[EETimes] TSMC starting 5 nm EUV Risk Production in April?

jpiniero

Diamond Member
Oct 1, 2010
6,193
174
126
#1
https://www.eetimes.com/document.asp?doc_id=1333827

It's only risk production, but unlike the 7+ node this is fully EUV like Samsung's 7 nm node.

TSMC said that N5 will deliver 14.7% to 17.7% speed gains and 1.8 to 1.86 area shrinks based on tests with Arm A72 cores. The N7+ node can deliver 6% to 12% less power and 20% better density; however, TSMC did not mention speed gains.
 

Vattila

Senior member
Oct 22, 2004
363
62
136
#2
From the article:

"TSMC announced that it taped out a customer chip in an N7+ node that can use EUV on up to four layers. Its N5 that will use EUV on up to 14 layers will be ready for risk production in April. EUV aims to lower costs by reducing the number of masks required for leading-edge designs."
 
Feb 23, 2017
461
347
106
#3
The quote in the OP is puzzling me.
The first thing it mentions is speed gains, and the last thing it mentions are that speed gains weren't mentioned.
My head is in a boot loop following that BSOD.
 

french toast

Senior member
Feb 22, 2017
919
5
106
#4
The quote in the OP is puzzling me.
The first thing it mentions is speed gains, and the last thing it mentions are that speed gains weren't mentioned.
My head is in a boot loop following that BSOD.
Speed gains for N5 is ~15%....no speed gains mentioned for N7+...where is the confusion?
 

krumme

Diamond Member
Oct 9, 2009
5,708
75
136
#5
A13 2019 h2 on 7+ then? A14 2020 h2 on 5nm?
Amd x86 half a year later than Apple.
Tock tock tock tock. Relentlessly.
It's perfectly clear Intel business model is not working today. In no way can they compete with the rest of the world sharing process cost.
They need to change something radical now or they will be stuck on a half mature complex 10nm while the rest of the leader pack is on 5nm euv. It's not viable.
 
Apr 27, 2000
11,201
701
126
#6
Intel may get some good transistor performance out of 10nm, but they will lose on cost and density.
 

DooKey

Golden Member
Nov 9, 2005
1,464
12
106
#7
A13 2019 h2 on 7+ then? A14 2020 h2 on 5nm?
Amd x86 half a year later than Apple.
Tock tock tock tock. Relentlessly.
It's perfectly clear Intel business model is not working today. In no way can they compete with the rest of the world sharing process cost.
They need to change something radical now or they will be stuck on a half mature complex 10nm while the rest of the leader pack is on 5nm euv. It's not viable.
I believe Intel is working on 7nm EUV process as we speak. That should catch them up if they can execute. Big IF these days.
 

krumme

Diamond Member
Oct 9, 2009
5,708
75
136
#8
I believe Intel is working on 7nm EUV process as we speak. That should catch them up if they can execute. Big IF these days.
Intel current profit at approx 20b a year gives plenty safety and funds for 10 and 7nm. But when 5nm arrives I think the size of investments even for Intel starts to become a serious problem for profit. 3nm funding could perhaps be unsustainable. Process cost is just exploding and always takes longer than anticipated.
 

DooKey

Golden Member
Nov 9, 2005
1,464
12
106
#9
Intel current profit at approx 20b a year gives plenty safety and funds for 10 and 7nm. But when 5nm arrives I think the size of investments even for Intel starts to become a serious problem for profit. 3nm funding could perhaps be unsustainable. Process cost is just exploding and always takes longer than anticipated.
At that point I think Intel opens up their fabs to a multitude of others and garners some outside capital. They aren't the only ones looking at a huge bill to continue down the shrink path. The next several years are going to be fascinating.
 

scannall

Golden Member
Jan 1, 2012
1,333
147
136
#10
And to think, all this capital infusion into foundries really started when Intel told Apple no, they weren't going to make a SOC for their iPhone project. ;-) If Intel had said yes, the landscape would look a lot different.
 
Sep 11, 2010
61
7
91
#11
A13 2019 h2 on 7+ then? A14 2020 h2 on 5nm?
Amd x86 half a year later than Apple.
Tock tock tock tock. Relentlessly.
It's perfectly clear Intel business model is not working today. In no way can they compete with the rest of the world sharing process cost.
They need to change something radical now or they will be stuck on a half mature complex 10nm while the rest of the leader pack is on 5nm euv. It's not viable.
Maybe also an A13X on N5 in 2020H1?
Apple seems all over the place in why and when they release X variants (presumably A11X was skipped and A12X will appear soon this month?).
But skipping the A13X in October next year and delaying it to 2020H1 so that it can be the early exploration of N5 seems to make sense, like A10X was first on 10nm. Also gives iPads an average cadence of 18 months or so rather than annual, which seems to kinda work.
 

maddie

Platinum Member
Jul 18, 2010
2,446
379
136
#13
"TSMC said that N5 will deliver 14.7% to 17.7% speed gains"

A rather strange number. Precisely imprecise estimate to 1 significant digit?
 

Dolan

Junior Member
Dec 25, 2017
13
0
41
#14
There are multiple ARM and other IP cores already implemented in pretty mature stage so very accurate results are already available. Although, these numbers might change in future re-spins (oriented for price and manufacturing economics).

Btw.: it is impressive that this huge area jump is still possible at "end of moores law". It is even bigger jump than what Intel plans at transition from 10 to 7.

BtwBtw.: I don't work for Apple but i have basically confirmed that Applle choosed DUV version instead of EUV. ...for this year Iphone. For EUV they will go directly at 5nm.
 

jpiniero

Diamond Member
Oct 1, 2010
6,193
174
126
#15
BtwBtw.: I don't work for Apple but i have basically confirmed that Applle choosed DUV version instead of EUV. ...for this year Iphone.
TSMC already kind of confirmed that when they said they weren't expecting any material revenue from 7+ this year. 7+ and 5 seem to be coming at the same time, which seems strange until you realize the bottleneck is EUV.
 

Ajay

Diamond Member
Jan 8, 2001
5,006
126
136
#16

fluxdeity

Junior Member
Apr 5, 2019
5
0
6
#18
Too bad Intel's Core series of processors is just refreshed from Core 2 Duo which is refreshed from Pentium 4. AMD has left the old Bulldozer and Piledrivers behind and made a new line that wasn't just a refresh. Zen and Zen+ were the first blows, Zen 2(3000 series) and beyond will be the fall of Intel.
 

Tuna-Fish

Senior member
Mar 4, 2011
949
147
136
#19
Too bad Intel's Core series of processors is just refreshed from Core 2 Duo which is refreshed from Pentium 4
This is not true.

Sandy Bridge (Core iX 2XXX) was a complete redesign, on a similar level to what Zen 2 is. Intel just didn't market it as such, and the performance gain over it's predecessors wasn't that dramatic, so only people who were watching closely noticed. Everything after SNB is incremental updates to the same base, without radical changes.

Before that, the Conroe (Core 2 Duo) was derived from the P6 base, that is, from Pentium 3 CPUs. Pentium 4 was a radical redesign that didn't work out well (much like AMD's Bulldozer), so Core 2 Duo was created from the base of it's predecessor instead.
 

jpiniero

Diamond Member
Oct 1, 2010
6,193
174
126
#20
I should add that I think the foundries making progress on EUV is what finally got Intel to more or less give up on trying to really fix 10 nm and get moving on their own EUV node. When you might actually see EUV from Intel remains to be seen.
 
Dec 6, 2018
53
36
51
#22
"TSMC said that N5 will deliver 14.7% to 17.7% speed gains"

A rather strange number. Precisely imprecise estimate to 1 significant digit?
maybe theoretically 16.2% + -1.5%
 

Ajay

Diamond Member
Jan 8, 2001
5,006
126
136
#23
I should add that I think the foundries making progress on EUV is what finally got Intel to more or less give up on trying to really fix 10 nm and get moving on their own EUV node. When you might actually see EUV from Intel remains to be seen.
It seems to me that Intel 7nm was going to use EUV all along and is being developed by a separate team from the 10nm group: https://www.anandtech.com/show/13683/intel-euvenabled-7nm-process-tech-is-on-track

As far as backing off on 10nm development; with 7nm on schedule, it didn’t make any sense to further delay what would be a short lived 10nm node.
 

jpiniero

Diamond Member
Oct 1, 2010
6,193
174
126
#24
2021 (Intel 7nm). When in 2021 is anyone's guess.
It would make a ton of sense if the public availability of Cooper Lake was a year from now, and 7 nm Granite Rapids two years from now. I don't know if that's realistic.
 


ASK THE COMMUNITY

TRENDING THREADS