New Zen microarchitecture details

Page 207 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
E

Enigma-

Junior Member
Feb 1, 2017
10
19
51
All "X" are the XFR one's! Zen is a masterpiece of engineering and now you can tell from the ISSCC... I would say that IPC is from 40% and up to much higher numbers depending on workload. IPC AMD and IPC Intel will vary as they differ much in their architectures.
 
K

krumme

Diamond Member
Oct 9, 2009
5,956
1,595
136
KTE said:
You once 'calculated' 4GHz base 95W LAUNCH too... With 4.5GHz Turbo. Based on your FO4+Neon claims.

Say so a billion wrongs things and you're bound to get something right.

Sent from HTC 10
(Opinions are own)
Click to expand...
At least he is optimistic and get many things right. I dont always agree with him - far from- but very much value his approach and try to learn from it.
But can you stop whining? You have been whining and making personal attack on bjt for 100 post. Ask yourself when you will stop crying.
 
K

krumme

Diamond Member
Oct 9, 2009
5,956
1,595
136
If prices are true it seems to me they got it right this time unlike eg phenom and bd that was overpriced.

This will enable stable prices during the year and an updated zen can continue this segmentation. Just like the core line. Stable and predictable. There might be shortages because this looks like an optimal server solution so that market will eat a lot, but this will again establish amd as a trustworthy brand with prices that dont colaps.

I am sure they did lisnt to my pricing advice a month ago. Good boys ;)
 
Z

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
I'
krumme said:
At least he is optimistic and get many things right. I dont always agree with him - far from- but very much value his approach and try to learn from it.
But can you stop whining? You have been whining and making personal attack on bjt for 100 post. Ask yourself when you will stop crying.
Click to expand...
The debate between bjt2 and KTE is an interesting one, and instead of jumping on bjt2's defense with excuses of 'optimism,' I'd much rather sit back and watch the reviews flow in and see who was right. bjt2 seems quite knowledgeable on the subject and I for one am interested in seeing whether he's right about 14nm FO4 process and it's potentially high frequency potential in the Zen implementation. (4.5/4.6Ghz, single core turbo).
 
tamz_msc

tamz_msc

Diamond Member
Jan 5, 2017
3,865
3,729
136
Looking at the per-core voltage regulation in Ryzen with the LDO, I wonder how Intel's voltage regulation differs from it? It seems that this is a big factor in lowering the power consumption of the Zen cores.
 
B

bjt2

Senior member
Sep 11, 2016
784
180
86
KTE said:
You once 'calculated' 4GHz base 95W LAUNCH too... With 4.5GHz Turbo. Based on your FO4+Neon claims.

Say so a billion wrongs things and you're bound to get something right.

Sent from HTC 10
(Opinions are own)
Click to expand...

I have not lost hopes that this can happen. If you remember well, i also stated that probabily 4GHz will not be achieved in the first batches, but maybe later in the year. Don't report only the part that is, for now, wrong... Finally, if the frequency leaked in those e-shops are base and not turbo, then i was right...
 
B

bjt2

Senior member
Sep 11, 2016
784
180
86
Sven_eng said:
Didn't you support Kanter who claimed a 3.2GHz base clock and "unknown" turbo? At least this bjt guy has balls enough to stick his neck out on turbo speeds.

If 3.6GHz base clock is true then Kanter is at least 12.5% out on his base clocks. If Zen launches at 4.0GHz turbo then bjt will also be 12.5% out on that with his 4.5GHz assumption. It would be smart to assume 3.6GHz is the very least and 4GHz too. ;)

Yeah I'm aware that clocks don't scale that way esp with power, but consider how different the scenario is to what you all thought it would be even only 6 months ago. :)
Click to expand...

I predicted 4 base and 4.5 turbo (but on another forum, more AMD friendly, i made a stronger prediction: 4 base and 4.8-5 max turbo).
 
A

Andrei.

Senior member
Jan 26, 2015
316
386
136
tamz_msc said:
Looking at the per-core voltage regulation in Ryzen with the LDO, I wonder how Intel's voltage regulation differs from it? It seems that this is a big factor in lowering the power consumption of the Zen cores.
Click to expand...
It looks vastly superior to what Intel is doing.
 
B

bjt2

Senior member
Sep 11, 2016
784
180
86
Rngwn said:
Anyone knows so far if XFR could be turn on-off in BIOS setting? This might sound weird, but this could be important in research works. You would want the clock frequency to be relatively constant while you are benchmarking some new algorithms.
Click to expand...

I think will be sufficient lower the TDP knob to the minimum to lock the frequency to the base clock...
 
B

bjt2

Senior member
Sep 11, 2016
784
180
86
Some thoughts on two slides: (excuse me for font size but with small font is unreadable)

2_s.png



Last line can even mean lower FO4 than Excavator.


3_s.png



Second line: XV was a 9T (HDL) design. The second phrase can mean that Zen is made with 10.5T libraries. Namely the Highest performance...
 
The Stilt

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
tamz_msc said:
Looking at the per-core voltage regulation in Ryzen with the LDO, I wonder how Intel's voltage regulation differs from it? It seems that this is a big factor in lowering the power consumption of the Zen cores.
Click to expand...

Intel's FIVR is a ultra high frequency (>= 140MHz) switching converter. AMD's converters are digital low-dropout (dLDO). The consumers should not worry about these, for reasons I will not disclose until the release.
 
  • Like
Reactions: Drazick, KTE, Phynaz and 6 others
E

Erenhardt

Diamond Member
Dec 1, 2012
3,251
105
101
The Stilt said:
Intel's FIVR is a ultra high frequency (>= 140MHz) switching converter. AMD's converters are digital low-dropout (dLDO). The consumers should not worry about these, for reasons I will not disclose until the release.
Click to expand...

Hey, can you share info if we will be able to overclock Ryzen past the stock TDP. Will motherboards allow disabling TDP throttling?

It would be huge issue if 65W Zen couldn't be OCed past the TDP limit without custom bios :p
 
CatMerc

CatMerc

Golden Member
Jul 16, 2016
1,114
1,153
136
The Stilt said:
Intel's FIVR is a ultra high frequency (>= 140MHz) switching converter. AMD's converters are digital low-dropout (dLDO). The consumers should not worry about these, for reasons I will not disclose until the release.
Click to expand...
Not worry about it as in the implementations are equal in results?
 
The Stilt

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
CatMerc said:
Not worry about it as in the implementations are equal in results?
Click to expand...

Personally, I absolutely hate both of them.
They are two completely different animals, but they have the same down sides.
Consumers never have to deal with any of the dLDO related stuff. Business as usual.
 
  • Like
Reactions: Drazick
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom