Some Ryzen Linux Users Are Facing Issues With Heavy Compilation Loads

Discussion in 'AMD Processors' started by juanrga, Jun 8, 2017.

  1. mvmiller12

    mvmiller12 Limp Gawd

    Messages:
    173
    Joined:
    Aug 7, 2011
    I have an Asus Prime B350+ mainboard, and I chose it because it was the ONLY Asus Ryzen mainboard with regular PCI slots on it. I have my R7 1700 OC'd to 3.8GHz and RAM @ 2933. System is 8 hrs Prime95 w/AVX stable. I stopped it after 8 hours, because IMO 8 hours in testing is good enough. I am admittedly liquid cooling the chipset and VRMs on this board though, and it is running Windows. I realize this is a Linux thread, but I've had no problems at all with my B350 chipset mainboard and overclocking an 8 core CPU.

    And this system just finished a 4 solid day video encoding marathon with the CPU pegged pretty much the entire time. I decided to re-rip and re-encode a bunch of my BluRays at a higher bitrate now that I can see banding in some of them on my new 4K TV (Movies are stored on my home NAS so I can keep my disks safely locked up. Children...).
     
    ManofGod and Gideon like this.
  2. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    4,738
    Joined:
    Jul 29, 2009
    But you have a clue , other people read this and could think: hey it is do able, they skip over the part where you explain you have the VRM and chipset liquid cooled (would be nice if you could share a picture of it on this forum somewhere :) ) .

    But without the cooling solution you created you would not ever attempt to run 8 core under load for 8 hours on B350.
     
    Gideon likes this.
  3. juanrga

    juanrga Limp Gawd

    Messages:
    440
    Joined:
    Feb 22, 2017
    Funny considering that your only post in this thread doesn't add anything of value.
     
    GoldenTiger and Shintai like this.
  4. juanrga

    juanrga Limp Gawd

    Messages:
    440
    Joined:
    Feb 22, 2017
    Did you read the links given including the one about the patch is being proposed to BSD kernel to alleviate the bug (i.e. the link that starts with "There is a bug in Ryzen related to the...")? Or will you continue in denial mode forever. A sincere answer would be clarify if i must add your subsequent posts to ignote.
     
  5. ManofGod

    ManofGod [H]ardness Supreme

    Messages:
    7,415
    Joined:
    Oct 4, 2007
    I read your Ryzen Predictions page and let me say, it does not surprise me that you would respond like this. The entire page was nothing more than, I am right, they are wrong, see, see, SEE! Therefore, feel free to ignore if you prefer but, that will not change who you are. By the way, Ryzen has kicked ass, just the way it is. :)
     
  6. JustReason

    JustReason [H]ard|Gawd

    Messages:
    1,780
    Joined:
    Oct 31, 2015
    yet by Feb we all knew what was coming, so not that special.

    OT seems if there were an inherent CPU issue it would affect all users. And being the few that do and 100%(required) load leads more to software scheduling or instabilities rather than an errata or similar issue.
     
  7. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    4,738
    Joined:
    Jul 29, 2009
    Which means absolutely nothing , there so many proposals for these kind of things unless a kernel developer can replicate the problem you can make proposals until you are blue in the face.

    Maybe you need to learn to quote but then again that is already asking to much of you. The only person that is in denial are you, since there are some people that do not have this bug. You linked to a reply by someone with a B350 based machine running 8 cores and having problems which I linked you the video which shows the B350 platform can't do this without problems of a hardware nature. upon this post https://hardforum.com/threads/some-...pilation-loads.1936605/page-3#post-1043060783

    Shows you that you can run it with 8 cores but only if you apply cooling solution as mvmiller12 did.
     
  8. Gideon

    Gideon Gawd

    Messages:
    761
    Joined:
    Apr 13, 2006
    We already know he is wrong cause if AMD had said there was a bug and they are fixing it he would have posted it all over. Also errata tends to be easily duplicated since everyone has the issue.
     
    JustReason and Pieter3dnow like this.
  9. mvmiller12

    mvmiller12 Limp Gawd

    Messages:
    173
    Joined:
    Aug 7, 2011
    I have a picture of the inside of the machine, but it is using soft tubing, so it looks like quite the mess of computer intestines :). Also, the pinched tubing in the picture has been repaired. When money presents itself, I am seriously considering re-doing the radiator, pump/reservoir, flow meter portions of the loops (there are 2) with hard tubing, and leaving the component pieces themselves on soft tubing with the quick disconnects - the best of both worlds for neatness and accessibility. There is currently a LOT of excess tubing in there to prevent kinks.

