, and how they're all trying to reach it. Arm naming a chip AGI is clearly meant to make casual observers think they cracked AGI.
The Hopper GPU isn't the same, because Nvidia isn't actively trying to make people think that it includes a lifesize cutout of Grace Hopper. Not a dig on her, but most people don't know who Grace Hopper is, people haven't been hearing on the news for the last several years about how having a Grace Hopper is going to make every job irrelevant.
[dead]
[dead]
If you showed someone what our computers can do with the latest LLMs now to someone 5 years ago they would probably say it sure looks a lot like AGI.
We have to keep defining AGI upwards or nitpick it to show that we haven't achieved it.
I would argue that LLMs are actually smarter than the majority of humans right now. LLMs do not have quite the agency that humans have, but their intelligence is pretty decent.
We don't have clear ASI yet, but we definitely are in a AGI-era.
I think we are missing an ego/motiviations in the AGI and them having self-sufficiency independent of us, but that is just a bit of engineering that would actually make them more dangerous, it isn't really a significant scientific hurdle.
Ok, but it's not AGI. People five years ago would have been wrong. People who don't have all the information are often wrong about things.
ETA:
You updated your comment, which is fine but I wanted to reply to your points.
> I would argue that LLMs are actually smarter than the majority of humans right now. LLMs do not have quite the agency that humans have, but their intelligence is pretty decent.
I would actually argue that they are decidedly not smarter than even dumb humans right now. They're useful but they are glorified text predictors. Yes, they have more individual facts memorized than the average person but that's not the same thing; Wikipedia, even before LLMs also had many more facts than the average person but you wouldn't say that Wikipedia is "smarter" than a human because that doesn't make sense.
Intelligence isn't just about memorizing facts, it's about reasoning. The recent Esolang benchmarks indicate that these LLMs are actually pretty bad at that.
> We don't have clear ASI yet, but we definitely are in a AGI-era.
Nah, not really.
13 replies →
My definition of AGI hasn't changed - it's something that can perform, or learn to perform, any intellectual task that a human can.
5 years ago we thought that language is the be-all and end-all of intelligence and treated it as the most impressive thing humans do. We were wrong. We now have these models that are very good at language, but still very bad at tasks that we wrongly considered prerequisites for language.
9 replies →
> If you showed someone what our computers can do with the latest LLMs now to someone 5 years ago they would probably say it sure looks a lot like AGI.
Would they? Perhaps if you only showed them glossy demos that obscure all the ways in which LLMs fail catastrophically and are very obviously nowhere even close to AGI.
Certainly, they wouldn't expect that an AI able to score 150 on an IQ test is unable to play a casual game of chess because it isn't coherent enough to play without making illegal moves.
2 replies →
It doesn't look anything like AGI and no one who knows what that means would be confused in any era.
Is it useful? Yes. Is it as smart as a person? Not even remotely. It can't even remember things it already was told 5 minutes ago. Sometimes even if they are still in the context window un compacted!
4 replies →
No they aren't
ChatGPT Health failed hilariously bad at just spotting emergencies.
A few weeks ago most of them failed hilariously bad at the question if you should drive or walk to the service station if you want to wash your car
6 replies →
> If you showed someone what our computers can do with the latest LLMs now to someone 5 years ago they would probably say it sure looks a lot like AGI.
But this is a CPU! It's not a GPU / TPU. Even if you think we've achieved AGI, this is not where the matrix multiplication magic happens. It's pure marketing hype.
I did AI back before it was cool and I think we have agi. Imo the whole distinction was from extremely narrow AI to general intelligence. A classifier for engine failure can only do that - a route planner can only do that…
Now we have things I can ask a pretty arbitrary question and they can answer it. Translate, understand nuance (the multitude of ways of parsing sentences, getting sarcasm was an unsolved problem), write code, go and read and find answers elsewhere, use tools… these aren’t one trick ponies.
There are finer points to this where the level of autonomy or learning over time may be important parts to you but to me it was the generality that was the important part. And I think we’re clearly there.
Agi doesn’t have to be human level, and it doesn’t have to be equal to experts in every field all at once.
1 reply →
> LLMs are actually smarter than the majority of humans right now
I consider myself a bit of a misanthrope but this makes me an optimist by comparison.
Even stupid people are waaaaaay smarter than any LLM.
The problem is the continued habit humans have of anthropomorphizing computers that spit out pretty words. It’s like Eliza only prettier. More useful for sure. Still just a computer.
5 replies →
A human can think logically with reason, not to say they are smart or smarter. But LLMs cannot. You can convince a LLM anything is correct and it will believe you. You can't convince a human anything is correct.
I can't argue that LLMs do not know an absolute insane amount of information about everything. But you can't just say LLMs are smarter then most humans. We've already decided that smartness is not about how much data you know, but thinking about that data with logical reasoning. Including the fact it may or may not be true.
I can run a LLM through absolutely incorrect data, and tell it that data is 100% true. Then ask it questions about that data and get those incorrect results as answers. That's not easy to do with humans.
4 replies →
The problem with definitions is that they are all wrong when you try to apply them outside mathematical models. Descriptive terms are more useful than normative ones when you are dealing with the real world. Their meaning naturally evolves when people understand the topic better.
General intelligence, as a description, covers many aspects of intelligence. I would say that the current AIs are almost but not quite generally intelligent. They still have severe deficiencies in learning and long-term memory. As a consequence, they tend to get worse rather than better with experience. To work around those deficiencies, people routinely discard the context and start over with a fresh instance.
AGI wouldn't lie to me every chance it got. Current LLMs are just slop generators, nothing more.
> I would argue that LLMs are actually smarter than the majority of humans right now
This (surprisingly common) view belies a wild misunderstanding of how LLMs work.
"look, it completely lied about params that don't exist in a CLI!"
1 reply →
This is just a Neoverse CPU that Arm will manufacture themselves at TSMC and then sell directly to customers.
It isn't an "AI" CPU. There is nothing AI about it. There is nothing about it that makes it more AI than Graviton, Epyc, Xeon, etc.
This was already revealed in the Qualcomm vs Arm lawsuit a few years ago. Qualcomm accused Arm of planning to sell their CPUs directly instead of just licensing. Arm's CEO at the time denied it. Qualcomm ends up being right.
I wrote a post here on why Arm is doing this and why now: https://news.ycombinator.com/item?id=47032932
This reminds me of Intel talking about faster web browsing with the new Pentium
Ha, I wasn't old or into it enough at the time to remember that, but it is consistent with just about every IC datasheet ever with their list of possible applications. (Like: logic gate; applications include Walkman, Rocket ship, Fuzzy Logic Washing Machine, mobile phone, AGI co-processor, ...)
A lot of this happening.
The Dell marketing machine in particular is bludgeoning everyone that will listen about Dell AI PCs. The implication that folks will miss the boat on AI by not having a piddly NPU in their laptop is silly.
This was exactly my first thought when I saw the title. And after reading the contents of the blog, it's pretty clear that ARM is laser focused on getting a piece of their customer's cake by competing with them. This is likely why they are riding the AI hype train hard with their ill-suited name (AGI).
Unfortunately for them, I think hardware vendors will see past the hype. They'll only buy the platform if it is very competitively priced (i.e., much cheaper) since fortune favours long-lived platforms and organizations like Apple and Qualcomm.
> Arm's CEO at the time denied it
That would have been dumb.
Do you have a source?
It's worse, because there are actually integrated SoCs that include NPU, which I would say are real "AI accelerators".
I think the interesting bit is actually this:
For the first time in our more than 35-year history, Arm is delivering its own silicon products
I can imagine a lot of ARM engineers being frustrated at seeing their cores being used in stupid ways for decades to finally flex what they can do (outside of Apple).
I can imagine many of those ARM engineers looking at Ampere's product line and surmising that an "AGI" ARM server is like building the Hindenburg 2.
2 replies →
Agreed, it will be _very_ interesting to see what waves this causes. It would be like TSMC deciding to make and sell their own CPUs, now ARM is directly competing with some of their clients.
Eh, I'm not so sure it'll be that big a deal. The whole supply chain is so twisted and tangled all the way up and down. Shuffling out one piece doesn't seem like it will, on its own, be so major. Samsung made the chips for the iPhone, then made their own phone, then Apple designed their own chips made by TSMC, now Apple is exploring the possibility of having Samsung make those chips again.
