45% faster using 75% less power. Whoda thunk it possible?
www.anandtech.com
2nm transistors
- Thread starter Shadow Reaper
- Start date
Semiconductor engineers are a whole different animal. I barely made it through the basics in my EE curriculum, and to see these folks continuing to develop the technology despite some very hard physics limits is mind-blowing.
Curious to see how long before this architecture makes it's way into products accessible to independent users.
Curious to see how long before this architecture makes it's way into products accessible to independent users.
7nm chips came to market in less than a year. It's important to rush because the competition is usually just right behind, and if you can beat them by an entire year you can own the market.
"75% less power" is especially interesting in that it may be suitable to break the 5Ghz wall. Brave new world.
"75% less power" is especially interesting in that it may be suitable to break the 5Ghz wall. Brave new world.
Single core performance is a big deal for people like me who have way too much software/games that still rely on that. Looking forward to the near future
another line of intel processors eh? hopefully makes the market more affordable.
amd is going to have to make the 2nm processors before they can compete to lower prices of the next gen. with that said though the current generation is going to probably break down alot if intel sells the 2nm at an affordable price. i'm banking on amd to get 2nm fast since they are right behind the curve with intel these days.
Last I checked, Intel has been pretty well stuck refreshing their 14nm chips for quite a few years while AMD has been gradually decreasing sizes steadily the past 6 years. I don't foresee, without some sort of new help, Intel getting down to 2nm with it being stable, reliable & mass producible faster than AMD with the help of TSMC & Samsung. I am however looking forward to technology advancing as well as software designers moving towards their products scaling in performance based on how many cores & processors a PC has instead of focusing on 1 or 2 main cores while the others pick up slack wherever else. Can you imagine how well SC would run if its software could make maximum use of all except a couple cores equally & leave those for background processes?
Considering DNA is 2.5nm wide, this is pretty big (small?) news!
Hmm, interesting. Look here:
(4 nm is mobile processor, but still)
These manufacturers keep talking of walls and barriers that stop them from making smaller semiconductors, but somehow every 10 years without fail they've managed to trim like 80% of the size. 1991 being an exception.
| 1971 | 10 000 nm | - | |||||
| 1981 | 1 500 nm | 15% of previous | |||||
| 1991 | 600 nm | 40% of previous | |||||
| 2001 | 130 nm | 21% of previous | |||||
| 2011 | 28 nm | 21% of previous | |||||
| 2021 | 4 nm | 14% of previous |
(4 nm is mobile processor, but still)
These manufacturers keep talking of walls and barriers that stop them from making smaller semiconductors, but somehow every 10 years without fail they've managed to trim like 80% of the size. 1991 being an exception.
These companies keep surprising themselves. They say they are reaching the physical limit of the materials they have been using, but we might soon measure our chips in picometers.
Good news as it will drive prices down for larger nm chips that can still be highly performing for Star Citizen ;-)
But I'm still waiting for advanced quantum computing in a positronic network. I'm not getting any younger here ...
But I'm still waiting for advanced quantum computing in a positronic network. I'm not getting any younger here ...
The problem up to now has been the ability to control the transistor size. There have been successful laboratory tests of 1.2 and 1 nm, but we now think that is the basic limit of silicon. 1nm is only 20 silicon atoms wide, and it becomes difficult to create a transistor smaller than that - because the electrons in the material begin to be affected by their quantum properties. You can make the transistor smaller, but you can't control where the electrons are, and they can leak across the transistor base and even between transistors. In a digital system, that means 0's can randomly become 1's, and 1's can randomly become 0's.
Don't hold your breath. The quantum processors they currently have are the size of a refrigerator and operate at temperatures like .1 degree above absolute 0. They do amazing things though, like calculate in a second what the fastest silicon based supercomputers take a week to process. But the technology is starting to see industrial applications now so development will accelerate.
Last edited:
Heres some good reads to skim through.
EDIT: Most of the stock I own is in Nvidia. I've been doing a little bit of research into the whole semiconductor industry and its all really interesting. The manufacturing process they use to manufacture the wafers is borderline magic.
EDIT: Most of the stock I own is in Nvidia. I've been doing a little bit of research into the whole semiconductor industry and its all really interesting. The manufacturing process they use to manufacture the wafers is borderline magic.
Last edited:
It is indeed amazing stuff. I'd be interested to learn more about how well it can be hardened against space radiation. There's no industry so affected by the need for smaller, lighter using less power than aerospace.
Most space radiaton hardened transistors are made from silicon carbide or gallium nitride. In general what you want is a substrate with a high band gap. Carbon forms like nanotubes, graphene and the like don't have well understood bandgaps. They can be modulated and current studies are exploring this, but short answer is yes. It appears carbon nanotubes show promise for next generation transistors, if you can say that while saying we don't know enough about carbon yet.
My point though was that 2nm transistors need to be characterized for endurance with space radiation. In general, the smaller the transistor, the more vulnerable it is to radiation of many types. However, it is also more easily and cheaply protected from radiation. So, needs to be looked at.
My point though was that 2nm transistors need to be characterized for endurance with space radiation. In general, the smaller the transistor, the more vulnerable it is to radiation of many types. However, it is also more easily and cheaply protected from radiation. So, needs to be looked at.
If the purpose of the radiation hardening is to ensure proper operation of computer systems, one way theyre already doing that is through triple redundancy. Maybe you could also build the chip to compensate for errors caused by radiation? I'd be surprised if that too wasnt already the case.
Triple reduncancy is an older method that works on volitile systems but is power hungry. There are actually several methods and some kinds of components like MRAM are radiation tolerant by nature. Each component needs its own fix.
I am just waiting on Intel and AMD to ditch the x86 architecture. Wonder if Apple will cause the pivot we need on the windows side.