    The VRMs can be seen in the picture, but not very well. The chipset cooler is also largely obscured by the video card.I ended up getting a little creative with the VRM block mounting because the VRM area on the right side of the board is shorter than the actual block. I had to use the smallest Koolance VRM block plate and mount the cooler using a mainboard support hole. I put a flat-topped plastic post in that location in the case to still provide support.Works a treat. I also modded a Koolance 390 AM4 adapter kit to work on the 380 block. The mounting holes for the block are on the adapter, but they needed to be drilled through to screw on to the 380 block - apparently the 390 no longer has it's brackets screwed onto the block.

    WP_20170316_15_54_55_Pro.jpg
     
  10. JustReason

    JustReason [H]ard|Gawd

    Messages:
    1,780
    Joined:
    Oct 31, 2015
    doesn't matter what it looks like as long as it works.
     
    BitMaster, cageymaru and Algrim like this.
  11. juanrga

    juanrga Limp Gawd

    Messages:
    440
    Joined:
    Feb 22, 2017
    Precisely the guy that identified the bug on RyZen is a kernel developer...

    Wait! Didn't you read the several post here or the links given with people on X370 boards is also experiencing the problem?
     
    Shintai likes this.
  12. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    4,738
    Joined:
    Jul 29, 2009
    If you only had some basic knowledge about hardware. Check out the picture from mvmiller12 , then wonder why I made the comments about the B350 board having trouble with errors if used without such a solution. Then wonder why are faults appearing because I'm not saying that there are no errors , just saying that the cause of the errors might not be Ryzen in the first place.

    There so many posts in the community thread some claiming this helped other saying it did not work for them. You pick which is which and of those are truly having problems with Ryzen or just hardware/bios problems ?
     
  13. juanrga

    juanrga Limp Gawd

    Messages:
    440
    Joined:
    Feb 22, 2017
    Attacking people that post information you don't want to hear doesn't add value. And what is more important, the problems that people is experiencing will not go away.

    Then what part of "There is a bug in Ryzen" from the BSD patch you don't get?
     
    Last edited by a moderator: Jun 16, 2017
    Shintai and gigatexal like this.
  14. gigatexal

    gigatexal [H]ardness Supreme

    Messages:
    6,946
    Joined:
    Jun 22, 2004
    +1 this.
     
    Shintai and juanrga like this.
  15. Findecanor

    Findecanor [H]Lite

    Messages:
    75
    Joined:
    Feb 21, 2016
    Please, guys. Be civil!
    Bickering between each other this way does not add to the discussion. If you can't post anything constructive, then don't do it. You are wasting mine and other's time because you are making us read your posts. And I can't even tell from half of the posts which side the poster is even on.
     
    Last edited: Jun 17, 2017
  16. ManofGod

    ManofGod [H]ardness Supreme

    Messages:
    7,415
    Joined:
    Oct 4, 2007
    B350's with R5 1600 or R7 1700's will do just fine. Overclocking probably will not but, then again, if you know what you are doing, you will buy the right thing to do so.
     
  17. JustReason

    JustReason [H]ard|Gawd

    Messages:
    1,780
    Joined:
    Oct 31, 2015
    I think the point is very specific to this situation as far as b350 is concerned. The load is extreme and why some think it is stability related be it OCs/voltages or software scheduling. I am more inclined toward software as some are stock clocked.
     
  18. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    4,738
    Joined:
    Jul 29, 2009
    What part of linking threads and posting your own opinion on top of it don't you get ?
    You post about things you don't understand. Case in point your 1st post.

    You claim that AMD is saying to disable SMT that would fix it according to what you wrote here. I haven't seen in any thread that would work for everyone , what you are doing here is creating a situation where you just say something suggest it comes from AMD and suggest it works, you did this not AMD.

    You acknowledge it here yourself:

    You interject your own opinion while there is no official answer from AMD.

    I'm saying you are purposely trolling this. You have been from day 1 on this board. You contradict yourself you don't post any relevant information.

    I have yet to see 1 post from you on this subject with a clear solution to this problem. Or a clear outline on which users have this problem. You move goal posts and claim to be a victim of personal attacks , then post something relevant to the issue.

    You can not even discern what is impossible in hardware to run a full load 1800x on any current B350 board and not get errors (Linux or windows for that matter) .
     
    JustReason likes this.
  19. juanrga

    juanrga Limp Gawd

    Messages:
    440
    Joined:
    Feb 22, 2017
    Do you even read what you quote? You are quoting me saying "AMD is recommending to disable SMT" and "their last recommendation was to disable uop or SMT". I didn't say it would work for everyone.

    From the amdcommunity link

    Disabling SMT only helps some users to reduce the frequency of repetition of the problem

    The problem is traced to a bug on RyZen that was identified time ago by a kernel developer. A temporary patch is being added to Dragonfly BSD


    AMD remains silent since the last post the day 1 June in the amdcommunity thread. They continue "investigating the issue". But I got unofficial word they are preparing a microcode update.

    It was also made clear in this thread that people is having the same problem on X370 mobos: MSI X370 Gaming Pro Carbon, ASRock Taichi x370, MSI X370 SLI Plus, ASUS PRIME X370-pro,...

    Could you and others stop pretending it is a problem with B350 mobos?
     
    Last edited: Jun 17, 2017
  20. JustReason

    JustReason [H]ard|Gawd

    Messages:
    1,780
    Joined:
    Oct 31, 2015
    Pretending is what your doing in regards to the subject. Look more closely at the issue and the reports. It looks a lot more like a software issue ie: kernel issue with the ryzen CPU. And sorry one person giving advice is not the same as an Official statement. You are drawing far too many conclusions only to serve your goals in this witchhunt.
     
    Pieter3dnow likes this.
  21. Shintai

    Shintai 2[H]4U

    Messages:
    4,066
    Joined:
    Jul 1, 2016
    So software doesn't run correctly on Zen you say? There is something called compatibility for a reason.
     
    juanrga likes this.
  22. JustReason

    JustReason [H]ard|Gawd

    Messages:
    1,780
    Joined:
    Oct 31, 2015
    So you agree with me then?
     
  23. Shintai

    Shintai 2[H]4U

    Messages:
    4,066
    Joined:
    Jul 1, 2016
    If you agree to that...

    So what else is Zen incompatible with if it doesn't follow x86 compatibility? :)
     
    juanrga likes this.
  24. gigaxtreme1

    gigaxtreme1 2[H]4U

    Messages:
    2,688
    Joined:
    Oct 1, 2002
    Looks like no-op code or wait state has to be inserted for dependent threads.
     
    Last edited: Jun 17, 2017
    Shintai and juanrga like this.
  25. juanrga

    juanrga Limp Gawd

    Messages:
    440
    Joined:
    Feb 22, 2017
    That person is working for AMD and responding to users about this problem in the name of AMD. Sure yuo have noticed the string "AMD" in his nickname. What he said is an official statement until someone else at AMD above him says otherwise.
     
    Shintai likes this.
  26. JustReason

    JustReason [H]ard|Gawd

    Messages:
    1,780
    Joined:
    Oct 31, 2015
    Wrong. He can make any number of statements none of which are official for AMD as a company. When AMD makes an official statement they will do it through official channels. I can go online and make any number of statements concerning products from the company I work for and trust me none of them would be official. Try making a legal case based on such statements and see how "official" they are received.
     
  27. juanrga

    juanrga Limp Gawd

    Messages:
    440
    Joined:
    Feb 22, 2017
    If you are working in the sales dept. of AMD and then posting during the week-end your personal statements about technical issues in third party forums external to your company, your posts don't have official validity. But this is not the case is being noticed. Matt is the person choosen by AMD to reply tickets about technical matters in the official support channel created by AMD. He is talking in representation of AMD, under the title of "Technical Support Engineer", to solve the tickets opened by users of products of AMD. That is why the talks in plural such as "we're looking at the issue", "We're still investigating the issue", "There is no need to open new tickets on this issue. We are investigating"... and that is why he creates documents with official instructions, for instance about how to enable stuff in AMD hardware

    https://community.amd.com/docs/DOC-1432
     
  28. JustReason

    JustReason [H]ard|Gawd

    Messages:
    1,780
    Joined:
    Oct 31, 2015
    Still not official just because he states it. I know of the guy and have had plenty of conversations. At best he has insider knowledge but in no way is he responsible for making official statements. He can ferry them if need be but forums are not and have never been the correct medium for official statements.
     
  29. mvmiller12

    mvmiller12 Limp Gawd

    Messages:
    173
    Joined:
    Aug 7, 2011
    Processors have erratum, whether they be designed by Intel or AMD. If it is fixable via microcode update, I'm sure they will. If ti requires a new spin of the processor, they'll do that, and you'll either buy the new revision or live with/work around the problem (it is exceedingly rare for CPUs to get replaced for erratum. In fact I've only ever heard of it happening with the ancient Pentium FDIV bug, and then only on a case-by-case basis)


    The point is that Intel and AMD are equally x86 compatible, but shit happens sometimes - to both.
     
