[wccftech] Rumor: Lisa Su Considers A Role Beyond AMD And Prepares A Successor

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

moinmoin

Diamond Member
Jun 1, 2017
5,064
8,032
136
Nah, N6 is an upgrade path for those who can't afford N5 tapeouts.
It still uses EUV, on more layers than N7+, too.
While that's the case N6 is using the same design rules as 7N whereas N7+ doesn't (and N5's design rules will build upon that instead N7). Also compared with N7, N7+ has higher area reduction as well as power/performance improvements than N6.
 

ondma

Platinum Member
Mar 18, 2018
2,999
1,520
136
In 2 years, AMD will have Zen 4 pushing out. What this means? Next year we are looking at Zen 3. With much larger IPC gain than Zen 2 is for Zen+.

Everybody here is suddenly believing that AMD is going to sleep for another 5 years, and that will allow Intel to catch up. Wake up guys. Its Intel who has to catch up to AMD. Regardless of what you think, and what you believe.

Intel's 10 nm process is really Intel's Bulldozer.
Zen 3 is going to have a higher IPC gain than Zen 2??? Haven't heard that one before. Zen 2 already has about 15% gain and you think in one more year they are going to make another gain larger than that? OK, whatever you say. Not to mention the max frequency still sucks.
 

cytg111

Lifer
Mar 17, 2008
23,993
13,519
136
Lisa is a role model for all women out there, if she chooses to go elsewhere I am sure she will dominate there too. God bless.
 

Glo.

Diamond Member
Apr 25, 2015
5,802
4,776
136
Zen 3 is going to have a higher IPC gain than Zen 2??? Haven't heard that one before. Zen 2 already has about 15% gain and you think in one more year they are going to make another gain larger than that? OK, whatever you say. Not to mention the max frequency still sucks.
Why is it so hard to think that AMD can gain once again higher IPC uplift with Zen 3, than they got with Zen 2?

Because genuine technological advantage of AMD, over Intel is unthinkable?
 
Mar 11, 2004
23,280
5,721
146
IBM kills everything they touch. Everything.



Welllll he certainly took his own post seriously. Which is kinda comedic in and of itself. Eventually someone has to just deadpan respond and say, "sorry, but no".



Why? Intel is overstating their successes in every sector wherever possible. Roadmaps are only going to get worse, not better. They haven't delivered on a node successfully and on-time since 22nm. Ponder that for a moment. After Haswell Refresh/Devil's Canyon, Kabylake, Coffeelake, Whiskey Lake, Comet Lake, Cascade Lake, and Cooper Lake, it should be obvious that Intel is slinging very little but band-aids onto the market. They have done nothing to inspire confidence other than cruise on momentum.

That's why they'd be wanting a change in focus and culture. If Lisa were being primed to take over, she'd be overhauling their plans. Imagine an IBM focused on technological prowess, where the engineers are given freedom operate. AMD would give IBM a big infusion of hardware that they can then leverage their massive support apparatus to maximize the potential of, while they pour R&D money into developing it further. AMD becomes less abound by their partners (OEMs, who have consistently dragged their feet with AMD, which is why AMD is where they are). AMD's stuff is practically targeted right at IBM's markets, with many able to develop from there (think Threadripper based workstations).

Heck some of the next ideas for Zen (more mulit-threading capability, probably more development of caches to alleviate the modular design) sound like they're right out IBM's work on Power.

Create a sort of skunkworks to develop new computing ideas, and anything promising find the business case to bring it to market (for instance, I've talked about a unified system that would be like a high end gaming console, that would be good for gaming while also being prime for workstations).
 

ondma

Platinum Member
Mar 18, 2018
2,999
1,520
136
Why is it so hard to think that AMD can gain once again higher IPC uplift with Zen 3, than they got with Zen 2?

Because genuine technological advantage of AMD, over Intel is unthinkable?
Obviously the answer is that as a product matures, improvements become increasingly difficult; i.e. the low hanging fruit has already been picked.
BTW, i though this thread was about Lisa Su, not hype for AMD next gen products. Isnt there a thread for that already?
 
Last edited:
  • Like
Reactions: happy medium

Glo.

Diamond Member
Apr 25, 2015
5,802
4,776
136
Obviously the answer is that as a product matures, improvements become increasingly difficult; i.e. the low hanging fruit has already been picked.
BTW, i though this thread was about Lisa Su, not hype for AMD next gen products. Isnt there a thread for that already?
Zen 3 is not a product that has low hanging fruits, because it is not Zen 2 iteration. Zen 3 is brand new architecture, just like Ice Lake is and Zen 2 was. From this moment on, AMD is on yearly architecture upgrades.
 

coercitiv

Diamond Member
Jan 24, 2014
6,626
14,035
136
BTW, i though this thread was about Lisa Su, not hype for AMD next gen products.
Nope, this thread is about wccftech pulling a fast one on the internet, and since they got denied by Lisa Su it rapidly became a cesspool of FUD and unrealistic expectations.
 

A///

Diamond Member
Feb 24, 2017
4,351
3,158
136
en 3 as well as 7nm+ are far bigger changes than Zen+ and 12nm.

Did you mean Zen 2? Zen 2 uses 7nm. Zen+ was 12nm. Or do you mean the change from 7nm to 7nm+ is massive?