Also, it takes a willful ignorance of history for ARM to claim this is the first time they've manufactured hardware. I mean, maaaaybe, teeeeechnically that's true, but ARM was the Acorn RISC Machine, and Acorn was in the hardware business...at least as much as Apple was for the first iPhone.
2 replies →
But really how different is TSMC than VLSI making the ARM1? By your logic I would say that ARM has already delivered it's own silicon product.
Well technically the ARM1 was a Acorn product (made by VLSI). ARM as a company was only incorporated in 1990 (as a joint venture between Acorn, VLSI and drumroll Apple), I guess that's where the mentioned 35 years and "first time in our history" come from.
1 reply →
Can this be read as finally the financial incentives to join the AI silicone race has become too tempting. Finally the incentives to sell chips are definitely stronger than the cost of competing with your own licensees?
Do they need to higher Design Verification engineers for this?
Thats a huge cost compared to the average RTL jockey
ARM already had tons of DV engineers. No company would license the RTL or any IP unless it has already been run through millions of simulations in DV.
What would be the real advantage of doing that?
AGI = Agentic AI Infrastructure
In case you were thinking about some other abbreviation...
I think this is a poetic encapsulation of the AI industry at this point. A beautifully poignant vignette.
It feels more like "blockchain" to me: https://www.cnbc.com/2017/12/21/long-island-iced-tea-micro-c...
Missed opportunity to call it AAII and market it as twice as powerful as regular AI.
We put AI in our AI so the AI is already baked in.
1 reply →
A^2I^2 or (AI)^2
Is it AGentIc ai infrastructure? Or AGentic aI infrastructure? Or AGentic ai Infrastructure?
I expected better from the people who brought us the ARM architecture, with A, R and M profiles.
It's... really something. Not good. Something.
It’s like they decided to moon all the onlookers while jumping the shark…
I don’t know if it was intentional or they were so far out over their skis that they got their bathing suit caught, but it’s impressive either way.
Should have called it A^3I^2 - Arm Agentic Artificial Intelligence Infrastructure.
I'd throw in an Inference there for the AAAIII symmetry. At a certain point it starts to just look like a scream haha.
The coast is clear to come up with your own expansion for AI!
AGI stands for Artificial General Intelligence.
Pretty sure it stands for "Artificial abbreviation & hype GeneratIon" nowadays
No, it's Agenzia Giornalistica Italiana.
Are you sure it doesn't stand for Advanced Guessing Instrument? That's what the result often seem to indicate after all.
what lenghts are they going to, just to say we have achieved AGI... now who's moving the goalpost?
What a terrible, terrible name.
I mean, they could at least use AI to figure out how to name their AI product.
> I work at ARM, we're launching a new CPU optimized for LLM usage. We're thinking of calling it "Arm Agentic AI Infrastructure CPU", or "Arm AGI CPU" for short. Do you think this is a good idea?
> No. I would not use it as the product name. “AGI CPU” will be read as artificial general intelligence, not “agentic AI infrastructure,” so it invites confusion and sounds hypey.
To bad these executives seemingly don't have access to ChatGPT.
They did ask AI if AGI what a great name. It said that it was the greatest name possible. It's bold, aspirational, and ... polarizing?!
Oh god! Mistral tell me it's highly polarizing, will make the buzz and it's risky but anyway people will know that ARM is doing CPU again now (maybe I did put too many context).
maybe they did and why they got this slop?
Not bait at all
I was thinking "Another Great Illusion".
They pathetically don’t mention what it stands for anywhere in this press release. Deceptive marketing at worst, shameless AI-washing at best.
I would've went for Agentic Neural Infrastructure personally
ARMANI for short /s
AGI (Agentic AI Infrastructure) is joining CSS (Compute Subsystems) in their lineup, apparently. Who’s naming this stuff?
The same people who abbreviate "generative" AI in a way that misleadingly conflates it with "general" AI.
Fraud is just the default lifestyle of marketers.
So Artificial General Intelligence and Cascading Style Sheets are not joining forces?
If there's ever a singularity as a result of AGI, it will likely look at CSS and decide that extermination is simply too good for the human race.
