Why is AMD's Zen more efficient than intel's Core?

Discussion in 'AMD Processors' started by Peppercorn, Mar 16, 2017.

  1. noko

    noko 2[H]4U

    Messages:
    3,091
    Joined:
    Apr 14, 2010
    Some consideration as well, some boards are overvolting the cpu. For example my board is putting 1.43v at stock speeds when it can do that at less than 1.3v plus the board is reading low on the volts making the actual volts going to the cpu even higher. So once all of this gets corrected I expect future power readings to be lower. At least on the ASUS Crosshair 6 Hero. So we are looking at roughly the same performance of a 6800K in multi threading, about the same power at half the cost to a third of the cost for the cpu if you go with a 1700 and OC. AMD is doing great here! To match efficiency, performance, power at much less the cost - good for all of us.
     
  2. Formula.350

    Formula.350 Gawd

    Messages:
    666
    Joined:
    Sep 30, 2011
    I just want to make sure that you are aware the Ryzen chips function different from past models, yes? Apparently most software (as I understand it) either isn't currently able to, or may even not be possible to, read what voltage the core is actually getting when the motherboard's voltage is set to Stock. When it's running default, as in when it hasn't set the "overclocked" MSR, the voltage will automatically adjust based on what it's internal voltage determines is needed (I think this is what AMD calls SenseMI?), which can sometimes exceed the "default" voltage amount. Often times though it is running lower than that, but the only way to get it to run consistently at a set voltage is to manually set the VCore (and crank the Load Line Calibration to max so it doesn't fluctuate wildly with load).

    Again, if I understood what TheStilt spoke about in his Anandtech thread, then the reading of 1.43V is just what the board sets as "default" voltage, but since it is set to "default" AMD's internal controller takes over to regulate everything. Therefore, your CPU being able to run stable at base speeds when manually set at 1.3V likely means that when set at "default" that's what it is typically running at even though you can't tell with software readings. :)
     
  3. noko

    noko 2[H]4U

    Messages:
    3,091
    Joined:
    Apr 14, 2010
    My understanding in a non OC condition the cpu SMU controls all the cores voltages and offsets making it hard for software to follow. Once you go to OC mode it is the motherboard now controlling the voltage and no longer the cpu, in that case the voltage readings should be correct but currently may not be with the current software.
     
  4. JustReason

    JustReason razor1 is my Lover

    Messages:
    2,363
    Joined:
    Oct 31, 2015
    That is what the Stilt reported. Asus has this issue with all AMDs in my house. Wifes 7870k overclocked to 4.5Ghz from stock on stock voltage. My 8350 lists as 1.2775 VID but Asus sets it at 1.38V stock, way above what I need.
     
  5. JustReason

    JustReason razor1 is my Lover

    Messages:
    2,363
    Joined:
    Oct 31, 2015
    LOL really? Well it is a good thing everyone has him on ignore then huh.
     
  6. noko

    noko 2[H]4U

    Messages:
    3,091
    Joined:
    Apr 14, 2010
    who ????
     
    Simplyfun likes this.
  7. CSI_PC

    CSI_PC 2[H]4U

    Messages:
    2,243
    Joined:
    Apr 3, 2016
    Tom's Hardware has it that you can correlate such power demand just need to look at the torture power test, unfortunately only way to really compare them is with a benchmark utility that can stress all cores and SMT.
    The 6800K in this situation has a lower power demand but then it is 6-Core.
    The 1800X at 3.8GHz all cores is pretty similar to 6900K albeit it could be argued if 1800X managed 4GHz all cores it would be a little bit worse than the 6900K;

    The one to be interested in for 1800X is the 3.8GHz (all cores) Luxrender:

    [​IMG]


    4.0 and 4.3 are all cores.
    Also worth noting the 3.2GHz default also ran all cores to 3.7GHz with turbo boost (possibly only a few of the motherboard manufacturers do this with the microcode), hence why it is closeish to the power demand of 4GHz.

    [​IMG]


    Here is the 6800K albeit 6-Core.

