How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip
spectrum.ieee.org117 points by maxall4 10 hours ago
117 points by maxall4 10 hours ago
Having worked with people doing bringup of specialized chips, I am awed at how the world has changed.
> When the first chips came back from the foundry in May, the team pointed its internal AI models at designing software to run benchmarks such as SemiAnalysis’s InferenceX. On DeepSeek’s multi-head latent attention kernel benchmark, performance climbed from 0.31 percent of the theoretical ceiling (set by the chip’s compute and memory bandwidth) to 88.94 percent in roughly 40 hours. Ho says this result is repeatable, so the time between when foundries deliver the first chips and when production ramps up can be reduced. “All our schedule assumptions are going to be based on the fact we have this capability now,” he says.
Back in the day you'd write the code before the chip came back but I guess today it's faster to wait.
The longer you wait the faster you will go.
Like space travel.
The successive generations of spaceships won't built themselves. Who will be responsible for setting up real world and software feedback loop?
> “All our schedule assumptions are going to be based on the fact we have this capability now”
is the world we live in, planning things while waiting for a more powerful LLM
With a few handy tips and tricks from apple insiders. But sure, I guess the LLMs helped too.
The lawsuit isn't for chip designers, but the consumer product lines. It's possible they also got IP for chips, but that was not brought up.
This is the part forgotten. Apple is claiming this is their IP embedded on chips that OPenAI stakes the future on. Will they settle?
The lawsuit appears to be about consumer devices, not NPUs or ASICs. If you think Jalapeno stole from anyone it would be Google.
You people are so weird - gossiping on a site that literally reported the news as it broke smh. It's like sitting in the back of class gossiping about the popular kids.
Newsflash that lawsuit is about product designs not accelerator ASICs. And there wouldn't be anything to steal because Apple doesn't have any DC class accelerators.
Yeah, how could those people even think of switching their owners!
Didn’t they not just switch employers but actually handed over a lot of proprietary documents from Apple?
Seems pretty obvious now that OpenAI is just hyping their models in order to get companies (in this case, chip developers) to use their products in order to learn from their (exfiltrated) IP. Any corporation would be foolish to use any of their or Microsoft’s products, particularly those with valuable IP. There’s nothing in the article that says AI did anything creative but rather that it was used for software development within the overall project. Clear misleading title. Suggest to mark this as clickbait.
IEEE Spectrum is such a good publication. Early in my career I worked at a place where the magazine would be passed around every month with a coversheet listing all us engineers we had to pass it around and sign we had read it. Been a while since I visited the website but love what they did with it.
Every time their content appears here, it's a very shallow analysis written for a barely technical audience. And this article is no different, it's just "slop machine wrote verilog; all the hard bits were done by Broadcom, who have access to public AI models (we didn't talk to them and don't know if they used them, but ClosedAI wants us to think they did)"
I grow Jalapeños. This conflation of AI and actual chili peppers irks me.
Guess how electrical engineers feel about the term "transformers".
> Guess how electrical engineers feel about the term "transformers".
There is more to this story than meets the eye.
I do generative art (no relation to AI prompting). I feel your frustration.
Cryptographers also got the same raw deal with cryptocurrency, and every one just said "crypto?"
Like traditional generative art? Like worms WMD map generation or something? Cool!
Just think of how the people of Xalapa, Mexico feel. They should send them a royalty check.
> irks me
It's jalapeño grill would you say?
Not sure what you're talking about. But I'll tell you, home grown Jalapeño peppers, fermented with 3% salt is the stuff of dreams.
I'm annoyed that the meaning of the word "Agent" has been obliterated.
Like, why couldn't they invent a new word and not hijack an existing word?
https://en.wikipedia.org/wiki/Agent
Computing
* Agent architecture, a blueprint for software agents and control systems
* Agent-based model, a computational model for simulating the actions and interactions of individuals
* Agentic AI, autonomous artificial intelligence that can make decisions and act on those decisions on its own
* Forté Agent, an email and Usenet news client
* Intelligent agent, an autonomous, goal-directed entity which observes and acts upon an environment
* Software agent, a piece of software that acts for a user or other program
* User agent, software that is acting on behalf of a user
Aw, I was expecting more details but this just seems to be a rehash of what they unveiled a month ago.
At some point people will use an LLM to design an Apple M series competitor.
Production grade CPU design is more than just the RTL (the source code.) To achieve the performance numbers that these companies get, you have to do a ton of optimization in your physical design to achieve the power/performance/area (PPA) metrics that make these products competitive. LLMs are not suitable for that kind of work.
There are people working on PPA optimization and trying to shake up how things are done, just not with LLMs.
> LLMs are not suitable for that kind of work.
I wonder why not or you meant not suitable yet?
This is just speculation on my part, but LLMs work best when they get immediate, verifiable feedback on their task, and the kind of physical optimizations they mean might not give that to LLMs.
Also a speculation but I'm almost certain that physical optimizations are first done through simulators running on a computer.
Something that I think is fascinating, though, is that labs are no longer beholden to the limitations of commercial design software. Want to replace your simulator and optimizer with a fully custom verifiable stack of Lean proofs of optimality and correctness? Just throw your unlimited token budget at it.
I don't work in the business, but my understanding was that even with these companies' budgets, it's still too expensive to do any kind of verified performance optimality.
Isn't that weird? The full knowledge of how to make such chips may one day be accessible to anyone, yet only the entrenched companies will remain the makers.
If we imagine machines being able to do the full process end-to-end, and the quality of that process only dependent on capital spent on tokens, I don't see how new companies could ever enter the market.
It is probably doable right not to push a risc-v design into that performance space.
They won't, because they'd need an ARM architecture license.
Why, the LLM can make up its own architecture.
The value lies in the design space exploration, which is what an LLM can easily do.
Qualcomm have an architecture license and the Snapdragon X2 Elite Extreme X2E-96-100 isn't too far off the M5 Pro.
[1] https://browser.geekbench.com/processors/snapdragon-x2-elite...
[2] https://browser.geekbench.com/macs/macbook-pro-14-inch-2026-...
Are they using LLMs to close that gap, or is this their Nuvia acquisition doing the heavy lifting?
Or they'll just build a competitor in RISC-V instead and that's fine.
Except the problem is not restricted to the actual ISA or its HDL implementation, etc.
It's even just getting space / time in a fab at that advanced of a process node.
I mean you can design anything without a license. Selling it is where the problems come up. Even then there are likely places in China that would still make it for you.
And be super-bankrupted by patent litigation from Apple. I don't think they're worried.
After all, they successfully threatened Adobe with spurious patent litigation unless they joined w/ apple in illegally fixing wages.
You don't think a criminal like apple would absolutely decimate any competition given the opportunity? They didn't hold back when it was a unambiguous crime, they surely wouldn't if it was merely bad for the world.
We were able to invent a chip that already existed so fast, you guys.
AI does not make anything new, it is not surprising that it can regurgitate what already exists much faster than humans can invent new things.
It is surprising to me that recursive self-improvement seems more plausible now than it did in 2023. Am I the only one to be surprised?
I remember the paper proving that hallucinations could never be fully solved back in 2024: https://arxiv.org/abs/2409.05746
I also remember the hang-wringing about running out of new datasets to train on. Now it appears humans are always generating more data. It's just not as cheap to acquire as legacy data? Meta has to give a deep discount on their API prices to entice people.
I thought back then that humans had a few more breakthroughs in them as meaningful as the seminal Attention is all you need paper. Enough to 100x the capabilities of LLMs back then (10x the smarts and 10x the speed simultaneously).
RSI with a 20 month turnaround for a chip to be made is not exactly breakneck speed though. Physical manufacturing and logistical constraints are going to be and remain a hard obstacle to that process for the foreseeable future.
> I remember the paper proving that hallucinations could never be fully solved back in 2024
The papers that use the halting problem or the Gödel's incompleteness theorem to prove something about LLMs are dime a dozen. The problem is they prove their results for any computable system. You need to also believe that the human brain contains "magic" to think that humans are exempt.
I believe I've said the same at the time this paper was published. There is no need for hindsight to notice the problem.
The required amount of compute and training data and whether the existing training methods were up to the task had the real potential to be show stoppers though.
> Jalapeño can reduce end-to-end latency (the time between prompt to last token) by up to 3.6 times
I'm never sure what on earth this kind of impressionistic math is supposed to tell me. Is the comparison between 4.6 and 1.0? 3.6 and 1.0? Clearly the comparison isn't supposed to be 1.0 and -2.6, even though that's what the words literally mean. I can't be the only person who finds this infuriating and distracting. These numbers shouldn't be impressionistic. They should be precise. That this is an article on spectrum.ieee.org makes the imprecision all the stranger. I'd expect their readershipt to care, for instance, about what's even being measured. Is this the geometric mean of something? The arithmetic mean? And what latency has improved?
That odor you are detecting is just good old fashioned bullshit, my friend. It’s just that nowadays everything and everyone is covered in it, and we are not supposed to notice. The emperor has no clothes… and is covered in shit.
It's... written right there? Like what?
Suppose you send in your marvelous prompt and hit Enter.
Machine churns for 18 seconds, types out a "reply", then yields back control.
18 / 3.6 = 5
So now the machine will only churn for 5 seconds before yielding back control.
This is confusing how exactly?
Why would an "up to" figure be a mean, or a geometric mean? It's clearly a max, that's why it's called "up to"...
Am I missing something?
So when can we start getting cheap chips? RAM anyone please!
Everyone's still bottlenecked on foundries, not designs.
Congrats to the former TPU team
I was surprised to see they were using XLS but then I remembered Chris went there a couple of years ago.
openai should figure out how to make lithography machine, so ASML don't have monopoly on it.