How many billions of transistors in an iPhone processor?
lemire.me
lemire.me
I've been reading datasheets a lot at work. The datasheet for an STM32F4 is 149 pages long. The technical manual is over 800 pages long. That puts it into perspective a little bit better for me. I might spend days or weeks writing a driver for some little few pages of that datasheet. It's a dense 800 pages. Those manuals indicate just a fraction of the complexity that lies inside of the device -- they describe features that I can use, not even how those features are implemented or made. And that's for a tiny chip that costs a couple dollars.
A billion is a tough number to understand, but I recommend reading a datasheet of a chip sometime. It will put into perspective how much complexity there is in just a few millions of transistors.
I sometimes have to take a moment, because I remember descriptions of alien technology from sci-fi books. Complexity, opaqueness, power. Chips.
Intel® 64 and IA-32 Architectures Software Developer’s Manual Combined Volumes: 1, 2A, 2B, 2C, 2D, 3A, 3B, 3C, 3D, and 4
[0] https://www.intel.com/content/www/us/en/developer/articles/t...
It's because they were originally 3 volumes. volume 2 is the ASM reference, 3 is the "how to do things" reference (System programming guide). Volume 1 is more of a general reference so maybe that's why it didn't grow so much.
I do miss working directly with small, simple (relatively), cheap chips. I don't miss dealing directly with the low level bus protocols though. I still have stress nightmares about diagnosing can bus and i2c bus problems with an oscilloscope. Like in the stress dream I'll either forget some important details about the protocol and not understand what is going on at all, or I will remember everything in perfect detail but be unable to speak and explain what's wrong to the other person who's there working on the issue.
There's no way in hell people can make sense of 19 billion separate concepts. Likely what's going on in the chip is just repeated patterns designed to make a very simple machine operate really really quickly.
But of course I'm not denying the complexity. But the 19 billion implication here is a bit too much. It's likely just as complex as say linux.
You are right about human designed chips. If you use something like simulated evolution or backpropagation or some other techniques from machine learning or mathematical optimization, you can get machines with billions of parts but not necessarily any clear pattern nor repetition.
The human brain might be an example of such a mechanism.
We've done it artificially as well with machine learning and genetic algorithms.
The amount of concepts in the neural network of an LLM is too large for us to fully understand.
At best we can only hope to understand these things as a sort of curve fitting optimization problem.
But you see how this is far from true understanding. It's like removing the WiFi module from a PC and you lose WiFi, therefore you know the WiFi module does WiFi stuff but in the end this action contributes nothing towards cohesive understanding of the PC.
And even then:
1) There's parts of the manual which are brief summaries, written with the assumption you're already familiar with the larger standards they interface with. The SDIO and USBOTG sections of the reference manual come to mind in particular, not to mention all the external manuals it takes to describe the ARM Cortex-M architecture.
2) That's a fairly simple microcontroller. (Given the page count, it sounds like you're working with STM32F401.) The STM32H7 series reference manual is 3500+ pages.
you should do some coding closer to the chip
some bitshifting, some bit level manipulation
you dont have to be good at assembly, just do a hello world and make a mental visualization of the registers being hit
look up the OPCODES in your specific chipset , make some code to hit that specifically
make a process that does it as fast as possible and benchmark it, do it faster or slower on a different chipset, or a resource constrained environment
it makes it all really easy to follow and visualize
I dont think you need to write a driver for that map
Sufficiently advanced technology is indistinguishable from magic.
I love using the money analogy for this:
"If you earned $1 every single second, that's $60 per minute, $3,600 per hour, $86,400 per day. More than most people make in a year!
Assuming no tax or expenditure whatsoever, it would take only 11 days to become a millionaire at that salary.
It would take over 31 _years_ to become a billionaire.
And that's only a mere $1 billion, some people are worth _hundreds_ of billions."
Also I don't really have an idea, but I wouldn't be surprised, if a very significant part of transistor budget has been spent on SRAM or something similar, which is not very fascinating.
i tried to read an Espressif ESP32 manual, blew my fucking mind on how complex it is relative to how cheap it is.
The GameBoy CPU is simple and easy to understand, sure, but modern ones are built upon the same basic principles, it's just that they run at much higher clock speeds and add lots and lots of parts (caches, pipelining, speculative execution, register renaming, multiple ALUs, SIMD ...) to execute as many instructions per clock cycle as possible. Also virtual memory support — MMU, TLB, all that stuff.
For the other side, if you want to build a CPU out of logic gates, there was a web-based game where you did exactly that. I don't remember the URL, sorry.
That’s not true for any post 1995 desktop CPU.
They just adhere to that model and pretend that that’s all they’re doing but in reality there’s layers of supercomputers inside the CPU: micro-op cache, branch predictor, reorder buffer, rename buffer.
They extract hidden parallelism out of the code, run it in blazing fast multiple instructions per clock cycle (!), and pretend that nothing out of the order happened.
CPUs are extremely complex and extremely smart. They can have errors in their implementation where the unseen branches that it executed can be accessed by some processes, etc. But there’s nothing dumb about CPUs. They are made of a million man-hours of engineering.
What's the difference between a logic gate and when my index and thumb have a static charge jump between them? About a couple trillion dollars of research.
That seems to me a highly reductive point of view. The whole is always more than the sum of its parts.
This is the one I played some time ago: https://nandgame.com/
https://news.gallup.com/poll/505745/depression-rates-reach-n...
I don’t post that to be negative, because I agree with what you are saying about a unique time to be alive and have done the same Atari to Xbox Series X glow-up and I marvel at the improvements there. But I do think happiness (and decreased mortality, increased lifespan) should be the ultimate end goal we are striving for for humanity and we seem to be failing, at least in the USA. The reasons are certainly multifaceted, but increased transistor density may be one of the reasons just like obesity, environmental toxins, etc.
Are you arguing that we are less secure against nature while sitting in climate controlled buildings with food stores in them, and a global industrial complex raised to keep these trains running?
But you're right: I'm not going to get lost in my own city - or yours - because I have an iPhone in my pocket. I may never really learn my city and all its hidden little secrets you find only by wandering around, but I'll be efficient getting to the bar in a "climate controlled buildings" where I'll just stare at the iPhone some more and not really talk to anyone, as that's what everyone is doing, too.
"It is a world suited for monomaniacs obsessed with the idea of progress - but a false progress, a progress which stinks. It is a world cluttered with useless objects which men and women, in order to be exploited and degraded, are taught to regard as useful ... Whatever does not lend itself to being bought or sold ... is debarred" –Henry Miller, The Air-Conditioned Nightmare
Sometimes I wonder if internet people are living in parallel universe. I see literally opposite, wherever I go, of what people like you describe.
Thanks to technology, we don’t have to try and have eight kids—losing half of them at a young age—just so that we don’t lose the farm and starve.
Thanks to technology, millions of children are protected from diseases that years ago they would’ve died from.
But sure, blame technology for your lack of self control causing you to stare at your phone at bars.
I think this is valid for most people. But as I've tried to integrate more gratitude into my life, I've noticed that it is possible to feel happiness from these things. I find it all too easy to disregard many of the things I have because they've always been there and available. But as soon as I don't have some of my comforts (being outside in the cold, camping and batteries are dying, my car needs an unexpected repair, a pain in my arm that I can't treat myself, etc.), it becomes easier to realize just now nice things really are.
One possible way of getting at this is to ask if people were more ‘happy’ when there wasn’t this level of tech available - and I’m not sure the answer is yes - I mean, there’s the mythos of ‘a simple life’ and ‘noble savages’, but from what I’ve read of life even a 100 years back - I would surmise folks were as unhappy / miserable then too - but likely for different reasons.
I think expecting just tech to make us happy is framing the wrong problem space.
so technological progress in itself, regardless of its use, is a consideration
From what available data would you derive this and not immediately question selection bias?
Is it? Because that is not clear to me in the slightest.
For instance I’m sure that a major factor is the need for people to move great distances away from family and the community they grew up in to be able to find education, living wages, and other forms of opportunity. That breaks existing connections and makes it harder to form new ones, but I never see that talked about.
So maintain connection has never been easier, and migration is not any worse. I have no doubt there are far more things than social media making people more unhappy than historically, but having to move away from where we were born isn't it.
Instead it’s often necessary to move repeatedly, for reasons varying from going to school to getting a job to finding housing with room for kids to just trying to keep rent from eating up the majority of one’s paycheck.
In my case I’ve moved 9 times in the ~15 years since I turned 20 and it’s very likely I’ll be moving again in the future. It’s been very disruptive for maintaining connections, even with the power of modern communications at my disposal. I’ve more or less grown used to it and am kind of introverted anyway so I’m not depressed but I could see where others might not be so lucky.
It was simply not spoken about, people either poisoned their husband if it got that bad, or lived a sad life in silence.
A washing machine is a fantastic piece of machinery you could even build on a small scale that saves hard labor
A patent by Sony that forces the TV viewer to stand up and lift their hands and say "McDonalds!" to continue watching TV, is not
I think you could apply a similar theory to the modern capitalist world.
Having to choose between one or the other would suck. I wonder if we can have both?
I would like to see some data comparing modern, fit worker against average hunter gatherer.
There are happy monks and unhappy billionaires.
Humanity must push forward on all fronts.
That's why it's important to understand context, both local and historical, to understand our place in the world and how we should feel about it.
https://www.wired.com/story/i-saw-the-face-of-god-in-a-tsmc-...
The Apollo guidance computers had "several tens of thousands of transistors"[0], we now have 19 billion in our pocket!
0: https://www.aei.org/technology-and-innovation/apollo-11-mank....
The thing about the series though was that NASA was self funded and didn’t have to justify spending billions in taxpayer money.
Also, do we really have better propulsion technology to make space travel more efficient?
Look after the people on Earth first. Once we show we can accomplish that, then and only then do we begin to earn the stars.
Curiosity has driven exploration, discovery, and subsequently science. It's fine that you're not as curious as some, but suggesting that we must only focus on things on Earth when we have literally billions of people on earth that all have individual interests and things they're curious about is just about limiting progress, not about preserving Earth.
If not, then we've established a mutually acceptable baseline for limits on human progress, and a subset of mutually acceptable reasons for imposing such limits. I suggest only that we need not stop there.
> How on Earth do you support the claim that one deliberately self-imposed limit on human progress can't be compared with another?
You're being obtuse. There's a much more clearly defined intersection of human ethics when it comes to performing science on human beings than there is with some sort of vague proposition that space exploration is a waste of resources that is harming humanity indirectly by not allocating those resources towards taking care of other human beings.
I don't think it has less meaning than the idea that the most sere and hostile environment for our species that we've yet been able to access - the deadliest imaginable desert, accessible only at vast effort and expense - has for our taking the boundless cornucopia of riches that space maximalists constantly promise, and with no more support than you complain I offer.
Meanwhile you blithely argue that some significant fraction of a trillion dollars, available in this happy counterfactual from what had been the war department's budget, should be devoted to this and nothing else, with no thought to the opportunity cost - and demand I justify myself in arguing we should better do otherwise!
That opportunity cost would be measured in human lives, consigned to neglectful death in the service of your cause - but whether by neglect or murder, dead is dead. You'd be far from the first so to act; history is replete with examples, even ignoring the one we live among now. Such people are sometimes called to account. What would you say if you were?
We've already benefited greatly from just the process of trying to get there. Even if there is no "riches" to be had in terms of extracting resources from alien planets, we still gain from our endeavors. This can be indirectly through technological advancements made in service to that mission (which we've seen plenty of already), or it can simply be in the pursuit of knowledge in the understanding of our universe.
> Meanwhile you blithely argue that some significant fraction of a trillion dollars, available in this happy counterfactual from what had been the war department's budget, should be devoted to this and nothing else, with no thought to the opportunity cost - and demand I justify myself in arguing we should better do otherwise!
Why do you assume that there is not thought to the opportunity cost? It's quite clear that we have different priorities on how humanity should allocate its resources but to say that it's not a consideration is just crass. The amount of human productivity we dedicate towards space exploration is a remarkably small percent of what's available, and a lot of that is driven by individual's interest in it.
The space program's budget is minuscale compared to what we spend elsewhere, maybe around 5% of a trillion dollars. That's a significant amount of money, but you seem to think that could be better spent on... what, exactly? Let's say we reallocate that budget to social security. That'd be adding less than 4% to its budget, which is already over a trillion dollars. Or the combined spend of medicaid and medicare, in which case it would be less than 3%.
> That opportunity cost would be measured in human lives, consigned to neglectful death in the service of your cause
A lot of things have indirect consequences, positive and negative. Many things that have dramatically improved the livelihoods of humanity have come at a great cost.
> but whether by neglect or murder, dead is dead. You'd be far from the first so to act; history is replete with examples, even ignoring the one we live among now. Such people are sometimes called to account. What would you say if you were?
The advancements made by curious people has made the world less impoverished and less "dead." Even if 100% of human productivity was allocated towards protecting human life, people will still die. That is life. At what point in your hypothetical world where you got to make these policies would it be acceptable for humanity to start exploring space again? When humans stop dying?
Humanity's driven by curious people willing to explore well beyond what seems rational to most people, not myopic luddites who would eschew entire avenues of exploration simply because they can't understand the utility in it.
> At what point in your hypothetical world where you got to make these policies would it be acceptable for humanity to start exploring space again?
When there is no human on Earth who suffers in a way that money can prevent, I'll no longer consider a dollar spent on putting a human into space a dollar culpably wasted.
> Humanity's driven by curious people willing to explore well beyond what seems rational to most people, not myopic luddites who would eschew entire avenues of exploration simply because they can't understand the utility in it.
I grew up suffused in the US space program. Huntsville was my favorite vacation. I read everything I could get my hands on - which, thanks to my grandfather, was a lot - I dreamed of astronaut school, at least until I learned it had visual acuity requirements. But even that little dampened the dream; I still spent most of my childhood studying with the intent of becoming an engineer and ultimately joining NASA. There are a lot of reasons that didn't happen, but none of them had to do with the dream itself failing - and in some ways, it still hasn't; do you think, if I still could believe there was anything for us either to find or to offer out there, I wouldn't want my people to go to the stars?
If, even in the face of all that, I still can be unconvinced by the case for putting humans as we are now into space, then that case needs to be better made. But dreamer though I was, the thing I came eventually to understand about a dream is that, however beautiful it may be, eventually you wake up and find reality still there.
As far as "human curiosity beyond rationality" goes, I would have been much more sympathetic to that argument before it became inescapably obvious that in its pursuit we have quite literally set the world on fire.
In the face of an error on that scale, I believe only a fool would fail to adjust his perspective, at least to encompass the possibility that, in the argot of my youth, we have gotten quite a bit too big for our breeches.
That we should reach beyond our grasp is something even I consider creditable in the species. That we should reach into the gearbox of the world and have it tear off a couple fingers, and then want to go right on reaching without a thought at least to stanch the bleeding, at some point becomes self-destructive madness. An individual human displaying such behavior would be restrained for his own good. As a species, we would have to do that for ourselves, or reality will do it for us. It is doing exactly that now. We would be wise to take that into account in deciding our next actions. We won't. But we could, and we should.
Humans aren't going anywhere any time soon, and I see nothing meaningful to choose between Mars today and Mars a hundred years from now, or two hundred, or a thousand. In that meantime, our efforts could make a meaningful difference as to how many billion humans are alive to celebrate that happy day.
You really want to convince me that space boosterism might still, in spite of everything, have merit? Show me you can take the long view.
The case is fine; the problem is your ignorance to seemingly many factors.
> As far as "human curiosity beyond rationality" I would have been much more sympathetic to that argument before it became inescapably obvious that in its pursuit we have quite literally set the world on fire.
Well, one of humanity's first great advancements was, in fact, to create fire.
We are making tremendous efforts in many different fields, but it's becoming abundantly clear that the only thing you actually care about is global warming, which, unsurprisingly, is something that we have people working on and hundreds of billions of dollars being spent annually on to solve. Oh, and wouldn't you know it, the space program's existence continues to contribute to climate science with earth-observing satellites [1] that provide all sorts of data that would only be possible from space. Thanks, space program!
> Humans aren't going anywhere any time soon, and I see nothing meaningful to choose between Mars today and Mars a hundred years from now, or two hundred, or a thousand. In that meantime, our efforts could make a meaningful difference as to how many billion humans are alive to celebrate that happy day.
Once again, you've not provided any rationale or reason as to why we can't work on more than one thing at once, or why the space program in particular has earned your ire. You can look up the dozens of advancements we've made directly thanks to our endeavors in heading to space. In fact i've already given you one very important example as to how it's relevant to your interests. You choosing to remain ignorant about the rest is your own choice, but it's quite clear that you're too ignorant to be making any value judgments on the space program at large.
Why not just admit that your argument is purely ideological? Or at least make an argument for what that money (that you don't seem to know even how much is being spent) should be spent on and how much further it could go in that area? The space program is not in ideological opposition with solving climate change; in fact it's playing a critical role in it.
> You really want to convince me that space boosterism might still, in spite of everything, have merit? Show me you can take the long view.
This is the long view; immense progress can happen from random discovery or it can happen slowly and gradually over time. We pursue these avenues because we don't know where it will take us or how long, just that we pursue these curiosities because it is what humanity was made for. If we only had 100 people on earth then sure, maybe it wouldn't be worth looking to the stars. But we don't have only 100 people.
Of course it is! And so is yours.
If not from ideology, where come you by the idea that "we pursue these curiosities because it is what humanity was made for?" That statement has zero factual content and can be meaningfully parsed only as a statement of belief. How do you mean me to imagine it as not ideological in origin?
I would, though, prefer it if we could stay on the same page with regard to what we're discussing. At no point for example have I said I have a problem with placing satellites in Earth orbit, and not by accident have I been specific in phrases like "putting humans in space". Robots in space, serving our ends, are cheap, easy, durable, and worth the trouble. Meat in space, at vastly greater expense and fragile complexity, is not.
Certainly we can do more than one thing at once, too. But every thing we do has an opportunity cost. My point is simply that the opportunity cost of putting humans in space has yet to be justified by any result - very notably including any output of the Space Shuttle program, where my opinion largely tracks Cegłowski's wellknown 2005 review [1], albeit perhaps with a touch more charity - and with no realistic prospect of that changing, I see no compelling reason to throw any more good money after such a vast sea of bad.
I don't need an ideological justification to allow people to do things that they want to. I'm not the one looking to reduce people's rights to pursue their interests; you are. That's why there is a distinction between my position and yours.
> At no point for example have I said I have a problem with placing satellites in Earth orbit, and not by accident have I been specific in phrases like "putting humans in space".
> > Look after the people on Earth first. Once we show we can accomplish that, then and only then do we begin to earn the stars.
I think a reasonable interpretation of this is that you do not think humanity has not "earned the stars." It would not reasonably be interpreted as only referring to sending humans into space.
> Robots in space, serving our ends, are cheap, easy, durable, and worth the trouble. Meat in space, at vastly greater expense and fragile complexity, is not.
Why not? What is the criteria for it being worthwhile?
> My point is simply that the opportunity cost of putting humans in space has yet to be justified by any result - very notably including any output of the Space Shuttle program, where my opinion largely tracks Cegłowski's wellknown 2005 review [1], albeit perhaps with a touch more charity - and with no realistic prospect of that changing, I see no compelling reason to throw any more good money after such a vast sea of bad.
