Intel shares details on its 14nm process, also states they are ready for mass production

[LukaP]

http://www.anandtech.com/show/8367/intels-14nm-technology-in-detail
 

 

Intel’s 14nm process has been their most difficult process to develop yet, a fact that Intel is being very straightforward about. Throughout the life of the 14nm process so far its yields have trailed the 22nm at equivalent points in time, and while yields are now healthy enough for volume production Intel still has further work to do to improve the process to catch up with 22nm.

 

 Intel has provided the minimum feature size data for 3 critical feature size measurements: transistor fin pitch, transistor gate pitch, and the interconnect pitch. From 22nm to 14nm these features have been reduced in size by between 22% and 35%, which is consistent with the (very roughly) 30%-35% reduction in feature size that one would expect from a full node shrink.

 

Intel is especially proud of their interconnect scaling on the 14nm node, as the 35% reduction in the minimum interconnect pitch is better than normal for a new process node.
Along with the immediate feature size improvements that come with a smaller manufacturing node, Intel has also been iterating on their FinFET technology, which is now in its second generation for the 14nm process. Compared to the 22nm process, the 14nm process’s fins are more tightly packed, thinner, taller, and fewer in number (per transistor).

 

 

they also state some 5% IPC improvement and 2x performance per watt. i hope this means a 140w broadwell e part and 80-90w broadwell k parts, not decreased TDPs...

 

also release is now H1 2015 and shipping from arizona, oregon and ireland

[flibberdipper]

No idea what the pics mean, but I like the news regardless.

[ViruZ_HUN]

the pics look like teeth

[Trik'Stari]

Could they at least do some price dropping on shit? I'd love to get a 4790k lol

[LukaP]

No idea what the pics mean, but I like the news regardless.

 

 

the pics look like teeth

its transistors, SEM picture and a schematic view. shows the gate substrate and fins. pm me if you wanna know what they do

[DaRk0]

Someone needs to make 'meanwhile at TSMC/GloFo' meme.

2x performance per watt without decreasing TDP would make worthwhile upgrade from current gen.

[Goldilock]

need cheaper i5

[azurezeed]

i want to see how good will be the performance boost on broadwell

[Xorbot]

The power efficiency of each generation keeps making me think that my old 386 was like a lawnmower.

[flibberdipper]

its transistors, SEM picture and a schematic view. shows the gate substrate and fins. pm me if you wanna know what they do

Naw, I don't wanna learn.

 

The power efficiency of each generation keeps making me think that my old 386 was like a lawnmower.

Same with my Pentium 4... Just it was a riding mower with a 30" deck.

[connorpiper]

You can see the difference, we aren't just jumping in transistor size, Intel has also done wonders to the actual shape and packing of them. 

[The_Strict_Nein]

 

they also state some 5% IPC improvement and 2x performance per watt. i hope this means a 140w broadwell e part and 80-90w broadwell k parts, not decreased TDPs...

 

also release is now H1 2015 and shipping from arizona, oregon and ireland

 

I doubt we'll see a doubling in performance per watt this generation, I'm guessing 60-70W TDP i7 K SKUs at most.

 

Also, the great thing about doing FET technology for my physics coursework is that I understand exactly what effect all this has.

[patrickjp93]

'oly fak but I'm worried about OC...

The transistors are much more efficient, so you should get more clocks per voltage increase than Haswell is capable of, assuming thermal displacement is good enough of course.

[patrickjp93]

i want to see how good will be the performance boost on broadwell

For most applications, 5%. For scientific computing where the floating point multiply and divide operations have been cut from 5 cycles to 3 can represent a 33% improvement.

[LukaP]

Naw, I don't wanna learn.

 

Same with my Pentium 4... Just it was a riding mower with a 30" deck.

your loss

[patrickjp93]

But I mean high voltages...

You seem to have missed my point. I'm saying at the same 1.2 volts you'll probably see a .5 GHz difference if your cooling is up to the task.

[LukaP]

I doubt we'll see a doubling in performance per watt this generation, I'm guessing 60-70W TDP i7 K SKUs at most.

 

Also, the great thing about doing FET technology for my physics coursework is that I understand exactly what effect all this has.

yeah same here was just studiying quantum mechanics as a part of getting ready for lectures on it this year. and its awesome, actually seeing it in action

[LukaP]

-.- IKR, I got it but I wasn't asking that... I was asking about how would be a high voltage on that...

bad. the smaller the transistor, the smaller the voltage as to (if i may put it in this analogy) not have the electrons erode the transistor like water erodes rock 

[patrickjp93]

-.- IKR, I got it but I wasn't asking that... I was asking about how would be a high voltage on that... voltage range if that can explain what I mean

I imagine Broadwell will be much better at breaking past 5 GHz than Haswell. As per what voltages it can endure, probably a little less. We're reaching the point where burst heat can possibly do damage before the thermal sensors can throttle it back. As per electrical erosion, I doubt that is as much a problem with this, as tri-gate tech is inherently more fail-safe than simple two-gate due to natural impedance, but like I said we're entering waters where overclocking headroom is going to begin to disappear.

[LukaP]

That's what I'm referring to... how BAD would it be?

a little worse than haswell, id imagine 1.2V to be a high voltage here, with max being 1.25-1.28V

 

I imagine Broadwell will be much better at breaking past 5 GHz than Haswell. As per what voltages it can endure, probably a little less. We're reaching the point where burst heat can possibly do damage before the thermal sensors can throttle it back. As per electrical erosion, I doubt that is as much a problem with this, as tri-gate tech is inherently more fail-safe than simple two-gate due to natural impedance, but like I said we're entering waters where overclocking headroom is going to begin to disappear.

this

[deviant88]

unlock all i3's i5's or fuck right off intel id rather have K unlimited OC and cheaper ulocked model with limited OC like 4.2-4.5 ghz only their K's be the ones that go as much as silicon lottery allows 5ghz or whatever

WTF my 3.6ghz 4570 probably can do easy 4ghz and its locked for no reason screw you intel,and screw amd for not doing any competition its like they got paid by intel to leave the desktop market to them.

[mr moose]

 

Not long now...

 

 

 

 

Well, I have been into this stuff for the better part of 32 years, I have several qualifications in EE (not that they hold any water these days), But every time I look at what they are doing and the sizes they are working with it still blows my mind.

[LukaP]

@deviant88 its called market segmentation, and its a business tactic completely reasonable for a big monopolistic firm... Do you think people would buy the 4690k if there was a unlocked i3 available, that performed the same in games for half the,money? No.

 

and AMD would do the same if they were in Intel shoes. They only have everything unlocked because they don't have the money to fuck around with their marketbase like Intel has.

[patrickjp93]

unlock all i3's i5's or fuck right off intel id rather have K unlimited OC and cheaper ulocked model with limited OC like 4.2-4.5 ghz only their K's be the ones that go as much as silicon lottery allows 5ghz or whatever

WTF my 3.6ghz 4570 probably can do easy 4ghz and its locked for no reason screw you intel,and screw amd for not doing any competition its like they got paid by intel to leave the desktop market to them.

Unlocking everything just isn't appropriate for most consumers and businesses. Let's use Intel's virtualisation and security features in vPro as an example. They're highly important to businesses and people who run virtual machine servers. These features happen to become fatally unstable when clock rates pass a certain point. Intel locks down the chips as high up in the clock rates as they can while guaranteeing xyz functions will work. This is why a number of instruction set extensions do not exist in K-series chips (though why the Hell CPU-Boss reports the 4790k has no avx or sse is beyond me).

[c3p0]

The only thing I'm wondering. New socket again or does it stay at 1150?