Intel Core i3-8121U 10nm processor hits ARK database

[HEXUS]

Status is officially launched. We saw this CPU earlier this week in the Lenovo Ideapad 330.

Read more.

[lumireleon]

is it true GFoundries 14nm is not a true 14nm but 20nm with certain FinFet sections at 14nm? So some speculators are claiming TSMC /SAMSUNG and GF 7nm process is actually equivalent to intel's 10nm process. Plain gossip or pure architectural facts that intel is still at the top of tech?

[3dcandy]

Samsung appear above Intel in fab tech these days - I mean they have been shipping 10nm for ages, they have great 3d flash and they seem to be best overall on dram. Intel in many ways seem to have lost their way, especially in the field of cpu's both with the actual cpu's and in the node they are produced on

[preter_s]

is it true GFoundries 14nm is not a true 14nm but 20nm with certain FinFet sections at 14nm? So some speculators are claiming TSMC /SAMSUNG and GF 7nm process is actually equivalent to intel's 10nm process. Plain gossip or pure architectural facts that intel is still at the top of tech?

Why the heck would u fools bother playing the marketing BS game set up by Intel. The performance gains from AMD already show that the independent fabs likes GloFo/TSMS/Samsung have already caught up if not surpassed Intel. AMD, without having to concern themselves with the R&D or retooling costs, can use ANY of them so long as they make the best. Boy are you 20 years out of date!

[preter_s]

What a crap BS CPU! It will be dead on arrival.

AMD doesn't even bother making dual core CPUs anymore. Its cheapest lowest end CPU/APU is quad cores at least. Are consumers really so dumb as to buy this dual core i3 crap?

[Ozaron]

What a crap BS CPU! It will be dead on arrival.

AMD doesn't even bother making dual core CPUs anymore. Its cheapest lowest end CPU/APU is quad cores at least. Are consumers really so dumb as to buy this dual core i3 crap?

Less dumb, more uncaring. There will be people for whom this does a job nicely. It just doesn't meet our personal performance standards, is all.

[DanceswithUnix]

I could understand if this was an Atom replacement, but 2 cores and no GPU just doesn't make sense to me. That would have to be a pad limited part, it needs more stuff to fill the space between the connections to the outside world. But in a low end 2 core part you would really want integrated graphics.

[Corky34]

is it true GFoundries 14nm is not a true 14nm but 20nm with certain FinFet sections at 14nm? So some speculators are claiming TSMC /SAMSUNG and GF 7nm process is actually equivalent to intel's 10nm process. Plain gossip or pure architectural facts that intel is still at the top of tech?

For the last decade or so node sizes, especially between different fabrication companies, have lost most of their meaning...

I was going to write why but then came across this article on semiengineering.com that does a far better job of giving a succinct explanation than me, basically since we moved away from traditional CMOS transistors there were more features to measure than just the length of the gate

Why the heck would u fools bother playing the marketing BS game set up by Intel. The performance gains from AMD already show that the independent fabs likes GloFo/TSMS/Samsung have already caught up if not surpassed Intel. AMD, without having to concern themselves with the R&D or retooling costs, can use ANY of them so long as they make the best. Boy are you 20 years out of date!

Because us fools know it wasn't marketing BS set up by Intel, in fact IIRC Intel even said a few years back that the industry should really move away from using node sizes. The semiconductor engineering world consists of far more than just Intel.

And no the independent fabs likes GloFo/TSMS/Samsung have not caught up with Intel in terms of performance, that is if you're defining performance as the switching speed of transistors (if not perhaps you could define what you mean when referring to performance), GloFo/TSMS/Samsung and Intel's design goals are very different, they're even different on particular lines within a single company (GloFo and TSMS have something like 3-4 different designs targeting different markets).

[3dcandy]

For the last decade or so node sizes, especially between different fabrication companies, have lost most of their meaning...

I was going to write why but then came across this article on semiengineering.com that does a far better job of giving a succinct explanation than me, basically since we moved away from traditional CMOS transistors there were more features to measure than just the length of the gate



Because us fools know it wasn't marketing BS set up by Intel, in fact IIRC Intel even said a few years back that the industry should really move away from using node sizes. The semiconductor engineering world consists of far more than just Intel.

And no the independent fabs likes GloFo/TSMS/Samsung have not caught up with Intel in terms of performance, that is if you're defining performance as the switching speed of transistors (if not perhaps you could define what you mean when referring to performance), GloFo/TSMS/Samsung and Intel's design goals are very different, they're even different on particular lines within a single company (GloFo and TSMS have something like 3-4 different designs targeting different markets).

Which is why I feel they are being left behind in many ways. Slow to react yet again....the good ship Intel is about to scrape an iceberg

[Corky34]

I'd say it's not really a case of being left behind, it's a matter of horses for courses, Intel have been, from what i can tell as they're not as open as other fabrication plants, targeting higher clock speeds, that's not to say that's their only concern as IBM 5Ghz power8 processor is stronger on clock speed, they have however placed the switching speed of the transistor pretty high in the list of design goals.

GloFo, TSMS and Samsung all have their own take on designing transistors that favor clock speed, power efficiency, density or any combination of those things, obviously if you want higher clock speeds you give up on some power efficiency and density and vice versa.

[3dcandy]

But one could argue that this 1 approach IS leaving them open to being left behind.... which is my point. Other foundries have a few processes and the designers can choose which one suites them best whereas Intel it's one as far as I can say. The fact that the latest 10nm processors are late, only dual core and have the integrated graphics disabled points to a problem somewhere. Intel was cutting edge in many ways and just appears to be slipping back - it's this which the OP was hinting at about being left behind

[Xlucine]

I could understand if this was an Atom replacement, but 2 cores and no GPU just doesn't make sense to me. That would have to be a pad limited part, it needs more stuff to fill the space between the connections to the outside world. But in a low end 2 core part you would really want integrated graphics.

There's rumours of a Y series (4.5W) chip on the same node, with the same 2 cores but with the IGP enabled
https://wccftech.com/intel-10nm-cann...essor-spotted/

Intel 10nm Cannonlake-Y ‘M3-8114Y’ Dual Core Processor Spotted – 4.5W TDP, UHD Integrated Graphics and 1.5 GHz Base Clock

Intel making a 2C4T die with IGP, and then disabling the IGP for yields wherever they can, makes a lot more sense to me than a 2C4T CPU. Then again I thought the quad core chips on X299 were too stupid to be real, so intel may surprise us yet

[Corky34]

But one could argue that this 1 approach IS leaving them open to being left behind.... which is my point. Other foundries have a few processes and the designers can choose which one suites them best whereas Intel it's one as far as I can say. The fact that the latest 10nm processors are late, only dual core and have the integrated graphics disabled points to a problem somewhere. Intel was cutting edge in many ways and just appears to be slipping back - it's this which the OP was hinting at about being left behind

True but it's the old argument of specialising versus diversifying, Intel's focus on transistor switching speed has made them very good at it as when compared to other designs focused on the same thing Intel's design is better, GloFo, TSMS and Samsung's high performance designs still lag behind Intel (afaik), Intel simply clocks higher.

Like i said it's horses for courses and arguably things have been moving away from high clock speeds for a long time, the markets want good performance (in the general meaning of the term) per watt, it's all about doing more with less, where the independent foundries really shine (imo) are in the mid to low end in terms of general performance as they're incredibly efficient, i mean seriously who would have thought we'd have a 4c8t 3.8Ghz CPU and GPU in a 15W envelope.

I get what you're saying about being left behind what with the delays to 10nm but their not really as despite other foundries approaching the same notional size the designs and end results are so different that they're not really comparable, it's a bit like trying to compare a motorbike with a car.

[DanceswithUnix]

There's rumours of a Y series (4.5W) chip on the same node, with the same 2 cores but with the IGP enabled
https://wccftech.com/intel-10nm-cann...essor-spotted/


Intel making a 2C4T die with IGP, and then disabling the IGP for yields wherever they can, makes a lot more sense to me than a 2C4T CPU. Then again I thought the quad core chips on X299 were too stupid to be real, so intel may surprise us yet

Well yes, it has to be a die harvested part. But you don't usually die harvest by turning off the entire GPU unless some awful design error crept in and the entire GPU is bad. In a world where Intel used to think it was fine for their GPU to hard lock systems on a regular basis I have to wonder just how broken a GPU would have to be to just turn it off.

As for the "my features are smaller than your features" race, it doesn't really matter that much any more. A process shrink used to get you twice the transistors and way higher clocks for the same cost. Now it gets you more transistors for more cost and you just hope you break even on the deal. TBH I'm not expecting to ever see a CPU twice as fast as the ones currently available.

[3dcandy]

True we are never going to see that again. All points argued well here, I guess we're being very diplomatic

[DanceswithUnix]

True we are never going to see that again. All points argued well here, I guess we're being very diplomatic

I expect the shouty tribalistic types aren't going to bother reading a thread with i3 in the title

For now I'm sure Intel can keep tweaking their 14nm process to excellent effect, so this pointing to still having problems at 10nm I don't think is significant. Other than a dent to pride when at one point Intel were generations ahead in process compared to everyone else.