Putting humans into space is a remarkably new thing that has largely been interrupted by its lack of funding. What are you even comparing this against? What returns would be required for it to satisfy your need for an ROI?
But for now, we have to just observe how countries enact progress, and it does seem to be that the ones that delegate and diversify do the best at it.
HN is one of the most reliable sources of those that I know, and developing a thread in this way is one of the more reliable methods I know to improve my odds of eliciting one. Luckily for me, I have enough of a prickly, disputatious streak in my own nature to appreciate the process quite as much as the result.
To get more explicit then: what did you mean to imply with this?
> It's rare I find myself reaching for the adjective "Panglossian".
2) There's also this graph, if you want to see how the other income brackets are changing: https://en.wikipedia.org/wiki/Extreme_poverty#/media/File:Th...
A rat done bit my sister Nell. (with Whitey on the moon)
Her face and arms began to swell. (and Whitey's on the moon)
I can't pay no doctor bill. (but Whitey's on the moon)
Ten years from now I'll be payin' still. (while Whitey's on the moon)
The man jus' upped my rent las' night. ('cause Whitey's on the moon)
No hot water, no toilets, no lights. (but Whitey's on the moon)
I wonder why he's uppi' me? ('cause Whitey's on the moon?)
I was already payin' 'im fifty a week. (with Whitey on the moon)
Taxes takin' my whole damn check,
Junkies makin' me a nervous wreck,
The price of food is goin' up,
An' as if all that shit wasn't enough
A rat done bit my sister Nell. (with Whitey on the moon)
Her face an' arm began to swell. (but Whitey's on the moon)
Was all that money I made las' year (for Whitey on the moon?)
How come there ain't no money here? (Hm! Whitey's on the moon)
Y'know I jus' 'bout had my fill (of Whitey on the moon)
I think I'll sen' these doctor bills, Airmail special (to Whitey on the moon)
And that's just a device for watching cat videos. The carry-over between rockets and modern technology like hypersonic missiles is even greater - which is why we invented the space race to begin with.
“If I had asked people what they wanted, they would have said faster horses.”
I sure never anticipated cosmetic surgeons twerking on Instagram in their scrubs to advertise themselves between the videos of a chick jumping into a confused Labrador's mouth and that one of the kid on a bike being shocked by the existence of brown chalk when they see a Pokémon drawn on a sidewalk.
But that doesn't mean that indirect spending and praying for spillover works better.
It’s not really the propulsion technology that’s the key, it’s not throwing away a whole bunch of rocket engines every time, just to launch something once. So it’s really a design methodology as much as the technology.
Having said that I think a key contributor to reusability is computation. As far as I’m aware all the experiments with propulsive landing in the 60s and 70s were manual. You could compute optimal thrust profiles given enough time and ideal conditions, but the conditions for an actual landing are never ideal. Particularly issues like atmospheric conditions such as air pressure by altitude, the effect of humidity, wind conditions and allows for the variability of turbulence around the vehicle. Modern automation of such must be vastly more efficient than manual control, but I don’t know when this became viable.
Go to the moon - check
Been there, done that, to expensive and the cocktails where shit.
Unmanned probes do good work for science. (Though even there I am not sure about the bang-for-the-buck compared to other alternatives.)
Have a look at '20 Breakthroughs from 20 Years of Science aboard the International Space Station' https://www.nasa.gov/missions/station/20-breakthroughs-from-...
That's the list of the 20 best things they could come up with. And it scarcely has anything that couldn't have been done cheaper with unmanned craft. And it includes gems like counting spending money as a benefit, instead of a cost.
Indeed, and I believe refusing to acknowledge that fact is actively harming not just spaceflight but manned spaceflight. We're spending tens of billions of dollars trying to set up for some bullshit useless moon base when we could be setting boots on Mars just to prove we can. If we're gonna waste money let's waste it pushing the boundaries.
It then went on to point out that the particle density of helium-3 in regolith of was so low that, despite it being a fusion fuel, it made more sense to catapult the ingots to earth and (1) magnetically decelerate them for energy, and/or (2) burn them.
I suspect one of the best things you can build on the moon and not on Earth is an orbital ring[0]. Well, two rings, first you build one orbiting the moon, learn what the failure modes are and how to mitigate them, then you use that learning experience to export the parts for another to Earth… but I am just straight up assuming that the cost of a moon-based factory won't blow away the cost savings from the difference in delta-v going down to LEO from the Moon (with or without a lunar orbital ring) vs. going up to LEO from the surface of the Earth.
(Also: not mining, but I imagine it would be a great place for dangerous research, inspired by all the people suggesting COVID leaked out of a gain-of-function research laboratory).
Especially since the signal lag from earth is so small, so you can even use remote human operators for many tasks instead of investing a lot in AI.
Single lifetime from flying a powered wooden glider to flying a rocket to the moon and back
What it seems like we're seeing is the compounding of science. Much like how compound interest works. I often wish I was born today just so I could see the progress in the next 50-100 years.
sort of like - a few guys fly across a field to me and my entire high school class can buy tickets and fly across the pacific.
I'm short Immortality is achieved by stopping aging.
I've heard the cache is the biggest part. Per https://en.wikipedia.org/wiki/Apple_A17, the A17 has 72MB of cache. Cache is typically SRAM. Per https://en.wikipedia.org/wiki/Static_random-access_memory, a single bit of SRAM typically uses 6 transistors, but it could be anywhere from 4 to 10. So that's between 2.3B and 5.7B transistors for the cache.
That leaves about 1B transistors for each of the 12 cores (6 CPU + 6 GPU). The instruction set is a few hundred instructions. Surely it doesn't take millions of transistors to implement each instruction. Other than the cache, which parts of the processor need millions of transistors?
Seems to me that all that extra processing power is still looking for a use case.
It is a useful feature. In particular in Google translate (lens function) which I am pretty was also there on the X (perhaps requiring a trip to Google’s server).
I like to keep around test devices that somewhat represent the average user’s device, and when it’s practical a low end device and test against those every so often just to make sure the app performs reasonably well.
This has been in my experience more important for Android apps than for iOS apps, though, because the gulf in power between currently-sold devices in the Android world is vastly more large than in the iOS world. With few exceptions, I find that reasonably efficient apps tested and developed chiefly on newer iPhones work fine on older iPhones, but developing and testing an Android app only on flagships will make for a miserable experience for users of low end or older Android devices.
LLMs on your phone are a few hardware generations out, but they are coming.
https://books.google.com/ngrams/graph?content=submicron+lith...
You're misremembering. Matrox cards were also terribly slow in DOS and anything adjacent to 3D. I own several models with SGRAM. Even GeForce 2 GTS: 25m transistors. G200: 10m. Voodoo2: 4m.
Nope, it was the Parhelia released in 2002 - so somewhat later than the hardware you're considering. The Parhelia was what I was thinking of, with a truly insane number of transistors: 80 million! ("something like" was bearing quite a lot of weight in my first comment :D ). However it was only in the same ballpark, not actually faster that the Geforce 4 Ti or the Radeon 8500, and it was more expensive, so failed in the market.
The relevant anandtech review: https://www.anandtech.com/show/936 (post 936, vs this week's Qualcomm which is post 21112 :D )
https://m.youtube.com/watch?v=lXUWmHgLiyU
At the end of the talk, he mentions the staggering complexity of the design and manufacturing of chips with billions of transistors. We have long reached a point where nobody fully understands how the chips work. IIRC, only two companies can do the manufacturing.
That modern chips work at all almost seems like a miracle. Do we really need chips this complex?
Now here that is on one chip. That change blows my mind.
It seems sort of like asking a programmer how many instructions there are in a binary we maintain. We could find out if we decided it mattered.
A gigantic chunk of those transistors are just SRAM memory cells for various caches. The number of transistors used to actually implement the logic of the SoC is significantly smaller. I would be surprised if it was more than 5%.
The growth of this number between generations is also likely not very large. Most of the growth is also probably cache sizes.