  30. Shintai

    Shintai 2[H]4U

    Messages:
    4,066
    Joined:
    Jul 1, 2016
    Agree, people just need to accept it instead of pretending nothing is wrong and its someone else fault.
     
    juanrga likes this.
  31. juanrga

    juanrga Limp Gawd

    Messages:
    440
    Joined:
    Feb 22, 2017
    It seems a new spin of Zen silicon is being prepared. This B2-step seems to focus only on Uncore/SoC erratas (PCIe controllers, etc.).
     
  32. JustReason

    JustReason [H]ard|Gawd

    Messages:
    1,780
    Joined:
    Oct 31, 2015
    LOOK... no links or proof. Got any?
     
  33. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    4,738
    Joined:
    Jul 29, 2009
    Reading what people say is not required I guess ... You are also ignoring what I wrote about the B350 board and 8 cores under load , maybe you have insight regarding that as well since you do know everything. Hows your piggy bank going these days Shintai ?
     
  34. Pieter3dnow

    Pieter3dnow [H]ardness Supreme

    Messages:
    4,738
    Joined:
    Jul 29, 2009
    So you finally posted the official message from AMD it took you almost 100 posts in the thread to do so. Someone might ask why is it that you will post stuff without any validation behind it?

    Since you posted this let me post what I found on the AMD community link:

    And follows up with:
    and he adds:

    the person also knows how to rule out certain aspects of how hardware might be at fault and posts:
    Now you know when people say things that are obvious can only be ruled out by people posting correct information. Your posts lack any valid information from sources which makes it hard to read.
    Throughout the thread you can read that SMT disabling does not help and the micro op cache can only be disabled on Asus motherboards. Check any of your posts before https://hardforum.com/threads/some-...pilation-loads.1936605/page-3#post-1043062721

    Still waiting for you to post some things that can actually help people with Linux ...
     
  35. gigaxtreme1

    gigaxtreme1 2[H]4U

    Messages:
    2,688
    Joined:
    Oct 1, 2002
    AMD Readies B2 Stepping of the Ryzen "Summit Ridge" Silicon
    by btarunr Monday, June 19th 2017 06:55 Discuss (28 Comments)
    AMD is readying a new stepping of its 14 nm "Summit Ridge" eight-core CPU silicon, which powers its socket AM4 Ryzen processors, according to Canard PC. The new B2 stepping reportedly addresses a lot of hardware-level errata which cannot be fixed merely by AGESA updates. According to Canard PC, the changes seem to be focused on the uncore components of "Summit Ridge." Typically, uncore refers to the integrated northbridge, which includes components such as the memory controllers, PCI-Express root complex, etc.

    If the B2 stepping is mostly focused on uncore-level errata, it could mean improved PCI-Express device support, and perhaps even memory support improvements beyond even what AGESA 1.0.0.6 brings to the table. Canard PC reports that it hasn't come across any CPU core-specific errata being addressed with the B2 stepping. The glaring FMA3-related bug has been patched through BIOS updates, and most newer batches of socket AM4 motherboards come with the patch pre-installed

    Think this is Canard PC
     
    Shintai likes this.
  36. Gideon

    Gideon Gawd

    Messages:
    761
    Joined:
    Apr 13, 2006

    It's a bunch of nothing. It fixes uncore-level errata according to Canard but then he has no clue what that is? There may be in fact a B2 stepping coming but so far we have no idea why and this little blurb is garbage without any substance. Not surprising Juanrga regurgitated it on here, more surprised he was unwilling to say where he got it from.
     
    JustReason likes this.
  37. gigaxtreme1

    gigaxtreme1 2[H]4U

    Messages:
    2,688
    Joined:
    Oct 1, 2002
    Be nice to see a 1900X though.
     
  38. Shintai

    Shintai 2[H]4U

    Messages:
    4,066
    Joined:
    Jul 1, 2016
    Seems to indicate no performance changes. Just bug fixing that requires hardware changes.
     
    juanrga likes this.
  39. ecmaster76

    ecmaster76 Gawd

    Messages:
    828
    Joined:
    Feb 6, 2007
    If I had to guess a lot of the uncore fixes are probably for the server world (multi socket and MDM) where the SoC features of Zeppelin are put to use and the Infinity Fabric needs to be 110% stable for off die communication

    They crowdsourced their server validation :cool: ;):p
     
  40. nunosilva

    nunosilva n00bie

    Messages:
    6
    Joined:
    Aug 6, 2013
    linux folks are working overtime finding CPU bugs:
    https://lists.debian.org/debian-devel/2017/06/msg00308.html
    Already in hardocp.com too:
    https://www.hardocp.com/news/2017/0...ocessors_allegedly_have_broken_hyperthreading
     
Tags: