Intel expects 7nm to be ready in two years

[WYP]

10nm will be a short-lived node for Intel.



Read more about Intel's plans to make 7nm production-ready in two years.

 

[NeverBackDown]

Honestly I would have gave up on 10nm put all the money into 7nm to speed that up or instead use that 10nm to speed up the design of the next architecture of whatever they were planning on. Would have saved them a lot of wasted time. I doubt 10nm will pay for itself with so much money going into it and so little time to sell.

 

[grec]

Yeah they only kept 10nm to save face to investors. Engineers were leaking that there were major issues with the node as far back as 2015, and once 2016 came along and it started to seem like EUV could be just about economical this decade there was a lot of internal pressure to drop 10nm's and move on to their EUV successor/7nm, but apparently it was seen as a PR disaster that would have too much of an impact on stock performance so they trimmed it down, massively cut projected production, cut out loads of products set to launch on it, moved more or less all major development to 7nm and products that use it and have left 10nm as a gap filler for certain specific markets with less than a quarter of the initial fab space committed to it.

 

[looz]

I wonder if the old rule still applies, as in Intel's manufacturing notation is more stringent and their 7nm is going to be "smaller" than others'

 

[grec]

What Intel calls 7nm will still be smaller than what everyone else calls 7nm because everyone still uses roughly the same naming scheme. However their 7nm launches when everyone else's 5nm does and will be their first EUV generation vs everyone else's 2nd or 3rd.

 

[AngryGoldfish]

Yeah they only kept 10nm to save face to investors. Engineers were leaking that there were major issues with the node as far back as 2015, and once 2016 came along and it started to seem like EUV could be just about economical this decade there was a lot of internal pressure to drop 10nm's and move on to their EUV successor/7nm, but apparently it was seen as a PR disaster that would have too much of an impact on stock performance so they trimmed it down, massively cut projected production, cut out loads of products set to launch on it, moved more or less all major development to 7nm and products that use it and have left 10nm as a gap filler for certain specific markets with less than a quarter of the initial fab space committed to it.

14nm was supposed to be a stop-gap itself, wasn't it? It was aimed originally at the mobile market. 10nm was supposed to be their big guns and the true followup to 22nm or whatever it was they were using.

 

[NeverBackDown]

14nm was supposed to be a stop-gap itself, wasn't it? It was aimed originally at the mobile market. 10nm was supposed to be their big guns and the true followup to 22nm or whatever it was they were using.

Honestly, it's been so long I have no idea anymore. LOL ^_^
Probably take a few hours just to find out if it was or not it was that long ago.

 

[WYP]

Well, 14nm did start with Broadwell, which barely had a desktop release. There is a reason Haswell got a re-release as Devils Canyon.

TBH it is similar to the 10nm rollout. Mobile first, bigger things later.

 

[NeverBackDown]

I think Intel need the bigger things now and mobile later. The big boy chips are where the money it at for them.

 

[WYP]

I think Intel need the bigger things now and mobile later. The big boy chips are where the money it at for them.

Yes, but mobile is where the early yields are at. After that then the big chips. It makes more financial sense to work your way up than the other way round.

I do agree that Intel needs big chips though. Mobile first also gives Intel the lead before 7nm Ryzen mobile chips come.

 

[NeverBackDown]

Well yields yes, but if the yields are already so low, why would you make the small chips first? Why not make more with the little you got? Makes more sense.

If they simply cannot handle the bigger dies, then sure. But that also signals the risk of bad low-end mobile dies too.

 

[Kleptobot]

Well yields yes, but if the yields are already so low, why would you make the small chips first? Why not make more with the little you got? Makes more sense.

If they simply cannot handle the bigger dies, then sure. But that also signals the risk of bad low-end mobile dies too.

Die size affects yields.
If you have defects per area that are roughly evenly distributed
The chance of having a 10mm2 patch unaffected vs a 50mm2 patch is 25 times more likely

 

[Warchild]

I think Intel need the bigger things now and mobile later. The big boy chips are where the money it at for them.

Well, I wonder if their interest in the GPU market slowed down their 10nm progress.

I think I am done with incremental expectancy. I have a decent setup now. I'll run it into the ground and do a full build in 3 or 4 years time, as opposed to constantly wasting money on bits n pieces all the time.

Last 2 years have been nothing but disappointment in terms of progress. Ok Zen was a success but I have Intel, and im not swapping everything out for something that yields minimal improvement.

 

[WYP]

Well yields yes, but if the yields are already so low, why would you make the small chips first? Why not make more with the little you got? Makes more sense.

If they simply cannot handle the bigger dies, then sure. But that also signals the risk of bad low-end mobile dies too.

Think of silicon defects as a number of defects per silicon wafer. If you use a wafer to make a tonne of small dies, a smaller percentage of the dies will be ruined by defects. This is why small dies achieve better yields.

 

[grec]

Yeah and these defects, if not large enough to damage the chip functionally, will often cause transition delays in logic gates, which reduces the maximum possible clock speed you can hit, making them suitable for mobile parts in that aspect too. Of course even with no defects at all, smaller dies are also significantly more efficient use of silicon wafers because you're fitting squares into a circle and smaller squares will leave you with less waste around the edges. Also worth remembering mobile is actually Intel's largest market by revenue and volume at the moment.