• 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."

New Zen microarchitecture details

Page 81 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

KTE

Senior member
May 26, 2016
478
130
76
In terms of FDSOI...
https://www.semiwiki.com/forum/content/5837-highlights-28nm-fd-soi-san-jose-presentations.html
http://semimd.com/hars/tag/14nm/
https://www.semiwiki.com/forum/content/3767-fd-soi-14nm.html
45nm ARM https://www.arm.com/about/newsroom/26070.php
http://www.advancedsubstratenews.com/2014/07/the-soi-papers-at-vlsi-14-part-1-breakthroughs-in-14nm-fd-soi-10nm-soi-finfets/

See what they presented. Mostly for smaller mobility devices where leakage is crucial but IBM and Co did present some impressive "profiling" (like usual).

In terms of GF 14nm...
https://www.semiwiki.com/forum/content/5119-globalfoundries-14nm-process-update.html
Global Foundries 14nm process is a FinFET on bulk process they licensed from Samsung and both companies supply the same process although as Shubhankar pointed out they have different targets for the process especially in light of Global acquiring IBM's chip business.
And yes, IBM has the more high performance tuned process at GF with 14nm expected to be a long node with many tiers.

EDIT:
LPP is a tweaked version of LPE... as GF has marketed it:




Sent from HTC 10
(Opinions are own)
 
Last edited:

KTE

Senior member
May 26, 2016
478
130
76
16 June: http://blogs.barrons.com/techtraderdaily/2016/06/16/advanced-micro-polaris-zen-impress-says-canaccord-buy-the/

Fortt asked if the “Zen” chip means the company is still “working out the kinks.”


“We are extremely exited [sic] about Zen,” said Su. “We’re going to tell you more about it as we go over the next couple of quarters.” “It’s on track, it’s special.”


Fortt asked Su if the chip will have “32 cores,” meaning, 32 individual CPU elements on each chip. “Nice try, Jon!” she said with a laugh.
:D
 

Abwx

Diamond Member
Apr 2, 2011
9,133
918
126
Sounds like it won't be released in the next 2 quarters, so 2017?
“We are extremely exited about Zen,” said Su. “We’re going to tell you more about it as we go over the next couple of quarters.” “It’s on track, it’s special.

We ll know more about Zen in the two coming quarters, she s talking of releasing infos, next significant infos is at Hotchips for the exact uarch in two months and absolute perfs will likely be known at launch in two quarters, that is in october..
 

LTC8K6

Lifer
Mar 10, 2004
28,523
1,570
126
“We are extremely exited about Zen,” said Su. “We’re going to tell you more about it as we go over the next couple of quarters.” “It’s on track, it’s special.

We ll know more about Zen in the two coming quarters, she s talking of releasing infos, next significant infos is at Hotchips for the exact uarch in two months and absolute perfs will likely be known at launch in two quarters, that is in october..
Tell us more about it = talk about it.

That's the way I read it.

If they are launching in October, then I can't see any downside to revealing Zen pretty soon.

The longer they wait, the more customers buy something else.

With a nice Zen reveal, many customers will wait for Zen, and we can dispose of the hype and talk about Zen realistically.
 

KTE

Senior member
May 26, 2016
478
130
76
Tell us more about it = talk about it.

That's the way I read it.

If they are launching in October, then I can't see any downside to revealing Zen pretty soon.

The longer they wait, the more customers buy something else.

With a nice Zen reveal, many customers will wait for Zen, and we can dispose of the hype and talk about Zen realistically.
We'll have a lot by late August...

Personally, I'm pretty adamant this is a paper November-ish DT launch for 2016. I don't think that's too bad, if the product is competitive.
 

Headfoot

Diamond Member
Feb 28, 2008
4,444
636
126
Makes the most business sense to drop Zen details (e.g. in order to slow competitor sales) around the peak buying period around Back to School season, so I'd assume we get some details around that time. Next couple of months, even without her making that statement it makes the most logical sense.
 

LTC8K6

Lifer
Mar 10, 2004
28,523
1,570
126
Makes the most business sense to drop Zen details (e.g. in order to slow competitor sales) around the peak buying period around Back to School season, so I'd assume we get some details around that time. Next couple of months, even without her making that statement it makes the most logical sense.
I thought we were getting server chips first, so there wouldn't be any Zen chips for back to school systems?
 

Ajay

Diamond Member
Jan 8, 2001
7,886
2,959
136
Can't wait till Hotchips now. It'll be nice to see a clean sheet x86 uArch!
I really, really hope that AMD isn't over hyping Zen. It would be awesome to have an Intel competitor again.
 

KTE

Senior member
May 26, 2016
478
130
76
Can't wait till Hotchips now. It'll be nice to see a clean sheet x86 uArch!
I really, really hope that AMD isn't over hyping Zen. It would be awesome to have an Intel competitor again.
Same here.

It will either be a revolution or a huge letdown

But the stocks are only going up due to renewed analyst ratings predicting $6 due to Zen and Polaris.




Sent from HTC 10
(Opinions are own)
 
Last edited:

scannall

Golden Member
Jan 1, 2012
1,751
1,177
136
Can't wait till Hotchips now. It'll be nice to see a clean sheet x86 uArch!
I really, really hope that AMD isn't over hyping Zen. It would be awesome to have an Intel competitor again.
I don't think it very likely that it will match Skylake IPC. Perhaps as high as Haswell. More likely Ivy Bridge, but that's still pretty darn good. But, if the price and core counts are right, then they'll be competitive enough for me anyway.
 

MajinCry

Platinum Member
Jul 28, 2015
2,488
557
136
Said it 'afore, but Skylake's only around 20% faster than Sandybridge. The most pessimistic expectations are for Sandybridge/Ivybridge perf. That's actually pretty damn good, unlike the current Piledriver situation.
 

KTE

Senior member
May 26, 2016
478
130
76
What is your "revolution" threshold?
If marketshare can be above the Duron days.

If it's competitive and makes them a profit.

Financial analysts have predicted 6% server share at most by 2020, from the current <1%. Bleak compared to the 20% during K8 days, but 5% in each market would be a win for them in my book.

A Lenovo sort turnaround from the IBM era (~5%) IS possible

Sent from HTC 10
(Opinions are own)
 

jpiniero

Diamond Member
Oct 1, 2010
8,709
1,564
126
Said it 'afore, but Skylake's only around 20% faster than Sandybridge. The most pessimistic expectations are for Sandybridge/Ivybridge perf. That's actually pretty damn good, unlike the current Piledriver situation.
It's probably more like 20-30% for the most part, but there are a few cases where it really takes off... the 6700 is like 45-50% faster than the 3930K in the Chrome benchmarks.

Financial analysts have predicted 6% server share at most by 2020, from the current <1%. Bleak compared to the 20% during K8 days, but 5% in each market would be a win for them in my book.
That seems tough unless it's competitive on perf/w. China may not care but they may be the only ones, and there's always going to be the question whether they will actually pay the royalties.

Especially with the way that the PC market is imploding I don't think you could really count on increased mainstream share really helping their cause much.
 

MajinCry

Platinum Member
Jul 28, 2015
2,488
557
136
Aye, there are fringe cases, like Haswell being a good 30% faster than Ivybridge in emulation, but again. Fringe cases.

Gotta average it out. Fer example, Piledriver's a dog, but in multithreaded Linux workloads, it's a god damn monster.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
What is your "revolution" threshold?
Didn't ask me, but if it hits FX-8150 clocks (3.6GHz base / 4.2GHz boost) <= 125W power draw (overclocked or not) I'd say Zeppelin is nothing shy of a revolution.

However I'd still call it as a success if AMD is able to ship the 8C/16T variant of Zeppelin at >= 3.0GHz base and >= 3.6GHz boost, at 95W.

IMO the only way Zeppelin can fail is a failure to reach clocks high enough.
 

Dresdenboy

Golden Member
Jul 28, 2003
1,730
554
136
citavia.blog.de
Didn't ask me, but if it hits FX-8150 clocks (3.6GHz base / 4.2GHz boost) <= 125W power draw (overclocked or not) I'd say Zeppelin is nothing shy of a revolution.

However I'd still call it as a success if AMD is able to ship the 8C/16T variant of Zeppelin at >= 3.0GHz base and >= 3.6GHz boost, at 95W.

IMO the only way Zeppelin can fail is a failure to reach clocks high enough.
They might delay it then. Using clocks and power, you might actually draw a line into a chart to define a range of succesful to revolutionary ZP chip specs.

Did you follow the Polaris clock reveals? I'm sure they did't move the FO4 metric around that much while doing the next GCN iteration. Power went down. Yet the clocks are significantly higher with 14LPP.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
They might delay it then. Using clocks and power, you might actually draw a line into a chart to define a range of succesful to revolutionary ZP chip specs.

Did you follow the Polaris clock reveals? I'm sure they did't move the FO4 metric around that much while doing the next GCN iteration. Power went down. Yet the clocks are significantly higher with 14LPP.
It will definitely be interesting to see both how Polaris overclocks, especially the smaller variant. Bonaire is basically the benchmark for Polaris, since it is the best clocking 28nm GCN GPU AMD has. AIBs have shipped these cards at 1200MHz out of the box and they hit 1400MHz quite easily on stock cooling.

Since certain people claim that shrinking virtually any design from 28nm bulk to 14nm FinFet LPP will immediately increase the Fmax by 50 - 100%, I'm really keen to see how that will work out :sneaky:
 

KTE

Senior member
May 26, 2016
478
130
76
Since certain people claim that shrinking virtually any design from 28nm bulk to 14nm FinFet LPP will immediately increase the Fmax by 50 - 100%, I'm really keen to see how that will work out :sneaky:
Be careful of process PR tho. IIRC it's 30% frequency gain or 50% less power from a process viewpoint... At a standard voltage and temp (like 1V). It most definitely won't be FMax. It's in one of the links I posted.

But yield is a different spanner altogether. 14nm is no where near 28nm (which is due to remain in high volume production till 2020 at minimum) and the production cost for 14nm is more than 2.5x.

And the way PC markets function with the TDP caps, additional features to the chip are typically favoured over frequency. Since those finFET power improvements tend to be mostly static rather than switching. 50% less sounds a lot but.. For example, out of 15W on a 100W chip, it's not so much.

I think I would agree looking at the design with many cores and SMT, to hit BD frequencies at <125W 14nm would be a remarkable achievement. Even more so with the limited Vt scaling these days.

I also feel that one of the next major features to come would be incorporating GPU SPs in place of the FPU with the breakthroughs in HBM/eDRAM. I think that's why the HSA/APU movement originally began.

Sent from HTC 10
(Opinions are own)
 

KTE

Senior member
May 26, 2016
478
130
76
BTW... Looking at DT Excavator, I wouldn't want AMD to project anything based off it!

It's certainly not even 5% better than PD on average. It wins few but loses more with a landslide!

Absolute performance wise, it's ppp. It needs FAR higher clocks but it doesn't scale at all. It's a mobile chip, simply put, low frequency+graphics optimized.

It's no better than the old Regor chips were at the time. Kuma caned them, literally.

Sent from HTC 10
(Opinions are own)
 

ASK THE COMMUNITY