Always have been :)
This is like naming your kid World President Smith.
This could work. Right? https://psycnet.apa.org/record/2002-12744-001
My realtor's last name is House
> Studies 1-5 showed that people are disproportionately likely to live in places whose names resemble their own first or last names (e.g., people named Louis are disproportionately likely to live in St. Louis).
When I lived in Austin, it seemed like a third of boys born were being named Austin. I presume many of them will end up living there as adults but not because of this particular bias, because they were raised there and have family’s there seems to be a more likely driver.
2 replies →
Seems more likely this falls under the replication crisis umbrella. My wife's favorite numbers are my birthday (mm-dd), which is a small reason she fell in love with me. Neither of those numbers are related to her birthday. My favorite number(s) do not overlap with my birthday. Maybe my mm-dd values just aren't low enough, like 02-02?
> Studies 1-5 showed that people are disproportionately likely to live in places whose names resemble their own first or last names
There are several cities in the US that share my last name. I don't live near any of them.
> Study 6 extended this finding to birthday number preferences.
D'oh!
My urologist, and I swear I'm not making this up, has the last name "Wiener".
3 replies →
See also: Nominative determinism in hospital medicine, by orthopedics Limb, Limb, Limb and Limb
https://publishing.rcseng.ac.uk/doi/10.1308/147363515X141345...
Reporting bias.
Meta are heavily invested in building their own chips with ARM to reduce their reliance on Nvidia as everyone is going after their (Nvidia) data center revenues.
This is why Meta acquired a chip startup for this reason [0] months ago.
[0] https://www.reuters.com/business/meta-buy-chip-startup-rivos...
Call this an “AGI CPU” just feels like the most out of touch, terrible marketing possible. Maybe this is unfair but it makes me think ARM as a whole is incompetent just because it is so tasteless.
> Arm has additionally partnered with Supermicro on a liquid-cooled 200kW design capable of housing 336 Arm AGI CPUs for over 45,000 cores.
Also just bad timing on trying to brag about a partnership with Supermicro, after a founder was just indicted on charges of smuggling Nvidia GPUs. Just bizarre to mention them at all.
Arm apparently now sells their own CPU's.
Today everything is AGI.
Yesterday everything was Agentic.
Everything was AI last week.
Waiting for AGI Agentic AI Crypto toilet paper to be in the supermarket shelves , next to the superseded Object oriented UML Rational Rose tuna.
What does "Built for rack-scale agentic efficiency" even means?
If you read past the marketing talk, this is basically a massively multicore system (136) with significantly reduced power usage (300W).
Where does Agentic come into this? ARMs explanation is that future Agentic workloads will be both CPU and GPU bound thus the need for significant CPU efficiency.
Big "but mongodb is web scale" vibes
We just say words now that sound good for marketing but have no real meaning.
> now
I’d argue we have always done that, and in fact it’s basically the definition of marketing!
It's volume of tokens consumed x number of agents x rack space. Basically agentic computation density.
How many angels can dance on the head of a pin?
https://en.wikipedia.org/wiki/How_many_angels_can_dance_on_t...
It's a code sentence for let's go to the utility room to cross pollinate ideas.
Lots of isolated firecracker instances for openclaw like agents.
It's when LLM agents are inefficient that you need a whole rack of servers to get shit done.
I was gonna say just big DCs in marketing yap but really wtf does that mean?
Translation: “Can you give us some money pretty please?”
This can’t come fast enough, I’ll finally be able to use CSS.
What a product name choice! I wasn’t expecting ARM to pivot to selling snake oil.
How fun would it be if due to improved chips handling more model state RAM needs are reduced and Sama cannot make all those RAM purchases he booked?
VC without a degree who has no grasp of hardware engineering failed up when all he had to do was noodle numbers in an Excel sheet.
He is so far behind the hardware scene he thinks its sitting still and RAM requirements will be a nice linear path to AGI. Not if new chips optimized for model streaming crater RAM needs.
Hilarious how last decades software geniuses are being revealed as incompetent finance engineers whose success was all due to ZIRP offering endless runway.
The thing they are good at is bullshitting and selling hype. Which we see here doesn't mean they are actually going to be good at running a business. Smart leaders understand they are not omnipotent and omniscient so they surround themselves who know how to get things done. Weak, narcissist leaders think they're the smartest one in the room and fail.
Unfortunately failing upwards is still somehow common, probably because the skill of parting fools from their money is still valuable.
No, he is also good at networking. When OpenAI was mission-driven and Sam was more respected, he could convince the most talented people to work for him.
Now the talent is going to other places for a variety of reasons, not all due to Sam (one of which is little room for options to grow). However it’s hard to believe his tanking reputation is not badly hurting the company. Other than Jakub and Greg, I believe there are not many top tier people left, those in top positions are there because they are yes-men to Sam.
What RAM? OpenAI booked the silicon wafers, they can print anything they want on them. I wouldn't call them "far behind" on hardware when OpenAI are actively buying Cerebras chips.
Yes exactly; he is behind in that he has to buy others chips with little say on how they work.
Apple and Google control their own designs.
Sama is 100% an outsider, merely a customer. The chip insiders are onto his effort to pivot out of meme-stock hyping, into owning a chunk of their fiefdom. They laughed off his claims a couple years ago as insane VC gibberish (third hand paraphrase from social network in chip and hardware land).
No way he can pivot and print whatever. Relative to hardware industry he is one of those programmers who can say just enough to get an interview but whiffs the code challenge.
He has no idea where the bleeding edge is so he will just release dated designs. Chip IP is a moat.
Plus a bunch of RAM companies would be left hanging; no orders, no wafers. Sama risks being Jimmy Hoffa'd imploding the asset values of other billionaires.
hello,
idk ... so ARM itself is now jumping on the AGI *) "train"... !?
how will they do this!? during system-bootup / splash-screens during execution of whatever we will run on these SOCs ... ;)
*) ads generated income
bruhahaha ... ;^)
just my 0.02€
This looks like an existing pre planned product hastily rebranded AI
For those wanting to know more about software stack,
> Arm is actively collaborating with leading Linux distributions from Canonical, Red Hat, and SUSE to ensure certified support for the production systems.
Taken from
https://developer.arm.com/community/arm-community-blogs/b/se...
6GB/s/core
That's...not much right? Maybe it's a lot times N-cores? But I really hope each individual core isn't limited to that.
Edit: 17 minutes to sum RAM?
It isn't obvious to me that they intended to give this as the maximum single-core performance, or just the proportional share of 844GB/s across 136 cores. Implementations of Neoverse V2 by Nvidia and Amazon hit 20-30GB/s in single-threaded work.
It's a decent amount. Cloudflare was happy to hit 3.2 GB/s/core yesterday. It is shared so cores can burst higher.
If I try to cut through the hype, it seems the main features of this processor, or rather processor + memory controller + system architecture, is < 100 ns for accessing anything in system memory and 6 GB/sec for each of a large-ish number of cores, so a (much?) higher overall bandwidth than what we would see in a comparable Intel x86_64 machine.
Am I right or am I misunderstanding?
It's the same memory bandwidth as Intel and moderately higher than AMD.
Even if you get the 136 cores or whatever?
1 reply →
I miss the all-capitals ARM spelling.
Seeing "Arm AGI" spelled out on a page with an "arm" logo looks slightly cheesy.
But maybe it's actually a good fit for the societal revolution driven by AGI, comparable to the one driven by the DOT.com RevoLut.Ion. (dot com).
Anyways, it sounds like an A.R.M. branded version of the AppleSilicon revolution?
But maybe that's just my shallow categorization.
I also miss the all-capitals ARM spelling. I think they've never been the same since they've changed that, since around the same time their business strategy went from sensible to nonsense.
It's an acronym (like Nasa), not an initialism (like the NSA). I think it might be a British English thing.
2 replies →
On a related note I miss LLaMA spelling.
Oh wow already in use by Meta, OpenAI, and more ?? https://www.arm.com/products/cloud-datacenter/arm-agi-cpu/ec...
The TDP to memory bandwidth& capacity ratio form these blades is in a class of its own, yes?
RISC-V will start making more waves now
Yep, smart people will jump ship since having a competitor control your product is not an amazing idea
Poor TSMC (and ASML)! They were already struggling with capacity to fulfill orders from their established customers. With ARM now joining the party, I don't know how they're going to cope.