    [​IMG]


    Links:
    http://www.tomshardware.com/reviews/amd-ryzen-7-1800x-cpu,4951-11.html
    http://www.tomshardware.com/reviews/intel-core-i7-broadwell-e-6950x-6900k-6850k-6800k,4587-9.html

    Cheers
     
    Last edited: Mar 21, 2017
  8. juanrga

    juanrga [H]ard|Gawd

    Messages:
    1,114
    Joined:
    Feb 22, 2017
    That link is irrelevant. Glofo does not always report all the processes available. For instance that link lacks the 28SHP used in Kaveri/Godavari or the 32PD used by Vishera chips.

    We know from IBM that Power9 uses Glofo 14HP. It doesn't matter if Glofo list that process or not.
     
  9. CSI_PC

    CSI_PC 2[H]4U

    Messages:
    2,243
    Joined:
    Apr 3, 2016
    Actually not irrelevant, in fact more relevant than you saying there is 14HP because IBM says so even though GF is the fab manufacturer.

    And I am still waiting on you to provide anything that is not IBM slide and source :)
    At least I bothered and showed one of the companies that use the foundry showing they do not have anything but 14LPP for customers even up to end of the schedule calendar year :)

    So your proof is that because IBM says HP it must be a general process available to all customers (I keep repeating I have reports suggesting it is an IBM specific process) and just because GF makes no mention of 14HP in any of their fab news/research it absolutely exists for all customers because well they do not mention something important like that :)
    Like I said I am not necessarily disagreeing, but you have given nothing beyond assumptions just because a slide of a customer mentions 14HP (who as I said potentially have their own process at GF that is not a general option).
    Just in case it gets skewed; I am saying if there is a High Performance process ( a big if as no other sources), so far it looks more like it is IBM walled garden.
    Cheers
     
    Last edited: Mar 21, 2017
  10. juanrga

    juanrga [H]ard|Gawd

    Messages:
    1,114
    Joined:
    Feb 22, 2017
    I guess you refer to Jim Keller and you are repeating two myths spread in forums. The first myth that a single man can design a modern microarchitecture, when they are so complex that are designed by teams of dozens of engineers. The second myth the pretension that Keller is the lead architect because he wasn't.

    K7 --> Two teams, one leaded by Dirk Meyer and other by Fred Weber, with Meyer being the lead engineer.

    K8 --> A team leaded by Fred Weber as chief architect.

    Zen --> A team leaded by Suzanne Plummer with Mike Clark being the lead engineer.
     
  11. juanrga

    juanrga [H]ard|Gawd

    Messages:
    1,114
    Joined:
    Feb 22, 2017
    Not me, but the people that believed that Zen was 60% more efficient than 6900k.

    Some of us knew since before launch that "95W" was marketing label and reviews have confirmed it.

    At contrary, reviews show that "95W" on the 1800X model is a marketing label, and that the real TDP is 125W or 130W. That is why Ryzen consumes (and dissipates) amounts of power similar to the FX-8350 and FX-8370.
     
    Shintai likes this.
  12. CSI_PC

    CSI_PC 2[H]4U

    Messages:
    2,243
    Joined:
    Apr 3, 2016
    Keller was heavily involved in the engineering at the earlier development stages of Zen (laying the fundamental design process) and then backed off to be a more high level engineer architect, he also made Mike Clark the lead engineer who worked/co-ordinated with Keller on Zen development.
    Cheers
     
    strategyfreak and ManofGod like this.
  13. ManofGod

    ManofGod [H]ardness Supreme

    Messages:
    8,108
    Joined:
    Oct 4, 2007
    Are you freaking kidding me? He is now reporting folks that have been here for a lot longer and share real experiences? Screw him, he is going to remain on my ignore and I am not reading jack of his anymore. :) Oh, and Ryzen is kicking butt, taking Intel names and all doing it well running cooler and more efficient as well! :D
     
    strategyfreak and JustReason like this.
  14. juanrga

    juanrga [H]ard|Gawd

    Messages:
    1,114
    Joined:
    Feb 22, 2017
    Keller was hired by Rory Read for developing the K12 and Skybridge projects. He worked on both projects leading a team of x86 engineers and he tell us some funny anecdotes about the lack of experience of the people in his team with the ARM ISA. Precisely Keller was hired for their past experience with ARM at Apple.

    Lisa Su first canceled SkyBridge and latter canceled K12 (which no longer appears in official roadmaps) and Keller left AMD because his work had finished.

    Keller has never been the chief architect of Zen, neither of K7, K8, or Apple cyclone.
     
  15. juanrga

    juanrga [H]ard|Gawd

    Messages:
    1,114
    Joined:
    Feb 22, 2017
    Relax. I wasn't reporting you. No one was reporting you. Just a problem with formatting that has been solved. But you are not going to read this. :p
     
  16. CSI_PC

    CSI_PC 2[H]4U

    Messages:
    2,243
    Joined:
    Apr 3, 2016
    Keller was employed by Papermaster who was looking to create the team also for Zen, Keller put Mike Clark as lead engineer on Zen after the fundamental early work was done.
    Rory Read is the CEO and not even an engineer, he allowed Papermaster who is the CTO and an engineer to do what he wanted for this project.
    And yeah I have worked closely with CTO/CSO in some high tech companies so I have a very good understanding how they operate.
    Cheers
     
    razor1 likes this.
  17. razor1

    razor1 [H]ardForum Junkie

    Messages:
    9,624
    Joined:
    Jul 14, 2005

    Yeah I'm going to have to agree with ya on that one, my direct bosses are , one is a COO and one is a CTO, the CTO has direct control over all tech departments, there is no questions about it. I would think any tech company would operate the same.
     
    CSI_PC likes this.
  18. juanrga

    juanrga [H]ard|Gawd

    Messages:
    1,114
    Joined:
    Feb 22, 2017
    And Papermaster was following Read strategic plan.
     
  19. CSI_PC

    CSI_PC 2[H]4U

    Messages:
    2,243
    Joined:
    Apr 3, 2016
    He is a freaking CEO.
    He just agrees to the proposal (and only presented as a very high level plan) from Papermaster and so my point is correct.
    Otherwise might as well say most of us are employed by the CEO because they give the goahead to VPs of departments!
    You really cannot argue on this one.
     
    NKD likes this.
  20. CompletelyBrokn

    CompletelyBrokn n00bie

    Messages:
    37
    Joined:
    Oct 27, 2014
    My point was they had hired the man on to work on Zen and the previous Athlon. I guess news articles could have it wrong, but I read several from many reputable sources that made it clear he was hired to work on it. Nobody else was really mentioned as far as bringing people into work on Zen. You don't just hire someone on because they are your buddy. You hire them on for a project because you NEED them. AMD had to have his help, regardless of the role he played, which must have been pretty important to make the news. My concern is why...is there nobody else that AMD has full time that can that? I could be wrong, but I don't recall Intel going out and bringing someone in just for a processor remake.
     
  21. razor1

    razor1 [H]ardForum Junkie

    Messages:
    9,624
    Joined:
    Jul 14, 2005

    Vision and experience is hard to replicate. Keller has both those, my guess is since the Athlon heydays AMD lost most of their engineering talent in both those categories.
     
  22. juanrga

    juanrga [H]ard|Gawd

    Messages:
    1,114
    Joined:
    Feb 22, 2017
    In reality the original plan was from Feldman that convinced Read to go ARM. And Read gave instrutions to Papermaster which hired Keller due to his former experience with ARM. Papermaster joined two teams: one leaded by Keller (K12 & Skybridge) another leaded by Plummer (Zen). Once Su replaced Read, she changed the company plans why 180º. The first to leave AMD was Feldman and the last to abandon was Keller. We also know from inside that Keller had some issues with Lisa Su when she canceled Keller's baby: K12. Papermaster remains at AMD and he push either towards ARM or towards x86 or both (ambidextrous) depending on what his chief decides.
     
    Last edited: Mar 21, 2017
  23. juanrga

    juanrga [H]ard|Gawd

    Messages:
    1,114
    Joined:
    Feb 22, 2017
    Keller wasn't the chief architect of Athlon. I already gave above the names of the chief architects of K7 and K8.

    Those reputable sources are wrong and used a fake history, that Keller was the chief architect of K8, to build hype around Zen and pretend that Zen was going to be the new K8.

    Lisa Su already admitted that the chief architect of Zen is Mike Clark and she doesn't even mention Keller, when people ask her about the people behind Zen. But hype sites as BitsandChips continue pretending otherwise.
     
  24. CompletelyBrokn

    CompletelyBrokn n00bie

    Messages:
    37
    Joined:
    Oct 27, 2014
    Sources? You are asking me to ignore multiple news sites and trust an anonymous forum member. Besides, I am not arguing who was in charge. I'm stating whoever was in charge wasn't enough at AMD.
     
    strategyfreak likes this.
  25. razor1

    razor1 [H]ardForum Junkie

    Messages:
    9,624
    Joined:
    Jul 14, 2005
    if you look at the time lines of when Keller was working on what its pretty clear to see what happened. He wasn't head of K8 development till the end of its development life cycle if I remember correctly.
     
  26. CSI_PC

    CSI_PC 2[H]4U

    Messages:
    2,243
    Joined:
    Apr 3, 2016
    And you still want to argue on this subject.
    The CEO gives a high level direction and signs off on projects presented by VPs and CTO/CSO.
    Papermaster was the linchpin behind Zen, see his presentations and interviews, not denigrating Rory as he was critical in the some key strategy plays for AMD but for this discussion and context his role is small.
    Read signs off on the high level plans and directions.
    Papermaster came up with the commitments and strategy for Zen and employing Keller as being integral.

    Do not take my words for it here is a quick search and what Clark said and what came out of HotChips:
    Papermaster the driving force and also the one who wanted to merge and have a single team working on both ARM and x86 and asked Keller to run the team and development strategy, in fact he stated this in the AMD video back in May/June 2014
    Anyway there are other articles with Papermaster and with Keller showing they were much more involved than you suggest, and Keller "involvement" drove the fundamental early design phase of Zen before handing off more to Clark while then always involved at a higher level engineer architect perspective and focus.
    Same with Papermaster who was championing Zen early without having to buy into it from Clark (who started talking about Zen after Keller left and was only more senior involved after the Zen R&D design started and not at the very beginning).
    Cheers
     
    Last edited: Mar 21, 2017
  27. juanrga

    juanrga [H]ard|Gawd

    Messages:
    1,114
    Joined:
    Feb 22, 2017
    K7 teams and the lead architect from official presentation of the chip at Microprocessor forum

    [​IMG]

    "Keller was never the guy behind the K8" confirmed by David Kanter from MPR.

    http://www.realworldtech.com/forum/?threadid=153598&curpostid=153941

    List of AMD Architects, including guys begin X86-64

    https://people.cs.clemson.edu/~mark/architects.html#wintel

    Lisa Su admitting Mike Clark is the lead engineer for Zen. She didn't mention Keller, despite she was asked about him:


     
    Last edited: Mar 22, 2017
    Shintai likes this.
  28. CSI_PC

    CSI_PC 2[H]4U

    Messages:
    2,243
    Joined:
    Apr 3, 2016
    Lisa saying all of this after Keller left, while ironically Clark actually stated: " Keller “was involved in the early days of Zen, we worked together on the arch and he made me lead architect for it because he was running the whole [processor design] group,” said Clark. “The engineering team loved him because he’s an engineer at heart and you felt you had a champion,” he said.
    Amazing how Clark managed to mess up Steamroller as lead engineer and at same time somehow able to work on Zen architecture (started late 2012)....

    Clark was not key in the early days of the R&D development of Zen because he was working on Steamroller and came on board with Zen once the foundations had been laid by Keller, only after Steamroller did Clark work closer with Keller on the design.
    Once Keller was happy with the direction-foundation he then stepped back into a higher more hands off technical architect role for Zen while promoting Clark to lead architect but still being involved from an engineering architect focus and certain decisions.

    But nearly every large company will downplay the loss of key personel on projects or product development, although it is fair to say the whole team of engineers contributed to Zen.
    Cheers
     
  29. DuronBurgerMan

    DuronBurgerMan Gawd

    Messages:
    632
    Joined:
    Mar 13, 2017
    Side note: the idea that AMD's marketing numbers are bullsh*t should surprise exactly nobody. 99% of marketing is bullsh*t, irrespective of the company in question. Never believe the marketing numbers. Always benchmark your sh*t separately. And regardless of what AMD's marketing department says the TDP is, it's actually pretty good relative to the competition, and miles ahead of where AMD was at with Crapdozer. So all to the good.
     
  30. juanrga

    juanrga [H]ard|Gawd

    Messages:
    1,114
    Joined:
    Feb 22, 2017
    I don't know if I miss this in the original Hardware.fr article or was added during the update. I am translating from French:

    This confirms what CanardPC and me have been saying since before the launch: that "65W" and "95W" are marketing labels. And it confirms why reviewers found that '95W' R7 1800X dissipates amounts similar to 125W FX 8370, whereas the '65W' R7 1700 dissipates amounts similar to 91W i7 7700k.

    Case closed.
     
    Last edited: Mar 25, 2017
    Shintai likes this.
  31. dgz

    dgz 2[H]4U

    Messages:
    3,971
    Joined:
    Feb 15, 2010
    Were you a sleep last time it happened? Cause they got fined even in the US, you know.
     
  32. Tsumi

    Tsumi [H]ardForum Junkie

    Messages:
    12,256
    Joined:
    Mar 18, 2010
    Because the people running Intel now are the exact same people as 10 years ago, and they did not learn their lesson.
     
  33. CSI_PC

    CSI_PC 2[H]4U

    Messages:
    2,243
    Joined:
    Apr 3, 2016
    Not sure any tech company would risk it (they will walk a fine line but not go overtly antitrust) these days considering just how aggressive Europe is on penalising monolopy behaviour, especially as Intel was found guilty by the European Commission back then as well.
    Not only does the European Commission do some hefty fines these days, it also links in with changing the company investigated business behaviour-operation or suffer continual penalties.
    Google is finding this out now, they will try court to delay/modify it but that just makes the commission add more pressure just as they did against MIcrosoft.

    Cheers
     
  34. Peppercorn

    Peppercorn Limp Gawd

    Messages:
    225
    Joined:
    Dec 8, 2016
    Simplyfun likes this.
  35. Simplyfun

    Simplyfun [H]ard|Gawd

    Messages:
    1,177
    Joined:
    Dec 17, 2016
    Some of the resident...... persons will be along shortly to explain this simply isn't possible and that mistakes were made instead of just accepting it for what it is.
     
    Last edited: Mar 27, 2017
    JustReason likes this.
  36. Peppercorn

    Peppercorn Limp Gawd

    Messages:
    225
    Joined:
    Dec 8, 2016
    No doubt. On the brighter side for intel, winter is coming in the Southern Hemisphere, so their CPUs will make fine spaceheaters for those people. :)
     
  37. Simplyfun

    Simplyfun [H]ard|Gawd

    Messages:
    1,177
    Joined:
    Dec 17, 2016
    Well I don't think things are that severe, I mean it's about time really they got parity with Intel and a decent process to build CPU's on. I'm really curious to see how they do with their new APU's given they will still be using GCN for the first round.
     
  38. Peppercorn

    Peppercorn Limp Gawd

    Messages:
    225
    Joined:
    Dec 8, 2016
    I think it's very bad. This is intel's what, 3rd or 4th gen 14nm product vs AMD's 1st on, as everyone claims, a much worse process node. There's no roses here for intel, their chips are running substantially hotter and inefficient compared to AMD's chips. This is especially significant for Naples.
     
  39. Tsumi

    Tsumi [H]ardForum Junkie

    Messages:
    12,256
    Joined:
    Mar 18, 2010
    The 4 ghz clock speed limitation is a good reason why those results are what they are. A low power optimized process will not clock high, and vice versa. AMD got close to beating Intel, but did not get there yet. If Zen had been able to hit at least 4.5 ghz, then you can claim Intel got its ass handed to them. But then again, if it did clock that high, we would probably see prices at $450 for the 1700 and $700+ for the 1800x.

    TL;DR AMD has competitive performance with very competitive pricing. Nothing more, nothing less. AMD did not steal any crown from Intel.
     
  40. Shintai

    Shintai [H]ardness Supreme

    Messages:
    5,301
    Joined:
    Jul 1, 2016
    Dont tell me you try and compare 2 completely different temperatures(cores vs single diode package) and also in most of your cases performances.

    AMD even mess up temperatures between its own SKUs. Not to mention the measurement as is done against an internal table :ROFLMAO:
     
    Last edited: Mar 28, 2017
    juanrga likes this.