New Zen microarchitecture details

Page 52 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Dresdenboy

Golden Member
Jul 28, 2003
1,730
554
136
citavia.blog.de
I find it interesting that they don't mention Excavator any more in the IPC prediction. Piledriver and Steamroller parts are still shipping (making them "current core") and it is easier to achieve the "over 40%" improvement over them than it is over Excavator.

As I've said before when directly compared to Excavator I heard and seen AMD using term "up to" as a prefix. And that started to happen quite recently.
Point is heard. But XV is also a "current core". And the last official slide making a clear distinction was this one:
AMDZenRoadmap_678x452.jpg


Maybe I asked this already, but does "up to" still refer to IPC or something else instead (for example per core performance)?
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
Well. Even if cb is a good measure of performance doing it at eg. 95w instead of 140w should make for a difference.

At same TDP Zen would be 21% faster than the 5820K, since it has 33% cores advantage this amount to 9% disadvantage/core for Zen.


TDP isn't necessarily the power consumption, especially on Intels Highend platform. The slowest sixcore and fastest Octacore have a 140W TDP but unsurprisingly the sixcore draws much less. And much less than a 125W TDP CPU from AMD
http://www.bit-tech.net/hardware/2014/09/03/intel-core-i7-5930k-and-core-i7-5820k-revie/8
http://www.hardware.fr/articles/940-8/protocole-cpu-consommation-efficacite-energetique.html


At the end of the year BDW-E is available anyway, HSW-E is old.
 

krumme

Diamond Member
Oct 9, 2009
5,952
1,585
136
I expect Zen cores the have similar IPC as Ivy Bridge, on average. Faster in certain workloads but the average should be quite well matched.

Seeing how much additional effort AMD has taken in regards of Zen power and clock management and how strict the VRM requirement (based on the existing boards) are, I don't feel that Zen will be able to scale too well in any aspect.

Because of that I actually have lowered my expectations for Zen's shipping frequencies from my original estimations (for the "halo" 8C/16T desktop FX), which originally were 3000MHz (±200MHz) base and 3600MHz (±200MHz) maximum boost. At the moment I'd expect 2600MHz (±200MHz) base and 3200MHz (±200MHz) maximum boost. However I have no idea what the clocks will actually be, so as always anyone elses guess is just as good as mine.
Even if amd is starved for ressources Keller would have brought tons of vital knowledge from apple. Add time since ib and hw. There is plenty to learn from. Ipc is stagnating so time is working for amd. From this perspective i would think like between ib and hw - if not more like the latter. Keller have a solid proven track record here. I think its safe they get to their ipc targets.
But gf also have a track record and judging from it there should both be freq issues and yield issues in spades. One have to wonder if that is not more important than ipc +/-10%. With low yield its utterly uninteresting.
 

Dresdenboy

Golden Member
Jul 28, 2003
1,730
554
136
citavia.blog.de
Nope, they very clearly took the effort to label the Carrizo slide but intentionally signaled a separation between the two and did not label the second.
Best case: OR = 1, SR = 2 (supported by horizontal grid with same step size as on the left)
Worst case: OR = SR

Lets check plausibility and remove any SMT, CMT, TDP, IPC factors and look at the resulting delta.

Example:
OR: 1.0 / 0.8 = 1.25 (20% CMT penalty)
SR: 2.0 / 1.2 / 1.4 =1.19 (20% SMT, 40% IPC)
 
Last edited:

zentan

Member
Jan 23, 2015
177
5
36
TDP isn't necessarily the power consumption, especially on Intels Highend platform. The slowest sixcore and fastest Octacore have a 140W TDP but unsurprisingly the sixcore draws much less. And much less than a 125W TDP CPU from AMD
http://www.bit-tech.net/hardware/2014/09/03/intel-core-i7-5930k-and-core-i7-5820k-revie/8
http://www.hardware.fr/articles/940-8/protocole-cpu-consommation-efficacite-energetique.html


At the end of the year BDW-E is available anyway, HSW-E is old.
Yeah and Broadwell-E will launch ahead anyway.
Just want to know the status on platform for Zen,would it have quad-channel and higher than 16 PCIe3.0 lanes from CPU?
 
Last edited:

Abwx

Lifer
Apr 2, 2011
10,847
3,297
136
TDP isn't necessarily the power consumption, especially on Intels Highend platform. The slowest sixcore and fastest Octacore have a 140W TDP but unsurprisingly the sixcore draws much less. And much less than a 125W TDP CPU from AMD
http://www.bit-tech.net/hardware/2014/09/03/intel-core-i7-5930k-and-core-i7-5820k-revie/8
http://www.hardware.fr/articles/940-8/protocole-cpu-consommation-efficacite-energetique.html


At the end of the year BDW-E is available anyway, HSW-E is old.

And the FX CPUs you re using as "exemple" are even more older node wise, so much for your "exemple" of CPUs that are 2 nodes behind what will be used by Zen....

FTR 14nm LPP require 0.8V at 2.41GHz when Intel s 14nm is at 0.88V at least, this imply that AMD could have as much as 20% advantage in perf/Watt for the process, you ll have to be used about it, Intel s gigantic process advantage will be over, we ll see how they ll do when they ll have no more a hudge advance in the starting line of the race...
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
Point is heard. But XV is also a "current core". And the last official slide making a clear distinction was this one:
AMDZenRoadmap_678x452.jpg


Maybe I asked this already, but does "up to" still refer to IPC or something else instead (for example per core performance)?

Excavator is indeed "a current core", however on the same page they compare to Orochi which is a desktop / server core. And Orochi Rev. C (i.e Vishera == Piledriver) is also a current core. And at the moment Excavator is not "a current core" for desktop, since it has not been released outside a single recycling SKU (Athlon X4 845) on desktop and never in server segment.

And the most recent phrase was precisely: "up to 40% more IPC over previous gen. core". Since Excavator is not mentioned by the name, IMO it can be mean the Bulldozer family as whole, since Zen has no previous generation (yet) :D
 

Abwx

Lifer
Apr 2, 2011
10,847
3,297
136
And the most recent phrase was precisely: "up to 40% more IPC over previous gen. core". Since Excavator is not mentioned by the name, IMO it can be mean the Bulldozer family as whole, since Zen has no previous generation (yet) :D

Could you provide a link for this alleged "up to"..?..

During the Q4 Earning call Lisa Su said:

In client computing our opportunities to regain share in 2016 will be driven by our design win momentum, continued progress expanding into the commercial market and reentering the high performance desktop market late in the year with our Zen based Summit Ridge CPU. Our second growth pillar is in the $15 billion plus datacenter and infrastructure markets, driven by our FirePro GPU’s and next-generation service CPU’s. Our Zen based CPU development is on track to achieve greater than 40% IPC uplift from our previous generation and we’re on schedule to sample later this year.

AMD-Zen-2016-Key-Growth-Pillars-2.jpg


Greater than 40% over previous gen on the latest slides, one more time, can you provide a source for your statement..?.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
Looked at the Cinebench R15 chart AMD provided in their PDF.

From the notes you can see that they used FX-7500 (Kaveri), FX-8800P (Carrizo) and FX-9800P ("Bristol Ridge") in the comparison. FX-7500 scores 179 in R15 MT (pclab.pl) while FX-8800P scores 234 (my result). For that part the chart is absolutely correct (Kaveri = 100%, Carrizo = 130%, "Bristol Ridge" = 150%). However between Carrizo and "Bristol Ridge" the things get interesting. According to the chart "Bristol Ridge" should score 15.4% higher than Carrizo, i.e 270 points in Cinebench R15 MT test. The chip would need to run ~3200MHz in order to score 270 points...

