• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."
  • Recovering from Halloween? Take part in our Spooky Giveaway to be in with a chance of replenishing your wardrobe. A simple riddle stands in your way. Visit the sweeptake thread over here.

Globalfoundries 7LP 7nm Leading Performance FINFET process and FX-7 ASIC platform

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

nad-

Junior Member
May 4, 2017
12
1
11
GlobalFoundries has only access to 14LPe and 14LPP. The Globalfoundries 14nm+ node is HPP. Not to be confused as a better version of 14HP, which is SOI FinFETs. 14HPP which is Bulk FinFETs and is basically 14LPP with 7LP-styled transistors, MOL, etc, and such has higher performance.
I'll require a source on that, that would imply AMD's 14nm+ "refreshes" (Vega and Zen) are going to be quite a bit different than I expected it.
Unless AMD is using the "14nm+" they used for Polaris 20.

EDIT: Now that I think about it, Raven Ridge might be using "14nm+" as well considering how late into the year it is.
 
  • Like
Reactions: ajc9988

raghu78

Diamond Member
Aug 23, 2012
4,002
1,232
136
I'll require a source on that, that would imply AMD's 14nm+ "refreshes" (Vega and Zen) are going to be quite a bit different than I expected it.
Unless AMD is using the "14nm+" they used for Polaris 20.

EDIT: Now that I think about it, Raven Ridge might be using "14nm+" as well considering how late into the year it is.
Yes. Raven Ridge in late q4 2017, Pinnacle Ridge in q1 2018, Vega 11 in q1 2018 and a Vega 10 refresh in mid 2018 are likely to use 14nm+.

Sent from my SM-G935V using Tapatalk
 
  • Like
Reactions: ajc9988

ao_ika_red

Golden Member
Aug 11, 2016
1,397
541
106
GlobalFoundries has only access to 14LPe and 14LPP. The Globalfoundries 14nm+ node is HPP. Not to be confused as a better version of 14HP, which is SOI FinFETs. 14HPP which is Bulk FinFETs and is basically 14LPP with 7LP-styled transistors, MOL, etc, and such has higher performance.
Wow, this is new to me, I never heard that GloFo is actually fine-tuning its 14nm node. What I heard was GloFo licensed all of their 14nm from Samsung because their own 14XM node had a handful of delay so they scrap it altogether and used Samsung's instead.
 
Last edited:

Lodix

Senior member
Jun 24, 2016
286
73
71
Wow, this is new to me, I never heard that GloFo is actually fine-tuning its 14nm node. What I heard was GloFo licensed all of their 14nm from Samsung because their own 14XM node had a handful of delay so they scrap it altogether and used Samsung's instead.
I don't think he is basing that on an official statement from AMD or SS.... and he is wrong about 14nmLPE because AMD didn't license it. And I remember reading someone posting the rx580 description talking about using a third generation 14nm FinFet process which probably is 14nmLPC.

EDIT: yeah, you can check on the official AMD page mentioning third generation 14nm FinFet.
 
  • Like
Reactions: ajc9988 and Tee9000

dark zero

Platinum Member
Jun 2, 2015
2,525
95
91
Yes. Raven Ridge in late q4 2017, Pinnacle Ridge in q1 2018, Vega 11 in q1 2018 and a Vega 10 refresh in mid 2018 are likely to use 14nm+.

Sent from my SM-G935V using Tapatalk
Also Raven Ridge is Zen +...
 

NostaSeronx

Platinum Member
Sep 18, 2011
2,641
403
126
Here is linkedin profiles just for reference:
https://www.linkedin.com/in/hyucksoo-yang-04068711
- 14nm Process Integration (14LPE / 14LPP / 14LPP+)

https://sg.linkedin.com/in/gyana-sahoo-33294015
- Working on Different GF technologies like 65LPe, 55LPe, 40LP, 28 (HPP, LPQ, SLP, LPS), 22FDX, 14(LPP, HPP, HP)

14LPP+ and 14HPP are the same node.

The Cobalt MOL, SSRW, etc was always planned for 14nm nodes... via the whole 20nm and beyond. (14nm is essentially 20nm @ GlobalFoundries)
https://www.linkedin.com/in/jiehui-shu-jeff-shu-63455b26
- Process owner of Cobalt Cap process for 20nm and beyond BEOL EM improvement

SSRW is a bit harder to find:
https://www.researchgate.net/publication/261024876_Low_temperature_ISSG_oxidation_and_its_application_in_SSRW_for_20nm_and_below_semiconductor_devices
http://www.rit.edu/kgcoe/eme/sites/default/files/Min-hwa Chi - abstract_ Challenges in Manufacturing FinFET.pdf
^-- patents of this (The present disclosure enjoys industrial applicability in any of various types of highly integrated semiconductor devices for 22 nm technology products and beyond, particularly for 14 nm and 20 nm technology bulk CMOS-based products and beyond.)
http://www.eetimes.com/document.asp?doc_id=1262902 (Instead of 14XM replace it with 14LPP/FX-14)

GlobalFoundries and AMD have done this secret node stuff before...
-> 28 SHP(Kaveri) / 28 A(Carrio) / 28 HPA(Bristol/Stoney)
For example, when the website only details 28SLP/28HPP, etc.

If anything the 14HPP node is aimed at the 12nm FinFET node @ TSMC.
 
  • Like
Reactions: CatMerc and Tee9000

ao_ika_red

Golden Member
Aug 11, 2016
1,397
541
106
Here is linkedin profiles just for reference:
https://www.linkedin.com/in/hyucksoo-yang-04068711
- 14nm Process Integration (14LPE / 14LPP / 14LPP+)

https://sg.linkedin.com/in/gyana-sahoo-33294015
- Working on Different GF technologies like 65LPe, 55LPe, 40LP, 28 (HPP, LPQ, SLP, LPS), 22FDX, 14(LPP, HPP, HP)

14LPP+ and 14HPP are the same node.

The Cobalt MOL, SSRW, etc was always planned for 14nm nodes... via the whole 20nm and beyond. (14nm is essentially 20nm @ GlobalFoundries)
https://www.linkedin.com/in/jiehui-shu-jeff-shu-63455b26
- Process owner of Cobalt Cap process for 20nm and beyond BEOL EM improvement

SSRW is a bit harder to find:
https://www.researchgate.net/publication/261024876_Low_temperature_ISSG_oxidation_and_its_application_in_SSRW_for_20nm_and_below_semiconductor_devices
http://www.rit.edu/kgcoe/eme/sites/default/files/Min-hwa Chi - abstract_ Challenges in Manufacturing FinFET.pdf
^-- patents of this (The present disclosure enjoys industrial applicability in any of various types of highly integrated semiconductor devices for 22 nm technology products and beyond, particularly for 14 nm and 20 nm technology bulk CMOS-based products and beyond.)
http://www.eetimes.com/document.asp?doc_id=1262902 (Instead of 14XM replace it with 14LPP/FX-14)

GlobalFoundries and AMD have done this secret node stuff before...
-> 28 SHP(Kaveri) / 28 A(Carrio) / 28 HPA(Bristol/Stoney)
For example, when the website only details 28SLP/28HPP, etc.

If anything the 14HPP node is aimed at the 12nm FinFET node @ TSMC.
Many thanks for the round-up. So we'll see another 14nm refresh from GloFo in the end of 2017 / early 2018 which should be better than their first revision early this year. I hope it will give AMD's much needed clock boost to all of their products.
 
  • Like
Reactions: ajc9988

Spartak

Senior member
Jul 4, 2015
298
191
116

CatMerc

Golden Member
Jul 16, 2016
1,114
1,145
106
RX 580 still uses 14nm LPP according to Anandtech's article. It's not HPP.
 

ajc9988

Senior member
Apr 1, 2015
224
142
116
No it isnt. You're posting as fact something that was wild, completely unsupported speculation by raghu78 in this thread: https://forums.anandtech.com/threads/zen-2-based-raven-ridge-and-pinnacle-ridge.2501770/

Ryzen lanched in 17Q2, so even with a yearly tock the raven ridge which will be released in 17Q4 for mobile is squarely part of the 1st gen Zen, just near the end of the cycle.
Yes, but it makes more sense. Su has said tape out Q4 2017 for Zen 2 7nm and suggested as early as Q3 2018. Even if we look to Q4 2018-Q1 2019, that is a short time. Glofo and AMD have worked on 14nm since 2015 and have already themselves suggested 14nm+ products in Q4 2017 and Q1 2018. If the 14nm+ brings further power reduction capabilities, it makes sense to switch, thereby being more competitive with Intel's offerings.
 

ao_ika_red

Golden Member
Aug 11, 2016
1,397
541
106
RX 580 still uses 14nm LPP according to Anandtech's article. It's not HPP.
Because RX5xx is using refined 14LPP, that's why there's not much gain from the Polaris refresh. GloFo somehow made a mess in adopting Samsung's LPP, they had to refine it along the way.
Based on NostaSeronx data, 14HPP will see the light in upcoming Zen (CPU and APU) and Vega refresh.
 
  • Like
Reactions: ajc9988

dark zero

Platinum Member
Jun 2, 2015
2,525
95
91
No it isnt. You're posting as fact something that was wild, completely unsupported speculation by raghu78 in this thread: https://forums.anandtech.com/threads/zen-2-based-raven-ridge-and-pinnacle-ridge.2501770/

Ryzen lanched in 17Q2, so even with a yearly tock the raven ridge which will be released in 17Q4 for mobile is squarely part of the 1st gen Zen, just near the end of the cycle.
So that's it then. However there must be some changes on it, at least an step B2 or something that fixes the bugs present on Ryzen.
 

ajc9988

Senior member
Apr 1, 2015
224
142
116
So that's it then. However there must be some changes on it, at least an step B2 or something that fixes the bugs present on Ryzen.
Most things called "bugs" are not bugs, rather software is not yet optimized on it.
 

moinmoin

Golden Member
Jun 1, 2017
1,049
845
106
Ryzen lanched in 17Q2, so even with a yearly tock the raven ridge which will be released in 17Q4 for mobile is squarely part of the 1st gen Zen, just near the end of the cycle.
Raven Ridge being 1x CCX will be only the second Zeppelin die in existence, the first being a die that was primarily designed for use in Naples/Epyc and then scaled down for mainstream consumer market. The Raven Ridge die will be optimized for mobile use, at least all server/MCM specific I/O with no use in an APU will be taken out. (I guess this is more a question of definition whether Zen+ will contain changes to the microarchiteture itself or just the uncore akin to e.g. Skylake-X. I expect only the latter which would explain why there won't be a Zen+ for Epyc etc.)
 

raghu78

Diamond Member
Aug 23, 2012
4,002
1,232
136
No it isnt. You're posting as fact something that was wild, completely unsupported speculation by raghu78 in this thread: https://forums.anandtech.com/threads/zen-2-based-raven-ridge-and-pinnacle-ridge.2501770/

Ryzen lanched in 17Q2, so even with a yearly tock the raven ridge which will be released in 17Q4 for mobile is squarely part of the 1st gen Zen, just near the end of the cycle.
http://hexus.net/tech/news/cpu/105772-amd-unveils-roadmap-ryzen-threadripper-ryzen-3-apus/

AMD has confirmed Zen being built at 14nm and 14nm+ . The 2017 products are all using 14nm aka 14LPP. The 2018 products are using 14nm+ . Raven Ridge is basically a 2018 product. The notebook version launches late in Q4 2017 and the desktop version in Q1 2018 along with Pinnacle Ridge.

2017 - Summit Ridge, ThreadRipper. - GF 14LPP
2018 - Raven Ridge, Pinnacle Ridge, ThreadRipper (using Pinnacle Ridge dies) - GF 14HPP (or whatever the name for 14nm+)
2019 - 7nm Zen 2 based CPUs - GF 7LP
 
  • Like
Reactions: ajc9988

Spartak

Senior member
Jul 4, 2015
298
191
116
http://hexus.net/tech/news/cpu/105772-amd-unveils-roadmap-ryzen-threadripper-ryzen-3-apus/

AMD has confirmed Zen being built at 14nm and 14nm+ . The 2017 products are all using 14nm aka 14LPP. The 2018 products are using 14nm+ . Raven Ridge is basically a 2018 product. The notebook version launches late in Q4 2017 and the desktop version in Q1 2018 along with Pinnacle Ridge.

2017 - Summit Ridge, ThreadRipper. - GF 14LPP
2018 - Raven Ridge, Pinnacle Ridge, ThreadRipper (using Pinnacle Ridge dies) - GF 14HPP (or whatever the name for 14nm+)
2019 - 7nm Zen 2 based CPUs - GF 7LP
Raven Ridge is a 2017 product. And nowhere does AMD assert every 'new' product shipped in 2018 will use 14nm+, let alone late 17Q4. Your linked article doesnt support your claim in any possible way outside the realm of your wild fantasies.
 

ajc9988

Senior member
Apr 1, 2015
224
142
116
Raven Ridge is a 2017 product. And nowhere does AMD assert every 'new' product shipped in 2018 will use 14nm+, let alone late 17Q4. Your linked article doesnt support your claim in any possible way outside the realm of your wild fantasies.
https://i.rbt.asia/g/image/1492/47/1492476182306.png
It was originally listed as a 2018 product and is listed with 14nm+ products, increasing the likelihood of being a 14nm+ product.
 
  • Like
Reactions: raghu78

Spartak

Senior member
Jul 4, 2015
298
191
116
That's the socket version of Raven Ridge that will be released in 2018. Not 'originally', still. If you think they'll use an improved version of their process for the socketed low volume desktop version of Raven Ridge, and thus having two different versions of Raven Ridge right out of the gate, you have completely lost your mind.

But my guess is you didnt notice the "socketed" part of that roadmap.
 

ajc9988

Senior member
Apr 1, 2015
224
142
116
That's the socket version of Raven Ridge that will be released in 2018. If you think they'll use an improved version of their process for the socketed low volume desktop version of Raven Ridge, and thus having two different versions of Raven Ridge right out of the gate, you have completely lost your mind.

But my guess is you didnt notice the "socketed" part of that roadmap.
I don't think they'll use an improved version. I'm saying that you would use the same process and that the mobile will be one of the first on that process. How hard is it for you to understand that?
 
  • Like
Reactions: raghu78

Spartak

Senior member
Jul 4, 2015
298
191
116
I don't think they'll use an improved version. I'm saying that you would use the same process and that the mobile will be one of the first on that process. How hard is it for you to understand that?
You just showed me a roadmap that is still current and does nothing to disprove my two arguments you replied to. Instead of trying to paint me as some retard why dont you just show me proof to what you claimed? Since you know...your roadmap refuted your own claim.
 

ajc9988

Senior member
Apr 1, 2015
224
142
116
You just showed me a roadmap that is still current and does nothing to disprove my two arguments you replied to. Instead of trying to paint me as some retard why dont you just show me proof to what you claimed? Since you know...your roadmap refuted your own claim.
It didn't refute it and I didn't say you were fully incorrect, nor that I was correct. You, however, did attack me and call me an idiot, assuming I thought they would be on different processes. What I showed is that, considering it will be on the same process for both versions, and that the sockets version is listed among 14nm+ designs, but NOT listed as a 14nm+ design, that neither can be confirmed, meaning you are wrong speaking in absolute terms, as is anyone speaking of it absolutely being 14nm+, unless you are an insider, which then you know, but are likely violating many NDAs.

Also, that roadmap didn't refute my claim. Look at my hedging language discussing likelihood.
 
  • Like
Reactions: Lodix

Spartak

Senior member
Jul 4, 2015
298
191
116
meaning you are wrong speaking in absolute terms
you are speculating and connecting the dots between two processors listed in 2018, as if both should be 14nm+, and I'm the one speaking in absolute terms? I'm saying there's no proof beyond wild speculation.

How hard is it for you to understand that?
...I'm the one accusing you of being an idiot?

Also, that roadmap didn't refute my claim.
"It was originally listed as a 2018 product and is listed with 14nm+ products"

I rest my case with you sir, this is my last reply. If you continue in this fashion I will report you for trolling.
 

raghu78

Diamond Member
Aug 23, 2012
4,002
1,232
136
you are speculating and connecting the dots between two processors listed in 2018, as if both should be 14nm+, and I'm the one speaking in absolute terms? I'm saying there's no proof beyond wild speculation.



...I'm the one accusing you of being an idiot?



"It was originally listed as a 2018 product and is listed with 14nm+ products"

I rest my case with you sir, this is my last reply. If you continue in this fashion I will report you for trolling.
In absolute terms you know nothing about the process Raven Ridge is built on and neither do I. What we do know is Zen is going to use 14nm+ . Given the late Q4 2017 launch of Raven Ridge notebook I am speculating it is the first one to use 14nm+ followed by desktop Raven Ridge, Pinnacle Ridge and finally Threadripper 2018. You can disagree with that. I won't argue further on this.

Sent from my SM-G935V using Tapatalk
 
  • Like
Reactions: ajc9988

ajc9988

Senior member
Apr 1, 2015
224
142
116
you are speculating and connecting the dots between two processors listed in 2018, as if both should be 14nm+, and I'm the one speaking in absolute terms? I'm saying there's no proof beyond wild speculation.



...I'm the one accusing you of being an idiot?



"It was originally listed as a 2018 product and is listed with 14nm+ products"

I rest my case with you sir, this is my last reply. If you continue in this fashion I will report you for trolling.
Pretty sure this is what you said previously, derisively, which insinuates what you accuse me of: "If you think they'll use an improved version of their process for the socketed low volume desktop version of Raven Ridge, and thus having two different versions of Raven Ridge right out of the gate, you have completely lost your mind."

Both Raghu78 and I have explained our rationale in why we believe that 14nm+ will be used with Raven Ridge on mobile. You have acted abusively to both of us. So please, report me if you must so they can see your behavior, especially as both of us have conceded that it is an unknown and could be either on 14nm or 14nm+. Go ahead.

Edit: also, the map I did was originally published 4 months ago. https://videocardz.com/67362/amd-pinnacle-ridge-to-feature-up-to-8-zen2-cores

But, the releases on these products have been in flux, such as the pro series, which one map shows for Q1 2018, but later information suggests may be earlier. As such, and the likelihood of energy improvements on a 14nm+ process combined with the Q4 mobile release, which was at one point rumored at Q3, I think there is plenty reason to speculate on the push and jump for it to be on the 14nm+ process.

I also speculate that both Epyc and TR will not see a 14nm+ refresh, instead leaving that only for the mainstream and mobile segments, which places it at a 12-18 month refresh cycle on those products going directly to 7nm. This makes sense, as you would sell fewer on a refresh and Zen 3 is rumored to roll out late 2019 into 2020, meaning you would leave too little time between the processes for adoption. It also matches with Intel moving the SP and X lines up to the front starting with Ice Lake-SP/X, which is taping out and may be seen as early as H2 2018. So by going to 7nm with TR and Epyc first, you have a head on battle with Intel's release, within a quarter or so, you refresh to 14nm+ on those lines in the first half of 2018, maybe as late as Q3 2018 on Vega 20, then you move to 7nm Zen 2 for mainstream and mobile throughout 2019, then repeat, starting at the high end, for H2 2019 and throughout 2020.
 
Last edited:

ao_ika_red

Golden Member
Aug 11, 2016
1,397
541
106
So by going to 7nm with TR and Epyc first, you have a head on battle with Intel's release, within a quarter or so, you refresh to 14nm+ on those lines in the first half of 2018, maybe as late as Q3 2018 on Vega 20, then you move to 7nm Zen 2 for mainstream and mobile throughout 2019, then repeat, starting at the high end, for H2 2019 and throughout 2020.
I'm not really sure about EPYC and TR will get first treatment of 7nm. Because, as usual, a new process means low yield and AMD can't afford such loss. I bet they will use the smaller chip, either small GPU like P11 successor or mainstream version of Zen2.
After all, AMD hold their schedule very close to their chest, even I won't dare to bet when first mobile Raven Ridge will be delivered, let alone which process (LPP or 14+) they will be used.
Keep all your heads down, guys. It's just about prediction and speculation after all.
 

ASK THE COMMUNITY