Apple, Intel to Become Alpha Customers for TSMC's 2 nm Manufacturing Node

[TheoneandonlyMrK]

I'm picking up on AMD because generally the crap on TPU is AMD vs Intel. Imo there is no doubt within 5 years Intel will have got their shit together and will be ahead of AMD again because they have their own FABS, again imo they are only using TSMC which is a recent thing is it not? until they get their shit together with their CPU making FABS.

Apple and Nvidia have no real correlation to AMD or Intel as Apple is mostly phones and computers that are not really in competition with AMD or Intel, and the other is GPU's only.

I disagree on Apple and Nvidia irrelevance , Apple has M2 due, not a phone chip , Nvidia has grace due, not a phone chip.

The Amd v Intel shit is caused by some having, and showing a bias.
I have been involved in the past I'll admit As an Adult should, I was wrong in that but realisation doesn't hurt and striving against your own biases helps a person be more reasoned in they're argument.

SI has little to do with naming convention, it should but by and large they're all decided by marketing men now not reality.

 

[Deleted member 24505]

Do vs discussion's on TPU include Apples M2 or Nvidia's Grace? no that is why i said they have no relevance. Maybe they will when they actually exist, but not at this time.

Imo at this time i have never known TPU so be so anti Intel as it is now.

 

[ARF]

If you look very carefully into IBM's 2 nm prototype chip, you can find a thin insulation or passivation layer around nanosheets which seems to be about 2 nm thick.

IBM—First to Produce 2nm Chip_05/16/21

IBM—First to Produce 2nm Chip IBM announced a breakthrough in semiconductor design and process with the development of the world’s first chip announced with 2 nm nanosheet technology. The new...

www.oled-a.org

Also in colour: https://semiengineering.com/new-transistor-structures-at-3nm-2nm/ (Fig. 3)

Yes, it looks like IBM sets the industry standard that should be followed by all the others. But TSMC doesn't follow it, so on its N2 node (keep in mind that it is N2, and nothing to do with 2 nm) the dimension of the smallest feature will be considerably larger than 2 nm.

So, what is it - race to who will make the thinnest nanosheet? lol How about to describe something more of use - like the overall size of the transistor?

 

[usiname]

I still can't understand why some are so outraged by the names of the processes. The names have nothing to do with the reallity even if they did, they would still have discrepancies, because each process is different from the others. Even the transistor density is not a guarantee of a good process, look at samsung, their process has great density, but is trash in terms of power consumption, heat and clocks. If you have problem, start with samsung, not TSMC. Samsung is first in the nanometer race and years behind everyone in the same time

 

[ARF]

I am wondering - can TSMC start a class-action law-suit against Samsung for potential misleading of the customers who would believe that both companies have comparable processes?
This is a type of anti-competitive practice.

 

[ghazi]

So Intel cucked AMD. Wonder how this will go. People who have been following this for a long time know that even if AMD has the better designs manufacturing limitations can and will screw them over.

 

[TheoneandonlyMrK]

So Intel cucked AMD. Wonder how this will go. People who have been following this for a long time know that even if AMD has the better designs manufacturing limitations can and will screw them over.

Thankfully, nothing in fab land is that certain, Euv is new enough to see vast improvement and they're all doing different 3D transistors, so Intel or Samsung could spend they're way to a winning position.

 

[simlife]

Good point, I honestly don't know what the purpose of this race to lower numbers actually means. Look at the discrepancy between the IRDS roadmap that sets the process specifications, and then look at how TSMC violates it, with one full node behind the roadmap, while Samsung is either two or even three full nodes behind the schedule.

I mean the customer is not stupid, he will see the final performance and will understand that Samsung is a no-go.

omg the consumer is 100000% stupid(i can show you like a million videos proving me right..) i worked in retail my whole life im 35 now and a vendor but still see this a exact same crap almost daily the 4.79 ruffles had 18ppl buy but it but JUST above it the 50% larger bag was 4.49... same flavor AND brand.. why? same chips same quility less product and more cost.... this is a thing x a million some ppl do.... tostitos chip flavor was 1 dollar cheaper for 50% more product if they got the party size and i took the time to point it out to 2 diffenrt customers but stopped and wont again because they looked at me funny and told didnt have the digital coupon... these are the ppl who cry about gas but drive aggresivly and drive trunks and break faster meaning wasted gas

so shadey naming sceames they should be sued for i have stories for years about saving even more the ppl care even less about.... the customoer is SUPER stupid i would in fact bet my entire life savings that... i bought a house while making less then co workers and they asked how... lss they nickel and dime themsleves when i watch what i do and track it your the "consumer is not sutpid" (ppl but smart phones every year) i could tanslate even higher to the tech indoustry... what comusmers are you actually talking to t make this INSANE claim ?!?!??

 

[R0H1T]

Intel's not an "alpha" anywhere in the world right now, this is just BS

 

[unknownk]

So Intel cucked AMD. Wonder how this will go. People who have been following this for a long time know that even if AMD has the better designs manufacturing limitations can and will screw them over.

The way things work is that newer nodes are used on smaller, slower and simpler chips until the foundry gets yields and speeds up enough for larger stuff like x86 CPUs. Apple uses the smallest processes for simpler and slower phone processors while AMD uses the much more mature processes for CPU and GPU chips.

Intel stumbled because they were still making large monolithic chips (whole CPU is 1 die) and yields went to hell on smaller nodes. AMD broke things down into more economical chiplets with an older process doing the controller interface which was more economical but slower in top speed because of the interconnects.

if you own your fabs and can have them running 100% capacity you are golden because of sheer production capacity. Once those plants start to idle you go bankrupt unless you have deep pockets and time to fix it.