FX-9800P has the same TDP as FX-8800P 15W / 25W (base boost), however it doesn't support cTDP > TDP configuration. Since Cinebench R15 takes less 200 seconds to complete in MT mode, both of the chips will have the 25W STAPM limit available throughout the test. With 25W power budget Carrizo is able to sustain <2700MHz frequency in Cinebench R15. FX-9800P has higher base and boost frequencies than FX-8800P, however the maximum frequencies doesn't matter when the effective frequency is limited by the power budget. So where does Bristol Ridge find the additional power worth of 500MHz, since the silicon itself is identical? FX-9800P also runs at higher voltages than FX-8800P does.

"50% performance-per-watt generational uplift" would also suggest that AMD claims that Bristol Ridge (7th generation) has 50% higher performance per watt than Carrizo (6th generation), which sounds pretty bold to me. Should be pretty hard to achieve, unless Carrizo has some serious errata which is yet to reveal itself...
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
Could you provide a link for this alleged "up to"..?..



AMD-Zen-2016-Key-Growth-Pillars-2.jpg


Greater than 40% over previous gen on the latest slides, one more time, can you provide a source for your statement..?.

As I said before, I know very well that we have heard both "40%" and "over 40% figures quoted by AMD themselves.

http://forums.anandtech.com/showpost.php?p=38228995&postcount=1180

I don't record the conversations I have so no, I unfortunately have no link to provide for the quote. Allegedly is perfectly fine by me :sneaky:
 

guskline

Diamond Member
Apr 17, 2006
5,338
476
126
The Dresdenboy and The Stilt, thank you both for keeping this an interesting and technical discussion ( much is over my head but I try to learn).

I sure hope the upcoming AMD presentations throw some new scraps of info about Zen to keep interest up. Dec 2016 to Feb 2017 sure seems a long way for those waiting for a new AMD cpu.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106

JK is a legend and he could very well say "I'm an artist, give me the time and the resources and I'll create something worth releasing".

I'm sure that he could and infact did create something worth releasing, however I'm not so sure that AMD necessarily gave him the time and the resources he asked for, nor if the resulting something is able to stand the more recent Intel designs toe to toe. It is impossible to give something you don't have in the first place (AMD / Intel R&D 952M vs. 11.7B$, 2014). One just doesn't pick a fight with a guy over twelve times of your own size, no matter if your henchman is JK or not.

Even if Zen would turn out to be even a bigger disaster than Bulldozer (nope) was, I wouldn't think any less of JK. The blaim would go to AMD and AMD only.
 

deasd

Senior member
Dec 31, 2013
513
724
136

guskline

Diamond Member
Apr 17, 2006
5,338
476
126
JK is a legend and he could very well say "I'm an artist, give me the time and the resources and I'll create something worth releasing".

I'm sure that he could and infact did create something worth releasing, however I'm not so sure that AMD necessarily gave him the time and the resources he asked for, nor if the resulting something is able to stand the more recent Intel designs toe to toe. It is impossible to give something you don't have in the first place (AMD / Intel R&D 952M vs. 11.7B$, 2014). One just doesn't pick a fight with a guy over twelve times of your own size, no matter if your henchman is JK or not.

Even if Zen would turn out to be even a bigger disaster than Bulldozer (nope) was, I wouldn't think any less of JK. The blaim would go to AMD and AMD only.
That pretty well sizes it up.:\
 

el etro

Golden Member
Jul 21, 2013
1,581
14
81
Point is heard. But XV is also a "current core". And the last official slide making a clear distinction was this one:
AMDZenRoadmap_678x452.jpg


Maybe I asked this already, but does "up to" still refer to IPC or something else instead (for example per core performance)?


Is what i want to talk too. Marketing was clear in to point that IPC measurements was about Core versus Core, not about processor versus processor. Maybe at the maximum the core with L2 included.
 

Abwx

Lifer
Apr 2, 2011
10,847
3,297
136
As I said before, I know very well that we have heard both "40%" and "over 40% figures quoted by AMD themselves.

http://forums.anandtech.com/showpost.php?p=38228995&postcount=1180

I don't record the conversations I have so no, I unfortunately have no link to provide for the quote. Allegedly is perfectly fine by me :sneaky:


That s also perfectly fine for me since Lisa su stated otherwise, after all she s CEO and would be at risk if ever she inflated the numbers, so she certainly is much more credible than someone who is talking of hearsay without even providing a source, a quote or anything that would support his claims...

Indeed if you are so sure you should tell your "source", if it exist, that AMD is to be heavily shorted since they will be in big trouble juridically speaking, with tons of lost trials in sight...
 

Dresdenboy

Golden Member
Jul 28, 2003
1,730
554
136
citavia.blog.de
I'm sure that he could and infact did create something worth releasing, however I'm not so sure that AMD necessarily gave him the time and the resources he asked for, nor if the resulting something is able to stand the more recent Intel designs toe to toe. It is impossible to give something you don't have in the first place (AMD / Intel R&D 952M vs. 11.7B$, 2014). One just doesn't pick a fight with a guy over twelve times of your own size, no matter if your henchman is JK or not.

Even if Zen would turn out to be even a bigger disaster than Bulldozer (nope) was, I wouldn't think any less of JK. The blaim would go to AMD and AMD only.
A small and quick contender might also place some punches.

But it would also be interesting, what amount of R&D costs (less fab dev) has been used for K7 and P4.
 
Mar 10, 2006
11,715
2,012
126
As I said before, I know very well that we have heard both "40%" and "over 40% figures quoted by AMD themselves.

http://forums.anandtech.com/showpost.php?p=38228995&postcount=1180

I don't record the conversations I have so no, I unfortunately have no link to provide for the quote. Allegedly is perfectly fine by me :sneaky:

The Stilt, it is always very telling that whenever the "over 40%" comes up, AMD always changes to wording to something like: "previous generation," "current core," etc.

They never come out and say "Excavator" because that's likely not what they intend. Piledriver is the most current core for HEDT, and the new Zen chip is HEDT first. Seems to me that they're saying >40% perf/clock over Piledriver.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
That s also perfectly fine for me since Lisa su stated otherwise, after all she s CEO and would be at risk if ever she inflated the numbers, so she certainly is much more credible than someone who is talking of hearsay without even providing a source, a quote or anything that would support his claims...

Indeed if you are so sure you should tell your "source", if it exist, that AMD is to be heavily shorted since they will be in big trouble juridically speaking, with tons of lost trials in sight...

Any idea why AMD doesn't specify Excavator by the name in these slides, when they have done so in the past (by Lisa Su for example)? Instead they use terms which leave plenty upon interpretation ("a current generation core").
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
As pointed out by Arachnotronic, as soon as the quoted figure is "greater than 40%" they start pointing to "previous generation" :sneaky:

Our Zen based CPU development is on track to achieve greater than 40% IPC uplift from out previous generation and we're on schedule to sample later this year.

01/19/2016 - Lisa Su (Earnings Confrence Call)



40% more instructions per clock*.

*Based on internal AMD estimates for &#8220;Zen&#8221; x86 CPU core compared to &#8220;Excavator&#8221; x86 CPU core.

05/06/2015 - Mark Papermaster (2015 Financial Analyst Day - Foundation For The Future PPT/PDF)
 

Abwx

Lifer
Apr 2, 2011
10,847
3,297
136
Any idea why AMD doesn't specify Excavator by the name in these slides, when they have done so in the past (by Lisa Su for example)? Instead they use terms which leave plenty upon interpretation ("a current generation core").

You ll agree that thoses slides are rather obsolete since we have now a comparison with BD on a CPU level, that is, 8 cores + SMT against a CMted 8 cores.

Also that s all good to use CB as metric, but if we were to use Hardware.fr software suite then the last slide suggest that in respect of a 5960X :

Zen is 10% better in 7 Zip.
Better in X264, X265.
Close in StockFish 5 and equal in Houdini 4 Pro.
Better in Lightroom and Dxo Optics (this latter use 6 cores).
Equal in Visual Studio and MinGW GCC.

