You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!




Page 1 of 2 12 LastLast
Results 1 to 10 of 11

Thread: AMD Ryzen is.... buggy?

  1. #1
    Member
    Join Date
    Oct 2013
    Posts
    37
    Thumbs Up
    Received: 27
    Given: 74

    AMD Ryzen is.... buggy?

    Recently, there have been reports of Ryzen instability. AMD is yet to answer. One forum thread where this discussion is going on in more detail is linked to below:

    Gentoo Forums :: View topic - Segfaults during compilation on AMD Ryzen.

    Many users reporting issues and they are yet to find a way to making their system stable. The issue is only manifesting itself in highly threaded applications, most often in multithreaded GCC compilation tasks.

    These instability problems could be a bug on the CPU or anywhere else on the platform (even voltage regulators could be to blame). Maybe it could even be a compiler bug. However, a platform is only as strong as its weakest link so wherevener this problem is, it may turn into a "ryzengate" story.

    AMD should quickly jump in and be transparent about it, helping debug the issue so that these rumors can - hopefully - be put to rest. But as of now, this story is spreading fast and if it turns to be a bug in Ryzen it could be a huge blow to Ryzen and maybe even EPYC.

    2 Not allowed!
     

  2. #2
    Admin Daniel Nenni's Avatar
    Join Date
    Aug 2010
    Location
    Silicon Valley
    Posts
    4,035
    Thumbs Up
    Received: 721
    Given: 2,044
    I have heard of similar issues from a friend who works for a high performance gaming computer systems company. AMD is actually swapping out chips as well as updating BIOS and system level drivers. So yes this is a big problem. I have not spoken to him in a couple of weeks, let me call and get an update.

    1 Not allowed!
    Now available in print or Kindle: "Mobile Unleashed: The Origin and Evolution of ARM Processors In Our Devices"

  3. #3
    Influencer
    Join Date
    Feb 2016
    Posts
    75
    Thumbs Up
    Received: 46
    Given: 42
    AMD ZEN is entirely new architecture. And simmilary to all new x86 architectures, it has some problems which needs time to resolve. AMD is working on it, but since there are lots of different Linux distributions, it will take time to fix them all. (priority is Windows) If you have some trouble, try update you kernel (I heard that updated Ubuntu distribution is working well so probably you need to update it in your distribution manually), update your development toolchain (since there was problems in parralel compilation), etc.

    In gaming systems there is different problem. AMD ZEN has new cache structure. 4 cores are tied together with L3 cache. It is major problem since Intel has small part of L3 attached to every core comunicating via internal ringbus. So more software optimisations will be necessary.

    But it is improwing with every update. Actually memory subsystem seems to be better than what Intel curently has... We will see in future.

    1 Not allowed!
     

  4. #4
    Top Influencer
    Join Date
    Aug 2014
    Posts
    375
    Thumbs Up
    Received: 161
    Given: 91
    How is it that no one else has problems compiling with gcc? For instance, this site has run benchmarks and had no problems at all: AMD Ryzen 7 1800X Linux Benchmarks Review - Phoronix
    Maybe, it's just a couple of guys who happen to have a rig with a hardware fault.

    1 Not allowed!
     

  5. #5

  6. #6
    Member
    Join Date
    Oct 2013
    Posts
    37
    Thumbs Up
    Received: 27
    Given: 74
    Quote Originally Posted by lefty View Post
    How is it that no one else has problems compiling with gcc? For instance, this site has run benchmarks and had no problems at all: AMD Ryzen 7 1800X Linux Benchmarks Review - Phoronix
    Maybe, it's just a couple of guys who happen to have a rig with a hardware fault.
    @lefty It is not happening only while compiling, and it happens on multiple Linux distros including CentOS, Ubuntu, Fedora and Arch. Also, users often report they need to run the compiling job for multiple hours to reproduce the issue and a benchmark run does not take that long.

    AMD is aware of the problem and the engineering team has stated they are working on it. A few days later they suggested people disable SMT and op-code cache (it does not fix the problem). Users are left wondering if that was the final "solution" and whether or not AMD is still working on it.

    AMD's support staff also told people to not open new tickets about this issue, they have enough of them already.

    https://community.amd.com/message/2796982

    1 Not allowed!
     

  7. #7
    Top Influencer
    Join Date
    Dec 2013
    Location
    EU
    Posts
    368
    Thumbs Up
    Received: 142
    Given: 36
    It could be software (linux kernel issue), memory or PSU related as well.
    Here is a similar case from Intel (although very old):
    Linux Machine Check Exception: Is it the CPU? |Intel Communities

    Here is another one about a kaby lake laptop:
    hardware - Understanding Machine Check Exceptions (MCE) - Unix & Linux Stack Exchange

    More about MCE here

    In summary, problems happen (and CPU RMAs are not so uncommon unfortunately)

    1 Not allowed!
     

  8. #8
    Member
    Join Date
    Oct 2013
    Posts
    37
    Thumbs Up
    Received: 27
    Given: 74
    @astilo While I understand what you are saying and agree with you, the number of users experiencing this same issue is too large to simply dismiss it as an occasional defect.

    Anyhow, preliminary conclusion is that it's indeed a bug in the Ryzen processor. The DragonFlyBSD devs had already faced an issue with Ryzen and had a workaround in their code already:
    gitweb.dragonflybsd.org Git - dragonfly.git/commitdiff

    Based on their findings, some Linux experts suggested disabling Address Space Layout Randomization as a workaround and initial reports from users who have tried it are encouraging, everyone that tried it so far (still very early abd very few users) have reported that the issue is gone for now.

    So, it looks that indeed it is a Ryzen bug, although it may be a bug easily fixable via microcode update.

    Let's wait and see how AMD will respond next.

    1 Not allowed!
    Last edited by mbello; 06-04-2017 at 03:21 PM. Reason: Typos
     

  9. #9
    Top Influencer
    Join Date
    Dec 2013
    Location
    EU
    Posts
    368
    Thumbs Up
    Received: 142
    Given: 36
    Quote Originally Posted by mbello View Post
    So, it looks that indeed it is a Ryzen bug, although it may be a bug easily fixable via microcode update.
    Thanks Mbello, very interesting. A microcode issue is definitely more plausible, considering Ryzen is a brand new architecture. Should not be a big surprise anyway, if the optimization focus is at the beginning shifted more towards Windows OS.

    1 Not allowed!
     

  10. #10
    Admin Daniel Nenni's Avatar
    Join Date
    Aug 2010
    Location
    Silicon Valley
    Posts
    4,035
    Thumbs Up
    Received: 721
    Given: 2,044
    Quote Originally Posted by astilo View Post
    Thanks Mbello, very interesting. A microcode issue is definitely more plausible, considering Ryzen is a brand new architecture. Should not be a big surprise anyway, if the optimization focus is at the beginning shifted more towards Windows OS.
    My AMD contact feels that these types of problems are normal for a new chip and also says that a micro code fix is coming.

    Has there been any news on Ryzen running Windows 10? I would have expected to hear more complaints from the crowd by now. I read one a while back about Windows 10 performance on Ryzen but nothing lately.

    1 Not allowed!
    Now available in print or Kindle: "Mobile Unleashed: The Origin and Evolution of ARM Processors In Our Devices"

Page 1 of 2 12 LastLast

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •