Zen2's 7nm Complications: Why not All Ryzen 3000 Cores Are Created Equal

Discussion in 'HardForum Tech News' started by Auer, Jul 30, 2019.

  1. 1_rick

    1_rick Gawd

    Messages:
    571
    Joined:
    Feb 7, 2017
    BTW, the infamous (?) thread AMD_Robert commented in says "
    No, the 3900X is not being overvolted. The onboard firmware in the CPU controls that, not the power plan. What the power plans do change, however, is whether or not the CPU downclocks at idle or jumps straight to power-gated sleep (cc6 power state).

    The out-of-box plans set a min CPU clock of 10%, so the CPU will downclock before going to sleep. Voltage will follow.

    Our plans set a min CPU clock of 90%, and then instruct the core to go straight to sleep if not in use. Windows is unable to probe the behavior of the core when it's sleeping, so voltage will not appear to follow."

    That's not what I see. AMD Ryzen Balanced sets min/max of 0/100%. AMD Ryzen High Performance sets 100/100. AMD Power Saver sets 0/100, but changes the System Cooling Policy from Active (which the other two have) to Passive. The mouseover text says Active increases fan speed before lowering clocks and Passive does the opposite. I have the 1.07.07, I think the version is, chipset drivers--the first one available, and re-downloaded and re-applied this morning.
     
  2. NKD

    NKD [H]ardness Supreme

    Messages:
    7,719
    Joined:
    Aug 26, 2007
    what is your ambient. I am around 20c and my temps are 70s for cinebench all core and stock boost is around 4150-4200. But it used to be better with out of box bios so thats why I am confident they will eventually tweak it as AMD updates their AGESA versions quiet often and they become mature as the time goes on.
     
  3. 1_rick

    1_rick Gawd

    Messages:
    571
    Joined:
    Feb 7, 2017
    The AC is set for 23C, but the thermostat's in another room. I just brought a portable thermometer in here; it needs a few minutes to settle but it was reading 2C higher (it had been near a west-facing window, and it was 100F today here in Dallas.)

    Edit: 22.7 C / 73 F is the thermostat setting; the room with my computer is 25 C / 77 F
     
    Last edited: Jul 30, 2019
  4. mothman

    mothman [H]ardness Supreme

    Messages:
    5,003
    Joined:
    May 28, 2009
    I can't speak with the technical prowess of most of you guys but if the Ryzen 3000's employ cores with speed variations causing erratic behavior how can any software/firmware/bios update have any effect at the hardware level ?
     
  5. 1_rick

    1_rick Gawd

    Messages:
    571
    Joined:
    Feb 7, 2017
    FYI: there's a new chipset driver available *today*. Check this out (from Reddit): "You want to make sure the new, new power plan is installed. Uninstall the old Ryzen Ballanced plan first. (EDIT: Remove all Ryzen plans, you might have up to three, Ryzen Ballanced, Ryzen High Power and Ryzen Power Saver - Power Saver seems to have been phased out.)"




    Edit: There is a new Ryzen Master today, too.
     
  6. gerardfraser

    gerardfraser Gawd

    Messages:
    911
    Joined:
    Feb 23, 2009
    Posted Power saver plan and how to import it after new chipset drivers .Post #35
    Zen2's 7nm Complications: Why not All Ryzen 3000 Cores Are Created Equal
     
  7. 1_rick

    1_rick Gawd

    Messages:
    571
    Joined:
    Feb 7, 2017
  8. 1_rick

    1_rick Gawd

    Messages:
    571
    Joined:
    Feb 7, 2017
    So--with the new plan and Ryzen Master, now the clock speeds RM reports don't match with CPU-Z, which I suppose could be differences in sample timing. RM doesn't report any better clocks, but CPU-Z says that during a bench, I get 4 clocks at 4.250, which is a tiny bump. I guess I just leave it until the next BIOS revisions.
     
  9. NKD

    NKD [H]ardness Supreme

    Messages:
    7,719
    Joined:
    Aug 26, 2007
    I agree. I turned off PBO and my boost clocks actually went up a little. But I am still hoping for new AGESA to further refine things. I did get better boost clocks up to what amd stated for single core with out of the box AGESA on my asus crosshair VIII hero. But damn asus wont send me the original file lol or has it available to flash back on their website. Asus seems to have gone down hill in support.
     
  10. primetime

    primetime [H]ardness Supreme

    Messages:
    6,024
    Joined:
    Aug 17, 2005
    that is unusual......they usually have what seems like every bios revision released for my board...and yours is just newer
     
  11. NKD

    NKD [H]ardness Supreme

    Messages:
    7,719
    Joined:
    Aug 26, 2007
    yea I chatted with support twice and they asked me for my last for numbers on the board- its written next to socket. I gave it to them and said that is the bios shipped with board, okay! Then said engneering will send it to me via email. Both times no response for a week lol.
     
  12. primetime

    primetime [H]ardness Supreme

    Messages:
    6,024
    Joined:
    Aug 17, 2005
    so your issue is you forgot to backup the shipping bios? before you tried the newer one? i woulnt worry about it....Asus will release ten bios revisions for your board in the next year alone. that little 1% difference in this and that.....ill take a bullet proof stable setup 24/7 anyways (which i think is more asus direction these days)
     
    Jim Kim likes this.
  13. Meeho

    Meeho [H]ardness Supreme

    Messages:
    4,470
    Joined:
    Aug 16, 2010
    Just wait for the next AMD driver/BIOS/power plan/OS scheduler/cosmic star alignment to fix everything.

    And AMD isn't lying about their CPU spec, they've just implemented surprise boost mechanic, and surprise core performance.
     
    Last edited: Jul 31, 2019
    N4CR and ZeroBarrier like this.
  14. NKD

    NKD [H]ardness Supreme

    Messages:
    7,719
    Joined:
    Aug 26, 2007
    Read somwhere it might be on the cd drive. I used to have a portable cd drive but I moved and too lazy to order another one lol. Yea its stable so far, so not too concerned.
     
  15. Ranger101

    Ranger101 [H]Lite

    Messages:
    109
    Joined:
    Sep 11, 2015
    Whoa, easy there cowboy, that might just be a bit too much reality for some folks to handle ;)
     
    N4CR and /dev/null like this.
  16. Auer

    Auer Gawd

    Messages:
    769
    Joined:
    Nov 2, 2018
    I think it's more in the line with "AMD is having some issues here with this launch and if youre not desperate maybe hold off for a little bit".
    Hopefully all is fixed soon and this is just the usual AMD launch stuff.
     
  17. Auer

    Auer Gawd

    Messages:
    769
    Joined:
    Nov 2, 2018
    Not bored, just [H]ard.
     
  18. viivo

    viivo [H]ard|Gawd

    Messages:
    1,309
    Joined:
    Sep 7, 2005
    Faceless corporations don't need White Knights.
     
    ZeroBarrier and DooKey like this.
  19. sabrewolf732

    sabrewolf732 2[H]4U

    Messages:
    3,999
    Joined:
    Dec 6, 2004
    Fwiw, my 3600 is pretty much boosts to 4.2GHz on all cores all the time on water.
     
    Auer likes this.
  20. mothman

    mothman [H]ardness Supreme

    Messages:
    5,003
    Joined:
    May 28, 2009
  21. 1_rick

    1_rick Gawd

    Messages:
    571
    Joined:
    Feb 7, 2017
    download.amd.com/drivers/amd_chipset_drivers.exe -- I got them last night.
     
  22. mothman

    mothman [H]ardness Supreme

    Messages:
    5,003
    Joined:
    May 28, 2009
    Hey thanks man, I pasted your link and it downloaded the chipset driver. Where on AMD do you actually go for these I was interested in that 5.0.0 Power Plan
     
  23. purple_monster

    purple_monster Gawd

    Messages:
    553
    Joined:
    Jun 1, 2018
    wjats the workaround
     
  24. SunnyD

    SunnyD 2[H]4U

    Messages:
    2,961
    Joined:
    Jul 6, 2004
    Buy the SKU with the next high base clock.
     
  25. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    28,446
    Joined:
    Oct 29, 2000
    This cannot be fixed. It's a limitation of the silicon, and how they worked around it to get even more out of it.

    Best you can do is upgrade to Windows 10 1903 to get the new scheduler and use the latest AMD chipset drivers. At least this way the OS is aware and can make the best out of it.
     
  26. 1_rick

    1_rick Gawd

    Messages:
    571
    Joined:
    Feb 7, 2017
    Go to their download site and look for chips drivers.

    I deleted the Ryzen plans manually before installing, and it recreated them automatically, only slightly tweaked.
     
  27. Trimlock

    Trimlock [H]ardForum Junkie

    Messages:
    15,157
    Joined:
    Sep 23, 2005
    You hardly notice when your ignore list includes several hundred.
     
    ZeroBarrier and Auer like this.
  28. defaultluser

    defaultluser [H]ardForum Junkie

    Messages:
    12,562
    Joined:
    Jan 14, 2006
    I'm not white Knighting when it's the truth. I see you failed to contradict a single point I made, just whipping out the tried and true name calling.

    You're spreading pointless click bait from a site that has been worthless since Pabst sold the place in 2006



    Come back when you have a better complaint than Ryzen 2 BIOS still a work- in - progress, and will be finalized in the next few months. In the meantime, you already get b98 percent of the final performance.

    It's not as if Ryzen 2 is slower than they promised over ryzen plus, so making a big deal out of a small boost clock deficiency is in my mind just looking for ANYTHING to whine incessantly about. It screams of unhealthy obsession
     
    Last edited: Jul 31, 2019
    blackmomba likes this.
  29. Auer

    Auer Gawd

    Messages:
    769
    Joined:
    Nov 2, 2018
    Calm down.

    Tom's stated: We aren't accustomed to AMD being the target of enthusiast ire, but Ryzen 3000's seeming inability to hit its advertised boost clocks has led to complaints on enthusiast forums and reddit. We ran a series of tests to determine if our Ryzen 5 3600X, which we purchased at retail, can hit its advertised 4.4 GHz boost clock. We discovered several new things along the way. We shared our discoveries with AMD before publication, and the company has confirmed several of our findings.

    Whatever people think of Tom's, there it is.
     
  30. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    28,446
    Joined:
    Oct 29, 2000
    I try to avoid putting people on my ignore lists, even when I find them annoying or irrational, because I worry more about placing myself in a self imposed echo chamber than I do about being a little annoyed by fanboys every now and then.
     
  31. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    28,446
    Joined:
    Oct 29, 2000
    Even a stopped clock is right twice a day.

    That, and while Tom's may not have been my favorite source for Tech reviews in the past (it's amazing how well they buried that one scandal where they took payment for good reviews) I feel like a lot of the garbage that is out there today is making them look comparatively better. We've lowered the standards in what we expect from tech journalism over the last 10 years, and in this new normal, what used to look like bad reporting and testing from Tom's now looks better than average.

    Sad right?
     
    Soulstorm brew and Auer like this.
  32. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    28,446
    Joined:
    Oct 29, 2000
    Can you think of even one instance in human history when telling someone to calm down actually worked? :p
     
    Mav451 and Auer like this.
  33. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    28,446
    Joined:
    Oct 29, 2000
    I sadly agree.

    I'm no Intel fan, but they would have been raked over the coals for doing something similar.

    There are some inevitable contributors to this situation though. Higher leakage current at lower node sizes pretty much made it inevitable that they would have to get more creative to get as much as possible out of the architecture. I think Intel is going to hit the same snags when they get to 7nm.

    We are used to thinking of AMD as the more honest company though. The one that contributes to open standards, is more community friendly and straightforward, but here they really failed in that expectation.

    While these types of methods may be the inevitable future state of things as we keep racing towards the limits of silicon, they should have explained this, explained how their new boost process works and included these things in the review guides. You know, educated their consumers, They also shouldn't have advertised boost clocks the chips couldn't hit. A Ryzen 9 3900x is a pretty awesome CPU even if the official boost clock is 4.3 or 4.4. They didn't have to misrepresent reality. In that regard I am disappointed in AMD.

    All that said, I am still probably buying either a 3950x or Threadripper 3 on launch, because I am happy to finally have some real competition in the CPU space, and I want to support the underdog.
     
  34. gerardfraser

    gerardfraser Gawd

    Messages:
    911
    Joined:
    Feb 23, 2009
    AMD Ryzen can hit advertised boost speeds.Why do I know well I posted in this thread and I can not speak for everyone but for myself and my 2 x Ryzen CPU's surpass advertised boost speeds.

    WTF am I missing here with the post above, does Intel not advertise a boost speed of 5000Mhz on the box but runs at lower clocks.Really what am I missing here.The Intel does not run at 5.Ghz all the time and the AMD parts are the same way,just some motherboards have crappy BIOS or Ryzen works different and people can not use. Or I am the luckiest guy in the world where my Ryzen CPU works different than everyone else. Oh crap my Intel machines I am lucky there also.Buying a lottery ticket tonight.

    Hey also does not the same price slow crap Ryzen 3900X beat 9900K in productivity at lower clocks,that is terrible.I disagree with Tom's article and bored so I am posting something.
    EG:Intel Core i9-9900K Coffee Lake 8-Core, 16-Thread, 3.6 GHz (5.0 GHz Turbo) LGA 1151 (300 Series)
     
    blackmomba likes this.
  35. Chimpee

    Chimpee [H]ard|Gawd

    Messages:
    1,451
    Joined:
    Jul 6, 2015
    You do realize just because you can hit the clock speed and surpass it, doesn't mean everyone can. Professional reviewers have seen a lot of variability in boosted clock speed (some cannot even hit the boosted rate) that should warrant investigation of what is going on.
     
  36. Zarathustra[H]

    Zarathustra[H] Official Forum Curmudgeon

    Messages:
    28,446
    Joined:
    Oct 29, 2000
    Of course it isn't.


    The definiton of overvolting is to use a voltage higher than specified by the manufacturer.

    AMD IS the manufacturer. They specify the voltages.

    If AMD does it, it can't be overvolting :p
     
    ZeroBarrier likes this.
  37. Gideon

    Gideon 2[H]4U

    Messages:
    2,303
    Joined:
    Apr 13, 2006
    Most of it is likely caused by different AGESA bios and people running 370 and 470 chipsets. Most everyone I have seen on a 570 have had no issues with reaching boost clocks or very close and you need the proper tools to measure it properly as it tends to happen only for a very short time. My chip is not boosting right but I am also on a X370 and running a beta bios as the non beta bios caused way to much memory grief for me. I expected to have issues anyway since I am using a first gen motherboard, price of being bleeding edge and pretty much been the same since I first started building computers for myself. However even with all that my chip has scored higher then most reviews did in multicore tests and is just barely behind in single core tests. Honestly people are making a bigger deal then it is, the chip performs just like the reviews said it would the top MHz it reaches is not a big deal.
     
  38. NKD

    NKD [H]ardness Supreme

    Messages:
    7,719
    Joined:
    Aug 26, 2007
    Nope its across the board. I have x570 VIII hero. Shamino today uploaded the 1.0.0.3ABB link and basically said don't expect better boost clocks from AGESA 1.0.0.3 FW version because 1.0.0.2 had the right boost clocks. So it looks like Agesa is the reason for the boost clocks and they just need to iron this shit out.
     
    Auer likes this.
  39. Algrim

    Algrim [H]ard|Gawd

    Messages:
    1,514
    Joined:
    Jun 1, 2016
    You and I must be reading different threads.
     
    ZeroBarrier likes this.