Only case it would be somewhat trailing is 3DS Max but according to Hardware.fr HW benefited much from new instructions in this soft, so that s a corner case more than anything else, thanks to this special case the two CPU would be about on par in the average of the softs i mentioned.

http://www.hardware.fr/articles/940-19/indices-performance-cpu.html

Notice that in 3DS Max the FX8350 is faster than Ivybridge in the two test, as it is faster in Povray, nothing like Cinebench, isnt it...
 
Last edited:

jhu

Lifer
Oct 10, 1999
11,918
9
81
Notice that in 3DS Max the FX8350 is faster than Ivybridge in the two test, as it is faster in Povray, nothing like Cinebench, isnt it...

Do you use any of these programs at all? Because I do.

First in comparison to 3D Stuido Max, Cinema 4D, and Blender, barely anyone uses Povray. Their development team is essentially 5 people so any improvements in the program happen at a glacial pace (e.g. official SMP support happened in late 2013!).

Second their benchmark scene has not kept up with the times. It's essentially the same as it was 15+ years ago. It doesn't even use the most commom feature that users use: global illumination. I've tried petitioning for an updated scene to ni avail. From my tests, FX is anywhere from 10%-30% slower per clock, even on custom compiled versions.
 
Last edited:

Abwx

Lifer
Apr 2, 2011
10,847
3,297
136
Do you use any of these programs at all? Because I do.

First in comparison to 3D Stuido Max, Cinema 4D, and Blender, barely anyone uses Povray. Their development team is essentially 5 people so any improvements in the program happen at a glacial pace (e.g. official SMP support happened in late 2013!).

Second their benchmark scene has not kept up with the times. It's essentially the same as it was 15+ years ago. It doesn't even use the most commom feature that users use: global illumination. From my tests, FX is anywhere from 10%-30% slower per clock, even on custom compiled versions.

In the link i posted it s 3DS max, so what is the problem..?..

http://www.hardware.fr/articles/940-9/cpu-rendu-3d-mental-ray-v-ray.html


It should interest you even more since the first test use Mental Ray renderer within 3DS MAX and that it s also used within :

The second test use V-Ray wich is also used in :

Is it good enough for your taste..?.

Or should we keep using Maxon s "home made" test for anything FP..?.
 

jhu

Lifer
Oct 10, 1999
11,918
9
81
Or should we keep using Maxon s "home made" test for anything FP..?.

Except you also keep bringing up Povray. In my opinion, most people aren't using benchmark results appropriately. The real questions that should be answered are: 1) is x program's benchmark indicative of x program's performance? 2) is x program's benchmark indicative of the type of work similar to x?

For Cinebench, the answers are 1) yes, 2) maybe. For Povray the answers are 1) no, 2) no.
 
Last edited:

el etro

Golden Member
Jul 21, 2013
1,581
14
81
You ll agree that thoses slides are rather obsolete since we have now a comparison with BD on a CPU level, that is, 8 cores + SMT against a CMted 8 cores.

Also that s all good to use CB as metric, but if we were to use Hardware.fr software suite then the last slide suggest that in respect of a 5960X :

Zen is 10% better in 7 Zip.
Better in X264, X265.
Close in StockFish 5 and equal in Houdini 4 Pro.
Better in Lightroom and Dxo Optics (this latter use 6 cores).
Equal in Visual Studio and MinGW GCC.

Only case it would be somewhat trailing is 3DS Max but according to Hardware.fr HW benefited much from new instructions in this soft, so that s a corner case more than anything else, thanks to this special case the two CPU would be about on par in the average of the softs i mentioned.

http://www.hardware.fr/articles/940-19/indices-performance-cpu.html

Notice that in 3DS Max the FX8350 is faster than Ivybridge in the two test, as it is faster in Povray, nothing like Cinebench, isnt it...


Did some comparisons at SA, there i get Zen getting 5960x performance on Integer MT, and 4960x performance on Float MT.