New Zen microarchitecture details

Page 80 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

cytg111

Lifer
Mar 17, 2008
23,049
12,719
136
Optimistic : Thats fantastic, cant wait to hear more.
Pessimistic : Just shows its ISA compliant. Lets hope it doesnt crash.
 

KTE

Senior member
May 26, 2016
478
130
76
Optimistic : Thats fantastic, cant wait to hear more.
Pessimistic : Just shows its ISA compliant. Lets hope it doesnt crash.
Doom isn't CPU intensive and a benchmark even PD does far better than usual in.

Sent from HTC 10
(Opinions are own)
 

cytg111

Lifer
Mar 17, 2008
23,049
12,719
136
So they demoed a working sample of Zen that didnt crash running DOOM. It is not nothing.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
BTW is it normal to have no labels on the ES chips at all? :sneaky:

ALL of the lasering missing (completely blank)?
I've had such production silicon, but they are for special purpose only. It can be used as a safety measure, since unless you have the right equipment and info, the CPU is nothing but a paper weight :sneaky:
 

Dresdenboy

Golden Member
Jul 28, 2003
1,730
554
136
citavia.blog.de
ALL of the lasering missing (completely blank)?
I've had such production silicon, but they are for special purpose only. It can be used as a safety measure, since unless you have the right equipment and info, the CPU is nothing but a paper weight :sneaky:
Yep, like this (from the video):
QKMk3mc.jpg
 

Doom2pro

Senior member
Apr 2, 2016
587
619
106
Thanks.

If you look where the lady is putting the trays into, it becomes pretty obvious why they are blank chips ;)

Unfused and unmarked parts about to be fused and marked.

Are they binned before that part or are they benched at that stage, then fused and labeled?
 
Mar 10, 2006
11,715
2,012
126
Thanks for adding to the conversation with insightful facts and knowledge.

IBM's processes, including 14nm SOI, are typically quite exotic. The wafers tend to be expensive, yields aren't great, and so on. This is one of the reasons that IBM never really gained much traction as a general purpose foundry, and why they ultimately paid GloFo to take the fabs.

GloFo, as part of a wafer supply agreement with IBM, will continue to run IBM's exotic processes in its newly acquired fabs for the POWER chips. But it is very unlikely that third party fabless customers will want in on the fabs.

A good clue here is that in recent presentations around OpenPOWER, IBM and its partners have talked about POWER-derivative chips from partners being built by third party manufacturing houses rather than on IBM's old fabs now owned by GloFo.

Anyway, no way AMD is going to use these expensive, low-yielding, and probably difficult to design for processes for future CPUs or GPUs.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
22nm FDSOI is supposedly already cheaper per transistor (not necessarily taking into account design complexity) than various finFET processes from the "next generation" of nodes. It shouldn't take that long for yield issues associated with 14nm HP finFET to even out simply from repeated iteration, to the point where AMD could seriously consider it as a "step up" node to use after 14nm LPP.
 

Glo.

Diamond Member
Apr 25, 2015
5,661
4,419
136
IBM's processes, including 14nm SOI, are typically quite exotic. The wafers tend to be expensive, yields aren't great, and so on. This is one of the reasons that IBM never really gained much traction as a general purpose foundry, and why they ultimately paid GloFo to take the fabs.

GloFo, as part of a wafer supply agreement with IBM, will continue to run IBM's exotic processes in its newly acquired fabs for the POWER chips. But it is very unlikely that third party fabless customers will want in on the fabs.

A good clue here is that in recent presentations around OpenPOWER, IBM and its partners have talked about POWER-derivative chips from partners being built by third party manufacturing houses rather than on IBM's old fabs now owned by GloFo.

Anyway, no way AMD is going to use these expensive, low-yielding, and probably difficult to design for processes for future CPUs or GPUs.
http://www.eetimes.com/author.asp?section_id=36&doc_id=1329887
I suppose you have not read this article on the matter of production costs between processes.

I think that AMD will use this(FD-SOI) process for both Zen AND Vega GPUs.
 
Mar 10, 2006
11,715
2,012
126
I think that AMD will use this(FD-SOI) process for both Zen AND Vega GPUs.

I doubt it. See:

“FinFET technology is expected to play a critical foundational role across multiple AMD product lines, starting in 2016,” said Mark Papermaster, senior vice president and chief technology officer at AMD. “GLOBALFOUNDRIES has worked tirelessly to reach this key milestone on its 14LPP process. We look forward to GLOBALFOUNDRIES' continued progress towards full production readiness and expect to leverage the advanced 14LPP process technology across a broad set of our CPU, APU, and GPU products.”

http://www.globalfoundries.com/news...logy-success-for-next-generation-amd-products
 
Mar 10, 2006
11,715
2,012
126
That article is terrible!!
As one responder points out, 14nm FD-SOI is still in development - so it's yields are not yet known (which invalidates any cost metric).

IBM's 14nm FinFET process that will be used to manufacture POWER9 uses SOI wafers, IIRC.