Like when apple releases a new iPhone and there’s a new A17 bionic chip, or whatever they’re on now, how is it 40% faster than the last one? What did they discover in 1 year to make it faster? Why didn’t they make these changes in the last one??
In: 513
Most of the times I feel like there’s an agreement between everyone to feed you small % increments each year even if they are able to do 2x performance.
Short answer: There are processor manufacturing companies constantly coming up with better technological processes, and processor design companies coming up with better designs. A lot of really smart people oversee the whole process.
Processors in very simplified form are just a bunch of transistor logic gates, wired together to be able to do specific instructions. A lot of the improvements over the last 50 years is due to reducing the size of each individual transistor gate. [Here’s more about the process node.](https://en.wikipedia.org/wiki/Semiconductor_device_fabrication) By making the gates smaller you can fit in more of them into smaller space and the energy required to operate each gate is reduced, making faster processors.
We are nearing the limits of what’s physically possible to manufacture, because of two things:
1. logic gates become small enough that quantum effects become a meaningful force, resulting in errors in logic gate operations.
2. logic gates become small enough that they’re made from only a couple dozen silicon atoms. We don’t know what’s the minimum amount of atoms we can make one logic gate out of.
It is estimated that we will not be able to achieve technology node lower than 1nm (we’re currently at 3nm for newest Apple silicon). However, we will then try to improve computational capacity elsewhere – for example by adapting different, faster instruction sets, or by improving memory caching / other possible optimizations.
It’s called “product planning”. Tech companies plan and design for incremental upgrades so that they can keep coming out with new products, and keep selling to their existing customer base.
When Apple is releasing a new iPhone, there are always features that they held back. Why? So that they can use some of those feature in next year’s iPhone.
It also helps to maintain price consistency; Apple’s iPhone prices are largely the same from year to year. Compare to the game console market, where they try to pack every feature in, knowing it’ll be seven years before they release the next upgrade. So the price of consoles is generally very high at launch, and gradually floats down over time as the technologies become cheaper.
Computer chips are made out of semiconductors. Every two years or so, we find another way to manufacture smaller semiconductors. If you make a chip smaller but keep the design the same, it will use less power. This also means that you can build more semiconductors on the chip and get more performance out of the same power budget.
The first year that a manufacturer moves to a smaller process, the yields tend to be worse (there is a higher rate of defective or faulty chips) so the designs tend to be more sophisticated on the second chip designed on that process.
Intel referred to this as their ‘tick–tock’ model, where there would be a process size shrink followed by a microarchitecture revision.
Well, technology keeps advancing and we keep discovering better transistors.
But also, they are a company who wants to make money. You can’t just give out all the goodies at once. You should try to slowly feed them into the market so you get more long term income.
one way processors get faster is by reducing the physical space between transistors and making them smaller. you might hear things like 8nm processor or 5nm processor. that’s saying how large the transistor is, the smaller they are,the closer you can put them next to each other, the closer they are next to each other the faster they can “communicate”
with each other, the closer they are the more you can put into the same area. it’s a very difficult to get things smaller and smaller. for instance a nanometer is extremely small. 1 meter is 100, 1 centimeter is 100 millimeters , 1 millimeter is 100 microns, 1 micron is 100 nanometers. that means 1 meter is 1,000,000 nanometers. that’s extremely small and you need a factory that can make things at that such small scale with extreme precision. if 2 transistors touch that can mean a failure, if transistors aren’t consistently sized that can mean a failure. a lot of big corporations don’t make their own processor dies (samsung, AMD, Nvidia, apple) they buy them through TSMC which spends billions in R&D to create ways to make such microscopic things
Second hand anecdotal story, so I can’t verify it, but it does come from a very close friend who I don’t think would have any reason to lie to me.
He used to work for ‘one of the big’ phone companies, not Google or Apple but a big well known name. He developed a way to extend the battery life by a substantial amount. I can’t remember the exact figure, but for the purposes of the post lets say a Three Day battery.
When he took it to the superiors, they asked him to put it down to a Two Day battery, as that was already a decent improvement on what they had, and they could save the Three Day battery for next years release.
So to answer your question, I reckon phones could be a lot faster right now if companies didn’t like money so much.
Do they?
Transistor size is still shrinking, so they can put 20-30-40% more with each smaller transistor size. How can they use more transistors to make things faster?
1. Add more computing cores.
With graphic procesors that is a bit easier, because graphics calculations are inherently easy to parallelize and you can have many cores and use them effectively.
CPU cores are a bit different. Some tasks can’t be parallelized, so you have no use from more cores.
2. Make cores faster
It’s complex topic, but there are some “calculations” that you can do a lot faster if you make a dedicated hardware for it. For example you can take hundreds or thousands of clock cycles to multiply two numbers or you can make a part of the cpu that does only that one thing and can do it in several cycles.
You can make calculations for multiple things in advance, and depending on outcome of some calculation just use the one you need. That’s called speculative execution and is root of some security problems.
Cpu clock still does rise with smaller transistor size, so there is that.
3. You can add more cache memory
Cache is a really fast memory sitting near cpu cores. Instead of waiting for RAM to deliver information, you have it at hand.
—
When phones are in question, power draw is a big factor. As transistors get smaller, they use less energy, so you can add more of whatever you were constrained by power usage before.
There are other ways, I’m sure people will write about them.
It wasn’t discovered in one year, it was discovered a while back but if the company just makes a huge leap to the limits of current technology then they won’t have anything new to add for another 4-5 years. So what they do instead is they trickle feed the tech in smaller increments so that people feel compelled every year to buy a new phone. You cannot imagine the kind of money they’re making off of that but also the amount of waste being produced.
>What did they discover in 1 year to make it faster?
Just for clarification: the stuff that goes into next year’s iphone doesn’t get discovered this year. The stuff that they are doing research on today won’t get used until about 5 years down the line. That’s why those phones are so expensive. Not because of the build cost, but because it takes a long time and a lot of money to develop new microchips.
Source: I work in a cleanroom where most big companies do a lot of their research on microchips.
Latest Answers