The refined 7nm TSMC process is actually 6nm. 7nm+ is a bigger change which will be the first major step toward full EUV on 5nm.

7nm+ being 6nm makes sense because Intel wanted an density increase they didn't have experience in to make the jump from 14 to 10 nm but couldn't and were too far in to give up. Right? How much of a pure transister increase will we see from 7 to 7+ nm?
Zen 3's changes may be smaller than Zen 2's (at least Papermaster indicated so before) but that may well be the case due to many changes planned for Zen 3 to be essentially backported to Zen 2 (the Zen core apparently was originally planned to have much fewer changes).

I'm aware of this. Thought it makes me wonder if they did it because they could or because they had more in stock for Zen3 that came up later and made the decision to back port. Who knows.

Increasing frequencies will be an interesting topic. As said 7nm+ is no process refinement at all, and new process nodes usually make it harder to hold the same frequencies, never mind increase them (hello Intel 10nm).
I threw frequency out as an example. I wasn't expecting such a major jump in performance for Zen2 but it happened. I suppose you could say it isn't impressive if you'd prefer to stick your head in the sand like some long time Intel consumers.
 

Atari2600

Golden Member
Nov 22, 2016
1,409
1,655
136
They pulled whatever they thought they could get done in the timeline into Zen2.

It will be the same for Zen3 with respect to Zen4. What they feel confident they can get done will get pulled in, what they feel is too much to take on gets pushed out.

I guarantee there is no conscious thought as to "pulling punches" and not trying to do everything they can in the timeline!

If they have an internal plan of where they want to go in 5 years and are constantly updating it and taking steps along it - then there is little reason to believe Zen3 will not be an appreciable step on Zen2. Will it be a step equally as large? Unknown to anyone external to the design team right now - but it might.
 

moinmoin

Diamond Member
Jun 1, 2017
5,064
8,032
136
Did you mean Zen 2? Zen 2 uses 7nm. Zen+ was 12nm. Or do you mean the change from 7nm to 7nm+ is massive?

The changes from 7nm to 7nm+ at TSMC are very significant, unlike the change from 14nm to 12nm at GloFo which allowed AMD to reuse its designs and sneak in Zen+ (a refresh that initially wasn't on the roadmap).

The refresh/refinement to TSMC 7nm is TSMC 6nm. (Yotsugi appears to disagree with that, but contradicts the news I linked above and doesn't link sources of his own.)

TSMC 7nm+ is the start of EUV that will lead to the even more significant refresh/refinement of full EUV TSMC 5nm. 6nm is a dead end.

7nm+ being 6nm makes sense because Intel wanted an density increase they didn't have experience in to make the jump from 14 to 10 nm but couldn't and were too far in to give up. Right? How much of a pure transister increase will we see from 7 to 7+ nm?

TSMC gives some details in their quarterly earning calls, like https://www.anandtech.com/show/14290/tsmc-most-7nm-clients-will-transit-to-6nm
In that article the density increase 7nm -> 7nm+ is ~17%, 7nm -> 6nm is ~15%. Note that density is always a performance trade off, i.e. high density = low frequency, so density is reduced again to achieve higher clocks. AMD uses performance optimized processes at TSMC where the stats mentioned above likely divert.

I'm aware of this. Thought it makes me wonder if they did it because they could or because they had more in stock for Zen3 that came up later and made the decision to back port. Who knows.

They likely have more in stock and had sufficient time to backport changes that promised the biggest improvements. AMD always claims to be on time, but I recall they wanted a yearly cadence in early 2017 without Zen+ on the roadmap, so technically Zen 2 is a year late. This was time well spent looking at the issues Intel still has as well as how reception to the Epyc ecosystem changed.
 
  • Like
Reactions: A///

Yotsugi

Golden Member
Oct 16, 2017
1,029
487
106

happy medium

Lifer
Jun 8, 2003
14,387
480
126
Zen 3 is not a product that has low hanging fruits, because it is not Zen 2 iteration. Zen 3 is brand new architecture, just like Ice Lake is and Zen 2 was. From this moment on, AMD is on yearly architecture upgrades.
Who told you this? Do you work for AMD or are you just guessing or just making stuff up?
What is new about the Zen 3 architecture, specifically.
 

inf64

Diamond Member
Mar 11, 2011
3,864
4,546
136
I expect Zen3 to have similar IPC gain as Zen2 over Zen+. So around ~15%, just about enough to be on top of Icelake cores (minus AVX512 stuff of course). Core will likely get wider, we have SMT4 rumors floating for a while now. As Forrest Norrod already stated, clocks are going to be stagnant or even regress with newer nodes so AMD must pursue IPC as much as they can. Since Zen is a grounds up design they have a lot of room to improve/evolve core, probably to the point it's not Zen anymore but something else (Zen4). We have AMD architects on linkedin having Zen5 in their resumes so it's safe to say AMD has Zen5 in design as well by now.
 
  • Like
Reactions: lightmanek

Glo.

Diamond Member
Apr 25, 2015
5,802
4,776
136

"We designed this part to compete with Ice Lake, expecting to make some headway on single threaded performance. We did not expect to be facing re-warmed Skylake instead. This is going to be one of the highlights of our careers" -F. Norrod

Oh, even AMD is starting to make fun of Intel, and their Bulldozer-type of failure.
 
  • Like
Reactions: guachi and Drazick