Edit: The new CPU will be built with the soon-to-be-former leading edge process of 3nm lithography.
TSMC has multiple fabs being constructed, they'll be okay. The biggest losers here are AMD, Intel and Apple who will be forced to pay AI-hype prices to mass-produce boring consumer hardware.
It only took a quarter century, but I'm glad that somebody is finally adding a little multicore competition since Moore's law began failing in the mid-2000s.
I looked around a bit, and the going rate appears to be about $10,000 per 64 cores, or around $150 per core. Here is an Intel Xeon Platinum 8592+ 64 Core Processor with 61 billion transistors:
https://www.itcreations.com/product/144410
So that's about 500 million transistors per dollar, or 1 billion transistors for $2.
It looks like Arm's 136 core Neoverse V3 has between 150 and 200 billion transistors, so it should cost around $400. Each blade has 2 of those chips, so should be around $800-1000 for compute. It doesn't say how much memory the blades come with, but that's a secondary concern.
Note that this is way too many cores for 1 bus, since by Amdahl's law, more than about 4-8 cores per bus typically results in the remaining cores getting wasted. Real-world performance will be bandwidth-limited, so I would expect a blade to perform about the same as a 16-64 core computer. But that depends on mesh topology, so maybe I'm wrong (AI thinks I might be):
I find all of this to be somewhat exhausting. We're long overdue for modular transputers. I'm envisioning small boards with 4-16 cores between 1-4 GHz and 1-16 GB of memory approaching $100 or less with economies of scale. They would be stackable horizontally and vertically, to easily create clusters with as many cores as one desires. The cluster could appear to the user as an array of separate computers, a single multicore computer running in a unified address space, or various custom configurations. Then libraries could provide APIs to run existing 3D, AI, tensor and similar SIMD code, since it's trivial to run SIMD on MIMD but very challenging to run MIMD on SIMD. This is similar to how we often see Lisp runtimes written in C/C++, but never C/C++ runtimes written in Lisp.
It would have been unthinkable to design such a thing even a year ago, but with the arrival of AI, that seems straightforward, even pedestrian. If this design ever manifests, I do wonder how hard it would be to get into a fab. It's a chicken and egg problem, because people can't imagine a world that isn't compute-bound, just like they couldn't imagine a world after the arrival of AI.
Edit: https://news.ycombinator.com/item?id=47506641 has Arm AGI specs. Looks like it has DDR5-8800 (12x DDR5 channels) so that's just under 12 cores per bus, which actually aligns well with Amdahl's law. Maybe Arm is building the transputer I always wanted. I just wish prices were an order of magnitude lower so that we could actually play around with this stuff.
Amdahl's law is about the maximum speedup obtainable from parallelism, not balancing memory bandwidth with compute.
AGI will just become the new "Smart Phone" or "Smart Car" losing all meaning.
Many of these words are unexplained. "Memory and I/O on the same die". Oh? What does this mean? All of the DRAM in the photo/render is still on sticks. Do they mean the memory controller? Or is there an embedded DRAM component?
All processors have memory on the same die.
How much, what kind, and what is your source?
All mainstream server CPUs have a megabyte or two of SRAM on a core, of course.
1 reply →
What is “agentic AI cloud era” referring to? I honestly don’t know what this buzz-speak is targeting. Running models locally on the server, for cloud workloads? Agentic, that is just a LLM pattern.
Don't overthink it. Shut up and buy some ARM stock.
And the stock is down >2% today
Hmm all my experience with using AI has been mostly VRAM. I haven't experienced any bottleneck on the CPU side. What does this chip offer over Intel or Apple Silicon? Anyone expert here know whatit is?
The arm family of chips (apple A series, m series, and qcom snapdragon) are better on energy usage (thus battery life) and performance and design compared to many x86 style chips (intel, amd).
Time will tell if ARMs owncpu is on par or better than Apple’s ARM based chips
Interesting that Jensen Huang joined in the congratulations for this new product!
More AI bullshit and hype is good for Nvidia. Until it isn't.
I no longer believe this is like the dotcom. Now it feels like the 1983 video game crash.
I was wondering who convinced ARM to manufacture hardware. Turns out it was Meta.
Now if only they would go back to being "Acorn RISC Machines" and make a nice desktop home computer again...
One can dream.
DGX Spark is pretty nice. It could be cheaper if they removed the NIC though.
1 reply →
I hate RISC OS architecturally, but if they made a new Archimedes or whatever that ran it I'd buy it
Fabless. Like AMD and Nvidia. So I would think about it more as branding and packaging than Manufacturing
Huh, many companies use TSMC, in fact, probably all of them use TSMC, including Intel, yet there are only a few who dominates in performance. There are much more in designing chips than what you just listed.
1 reply →
There's a big difference between just providing IP and actually doing the physical design, manufacturing and packaging. You can't just send your RTL to TSMC and magically get packaged chips back.
I haven't ever ordered an ARM SoC but I also wouldn't be surprised if there were significant parts that they left up to integrators before - PLLs, pads, SRAM etc.
Nuvia/Qualcomm lawsuit and Softbank.
The non marketing fluff version of the press release can be found here: https://news.ycombinator.com/item?id=47506641
ARM must be feeling the heat from all those RISC-V AI startups.
"AGI" continues to lose all meaning.
ARM riding the "everything is AI" train.
So sad.
Yeah dumb name, but we will still use these we have been using Ampere in our office.
is this a cpu that's meant for AI training or is it more for serving inference? I don't quite get why I would want to buy an arm CPU over a nvidia GPU for ai applications.
It is for orchastrating inference/creating firecracker instances for agents etc. It does'nt have anything to do with actual AI usage.
Interesting thanks
finally, a CPU capable of making API calls to cloud providers
Agl? @gi? Heck if we can’t compete we’ll confuse!
"The I is for IPO" :D
Seems like hubris to use this name.
Now every product will have the AI buzzword in it's name, just like 25 years ago product names started with letter e, from electronic.
So we will see AI Toilet Paper launching in the next months.
Why not ASI? They aim too low.
If rich people are this stupid then they deserve to be parted with their cash
arm what we want is an arm chip that can rival m-series not this
We have C1 Ultra at home.
I feel like this is one of the things that people will look back on as the peaking of the bubble.
Like, c’mon, this is ridiculous.
Why?
[dead]
[flagged]
[dead]
[dead]
[dead]
Well that explains it, the guy in charge is a wad.
I found this article extremely frustrating to read. Maybe I lack some required prior knowledge and I am not the target audience for this.
> built on the Arm Neoverse platform
What the heck is "Arm Neoverse"? No explanation given, link leads to website in Chinese. Using Firefox translating tool doesn't help much:
> Arm Neoverse delivers the best performance from the cloud to the edge
What? This is just a pile of buzzwords, it doesn't mean anything.
The article doesn't seem to contain any information on how much it costs or any performance benchmarks to compare it with other CPUs. It's all just marketing slop, basically.
> The ARM Neoverse is a group of 64-bit ARM processor cores licensed by Arm Holdings. The cores are intended for datacenter, edge computing, and high-performance computing use. The group consists of ARM Neoverse V-Series, ARM Neoverse N-Series, and ARM Neoverse E-Series.
https://en.wikipedia.org/wiki/ARM_Neoverse
More precisely, this Neoverse V3 core is the server version of the Cortex-X4 core from smartphones. The actual core is pretty much identical, but the cache memories and the interfaces between cores are different.
Neoverse V3 is also used in AWS Graviton5 and in several NVIDIA products.
I feel like this is most products in the AI space lately. More marketing fuzz than substance. Hard to figure out what thing even does.
You should look at the benchmarks of the Cortex-X4 cores used in many smartphones from 2 years ago, because it is the same core as Neoverse V3.
AWS Graviton5 uses the same cores, but it has 192 cores per socket.
So Graviton5 has more cores per socket, but I think that it does not support dual socket boards.
This Arm AGI supports dual socket boards, so it provides 272 cores per board, more than Graviton5 MBs.
However, this is puny in comparison with Intel Clearwater Forest, which provides 576 cores per board, and the Intel Darkmont cores are almost exactly equivalent for all characteristics with Arm Neoverse V3.
Thanks for this article. Very important news.