• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

[SA] Intel guts 10nm to get it out the door

NTMBK

NTMBK

Lifer
https://semiaccurate.com/2018/08/02/intel-guts-10nm-to-get-it-out-the-door/

Sadly the technical details are behind the paywall, but the key quote from the intro is this:

What Intel is not telling you, or the analysts, is that the 10nm you may get in late 2019 is not the 10nm they had intended to come out in 2015. More importantly this new process is a significant step backward from the 10nm they promised, as touted in their manufacturing day. How much of a step backwards? Several of SemiAccurate’s moles are saying it is effectively a 12nm process rather than a 10nm process, and the technical changes more than back that claim up.
Click to expand...

Anyone more technically informed about transistor design than me care to speculate what Intel might have dropped/relaxed?
 
Well the first thing I would imagine has gone out of the door is the touted 2.7x density scaling.
 
A comment on Charlie's Twitter is saying that Intel targeted a minimum metal pitch of <40 nm, something that the rest of the industry suggest won't be possible without EUV. Incidentally Samsung has the same target, except they already plan on using EUV.
 
Charlie said Intel is going to introduce a 14nm+++ process which didn't happen and won't happen. He is a clown.
 
Most likely relaxing the layers a bit, especially the one that's 36 to at least 40, and changing the layers which are using SAQP (or worse?) back down to double. And the cobalt use is probably being reduced. The actual density hit from doing this remains to be seen, but Charlie is likely exaggerating that part.

mikk said:
Charlie said Intel is going to introduce a 14nm+++ process which didn't happen and won't happen. He is a clown.
Click to expand...

Tough to say that without knowing what Comet and Cooper are using.
 
Interesting. Maybe the shift from prosperity to not so prosperous will happen sooner than anticipated. The next year or two should be interesting.
 
Kenmitch said:
Interesting. Maybe the shift from prosperity to not so prosperous will happen sooner than anticipated. The next year or two should be interesting.
Click to expand...
It's what puts the fire under people's feet to innovate and be competitive again.
 
Kenmitch said:
And what makes the weak throw in the towel, crumble, go postal, etc. Who knows maybe some will defect and go elsewhere.

Time will tell.
Click to expand...
Intel... will be fine. They aren't weak. They just got fat and lazy in prosperous high margin low innovation times. With the fire now under their feet and their margins/market share under attack, they'll cut the fat and become great again. Hopefully their pricing will too.

Time will tell. Until then, I'm with the prior underdog that made a turn around in tough times : AMD.
 
If coffeelake refresh is really hitting those clocks, then it has to be on 14+++ imo.
Problem is they are not going to reach the same frequency with icelake, so hopefully they have some IPC gains.
 
In the past, with the Q6600 for example, processor steppings allowed higher clock frequencies with no changes to the process node. Coffee lake refresh could be the same.
 
There is also the elephant in the room. Intel's architecture is simply incompatible with shrinking FinFETs. 22-nm FinFETs had high intrinsic perf over FDSOI/Bulk, but as the shrinks go on FinFET is losing performance. In turn, each node after 22-nm required more and more complex perf enhancers.

10-nm is the grave of the old core. (essentially, delay was twofold: improve architecture extrinsic boost and improve finfet intrinsic boost.)
7-nm is the beginning of a new core. (The delay on 10-nm does not cause a delay on 7-nm. So, 2017 (10-nm ramp) and 2019 (7-nm ramp) is held.)

7-nm will most likely be FinFETs and on either EUVL or JFIL. If JFIL there will be a drop of EUV backorders. Got dibs on NZ4C which is a 450mm J-FIL machine slotted around 300mm High NA EUV introduction. J-FIL is more in line with Intel's wants currently for future logic nodes.
 
Last edited:
moonbogg said:
I wish Intel would just tell us straight up like, "Ice Lake will have 5% more IPC and -10% clock speed. Buy i9 9900K now or something".
Click to expand...

Given all the scrambling around, I doubt they know what the clocks are possible at this point.

Hell, I wouldn't assume Icelake on desktop will even be released and not just replaced by Comet Lake or even Coffee Lake Refresh Refresh. If they end up doing another 14nm revision which manages even higher clocks, it might be better for them to do it just from a marketing perspective than say any IPC gains Icelake gives.
 
14nm++ vs 14nm+++. If my experience with Devil's Canyon is any guide, Intel should be able to get higher performance from simply working the kinks out. Better, if they solder the IHS, the resultant i9 will sustain much higher clocks.
 
Kenmitch said:
Interesting. Maybe the shift from prosperity to not so prosperous will happen sooner than anticipated. The next year or two should be interesting.
Click to expand...
Intel will remain insanely rich for years to come. They have more fab capacity than the entire world combined, even if AMD keeps a clear lead (Which isn't a sure thing), Intel will still have a huge market simply because they're the only ones that can meet demand.

It would take at minimum 5 years for even a giant like TSMC to prepare enough capacity to fight it out with Intel.

firewolfsm said:
In the past, with the Q6600 for example, processor steppings allowed higher clock frequencies with no changes to the process node. Coffee lake refresh could be the same.
Click to expand...
There are multiple layers to what you said. First of all, node improvements aren't a new thing for either Intel or AMD over the lifetime of a processor. It was just never advertised and given a + moniker. That's one of the benefits of being an IDM (which AMD lost). You can continuously improve and modify the process to fit your exact needs. Those new stepping no doubt had manufacturing improvements and not just design behind them.

Second, a new stepping only gives you more frequency if there was some way to optimize the design you didn't have time or budget to do before. Kaby Lake and Coffee Lake are both so insanely physically optimized at this point, wringing anything more with the same node would be, uh, a fool's errand shall we say.
 
Last edited:
You must log in or register to reply here.

TRENDING THREADS

Back
Top