 

[R0H1T]

Apple uses the smallest processes for simpler and slower phone processors

Apple till recently only used the one (type of) chips, so this doesn't really hold true. M1 is what barely two years old? They did use the oversized A12x or A12z for top of the line tablets but that was just a one off IIRC.

 

[kony]

Alpha customers scare me not, I'm a sigma.

Betas better hide.

 

[Vayra86]

So what's the actual nm size, is the 2nm an actual 5?

Oh no the gap is bigger. Marketing is easily two to three shrinks ahead of reality.

I am wondering - can TSMC start a class-action law-suit against Samsung for potential misleading of the customers who would believe that both companies have comparable processes?
This is a type of anti-competitive practice.

Its marketing. TSMC requires a new marketing approach to fix it, Intel already adjusted that way, and eventually the product speaks for its real quality.

 

[unknownk]

Apple till recently only used the one (type of) chips, so this doesn't really hold true. M1 is what barely two years old? They did use the oversized A12x or A12z for top of the line tablets but that was just a one off IIRC.

M1 is 5nm process while Zen 3 is 7nm. Zen 4 being produced now is a special 5nm production. Its not like AMD is making Intel 14nm chips or something outdated.

 

[R0H1T]

Apple didn't switch to just using small dies for a new node, they've always used the latest node for their biggest chips ~ it's just that they had this one (singular) size chips before A12x/12z & transitioning to multiple die sizes is barely 2 years old.

Its not like AMD is making Intel 14nm chips or something outdated.

Not sure where I said or even implied that? The 7nm,14 nm & older nodes will live for a long time!

 

[mama]

This indicates two things: Intel is desperate for it's competition to have limited supply from TSMC in the future and secondly that likely it has limited confidence in it's own chip making development.

 

[unknownk]

Intel still has enough money to invest in its own chip making facilities AND purchase a supply line from TMSC for high value items. Could be wrong but I think they want to make GPU or compute cards at TMSC and not CPUs. Intel could make uncompetitive garbage and still sell it because AMD can't make enough chips for more than 20% of market anyway.

 

[ARF]

Intel probably spends more money on development than AMD's entire budget.

And Intel's products still remain mediocre at best compared to AMD's. AMD is the technology leader in the CPU x86-64 segment..

 

[ARF]

I wouldn't call the 12900k mediocre or the 12700k but hey ho i see your AMD so your response is typical of that.

1 year late and power consumption through the roof. Fail from all views...

Intel Core i9-12900K and Core i5-12600K Power Consumption and Efficiency - Intel Core i9-12900K and Core i5-12600K Review: Retaking the Gaming Crown | Tom's Hardware (tomshardware.com)

 

[ARF]

Well my 12700k in your chart uses less than a 5950x so whatever the 12900k uses is irrelevant to me. It might use high power when running apps that push it, but for gaming it uses less than some AMD.
using high power is the only thing AMD users can continually spout about a CPU that is as good or better than a 5950x which is the best AMD CPU atm.

Performance per watt is a direct result of the process node being used. In this case, clearly shows that Intel is years behind and the only thing that saves them is the heavy factory overclock of their otherwise mediocre offerings.

 

[ARF]

Whatever you say mr TPU expert. My mediocre 12700k matches or betters your 5900x in everything, not bad for mediocre.

I guess everyone who bought Intel 12th gen should have bought AMD instead eh? Go forth and multiply.

Your 12700K is slower in everything except 720p and 1080p old-school gaming than my Ryzen 9 5900X

 

[TheoneandonlyMrK]

Whatever you say mr TPU expert. My mediocre 12700k matches or betters your 5900x in everything, not bad for mediocre.

I guess everyone who bought Intel 12th gen should have bought AMD instead eh? Go forth and multiply.

It's worth remembering that Intel's 12th generation is their latest pciex 5 Ddr5 platform, one that will be a basis for moving forward.
It's against AMD's present, out nearly a year and a half platform which is known to be approaching EOL.
Just about beating an old platform with your latest is not in many people's eyes a great achievement.
Hence some people's negativity.

At all times if you're platform does what YOU want, what else matters.

I hope node conversations start back into this thread soon.

And to start, after thinking about it I would be surprised if Intel usurp Apple as customer number one, this rumour needs more meat on it.

 

[Denver]

Sadly, Intel always over promissing and under delivering... What happened with the plan to get back to the top ? XD

Go Samsung!!

 

[Vayra86]

Well my 12700k in your chart uses less than a 5950x so whatever the 12900k uses is irrelevant to me. It might use high power when running apps that push it, but for gaming it uses less than some AMD.

This is pulling wool over your eyes. "If I dont put serious load on it, it might use less than some AMD CPUs" lol. Not very objective is it.

There is no anti Intel sentiment here, only people calling things what they are. I remember an Intel Quadcore era where most people shat all over inefficient Bulldozers and Piledrivers and I also remember AMD getting flak for high power usage on GPU. I was one of those, just like I shit over ADLs 241W ridiculousness.

The constant in all of this is: performance per watt appears to be very highly valued by a large group of customers. And I totally get that. Look at the sentiment towards Ampere and incoming Lovelace announcing 450-600W potential TDPs!!

Id suggest you stop seeing the good old green red conflict and look at the market instead.

 

[R0H1T]

Tbf there's good arguments on both sides ~ AMD can't clock as high as Intel on all cores because of IF & their chiplet route, this is also the reason why they are more efficient on desktops. Intel 12th gen on laptops is super efficient, albeit with a slight node & massive IPC advantage as compared to zen 6xxx chips. zen4 should be interesting extremely so if they can pull a zen2 -> zen3 leap again!