Showing results for 
Search instead for 
Did you mean: 

Drivers & Software

Adept I

Re: gcc segmentation faults on Ryzen / Linux

I had a great experience doing an RMA with AMD to replace a faulty Ryzen 5 1600.

Old CPU:

SN: 9R67974070167


When attempting to compile 12 instances of GCC using all twelve cores, a segmentation fault occurred on at least one build before finishing.

When left running, my computer would randomly reboot or crash every 2 to 6 days.

New CPU:

SN: 9GY9258U70272


When attempting to compile 12 instances of GCC using all twelve cores, it successfully compiled all 12 instances for 3 loops, for a total of about 12 hours of compiling.

When left running, my computer was up for a complete 10 days before I manually restarted it.

I have no complaints!


Re: gcc segmentation faults on Ryzen / Linux

@drcouzelis What motherboard are you using?

Adept I

Re: gcc segmentation faults on Ryzen / Linux

I have the MSI B350 Tomahawk, with CORSAIR Vengeance LPX 16GB (2 x 8GB, CMK16GX4M2B3200C16R).

Adept II

Re: gcc segmentation faults on Ryzen / Linux

As I mentioned earlier, my new CPU (UA37SUS) still exhibits random reboots about once a day unless I disable C-State Control, which of course I did. CPU idle temperature seems unchanged, so no apparent difference in regular idle power use. At least the compiler segfaults are gone, so I have an apparently reliable CPU now. I'll only touch anything when I see problems again.

Maybe my board contributes to the reboot bug? Asus Prime B350-Plus, UEFI 0902 = latest as of now.

Maybe running RAM close to the stability limit (I tested with several tools and then upped the voltage slightly, it should be fine) contributes.

Who knows!

Journeyman III

Re: gcc segmentation faults on Ryzen / Linux

I am segfaulting. 1721PGT. Happens in less than a minute. CPU at stock clocks, RAM at default clocks. Increased SoC, VCore and VDDR do not help. Runs very cool. AX370 K7.

AMD's e-mail form is broken, and RMA form is down until Sunday.

My last AMD build was Barcelona...that was bad enough to leave trauma. And now this...

Adept II

Re: gcc segmentation faults on Ryzen / Linux

@true - lol... there should be more copies of that claim form floating around than the latest copy of Game of Thrones episode.

Warranty Claims form.xlsx - Google Drive

Here is the form.  They are going to want a copy of your serial number.  That was hard, it's quite small and my computer came assembled, so no box.

I'm still waiting for AMD to ship out my replacement.  No inventory of good 1800X processors.

Adept III

Re: gcc segmentation faults on Ryzen / Linux


I also have freezes on idle, which do not show up in Windows. Only in Linux. After several tests, I believe I made my system freeze free by disabling C6 state with . I have a 1700X cpu, running now at stock until I am 100% sure that my system no longer freezes on idle. My motherboard is an ASUS X370 Prime with latest BIOS. With default settings (c6 state enabled), one core can go up to 3850-3900 with turbo + xfr, while with c6 state disabled in can only go up to 3500, same freq that all-core turbo can achieve. WIth c6 enabled, my it87 voltage sensor shows values down to 0,4V while with c6 disabeld, low voltage on idle never goes lower than 0,83 . In Windows I mostly use Ryzen Balanced power plan, which also disables core parking (c6 core), but I don't have freezes even when using the standard Balanced power plan that does core parking.

My conclusion is that something in my system (cpu-ram-motherboard-psu) cannot handle well the very low voltages on idle when c6 states are enabled. It could be the CPU but I really don't care RMAing again since disabling c6 core parking is something I want and btw many thanks to the author of that made this available.

Adept II

Re: gcc segmentation faults on Ryzen / Linux

About a week after I finally rx'd my replacement, (AMD received my return two weeks before, verified by fedex), I received a message saying that they had not received my return and were upset with me. The links in that message were broken so I couldn't directly give them a wtf message. I hope this is safe to ignore. =|

Re: gcc segmentation faults on Ryzen / Linux


I had the same case!

AMD confirmed they received my old chip and will proceed with replacement.

Please note original RMA number was changed in the meantime(!!!) but I already sent CPU with old RMA number on it.

Then week or so later email "AMD RMA# xxxxxxxxxxxx 15-Day Reminder"

This is a courtesy notice from AMD Global Customer Care with regards to

  Warranty Request RMA# xxxxxxxxxx. for your AMD retail packaged

  Processor(s) in a Box.

  Product         Serial#         Result


  xxxxxxxxxxx  xxxxxxxxxxxxxxxx   RETURN NOT RECEIVED

  Your warranty replacement processor was sent to you as part of the AMD

  Advanced Warranty Fulfillment service which is offered as a privilege

  to our warranty service, if we do not receive the original processor

  within the next 15 business days, we may not be able to continue to

  extend this privilege to you in the future.

  Next Steps

  If you have already returned the aforementioned product and feel you

  have received this notification in error please check your transaction

  history for recent updates or contact us via our services portal.

  Please provide your warranty Request’s RMA# for reference.

Which is not true - I got confirmation in ticket conversations. Links don't work of course

I replied to old ticket and told them to fix it - but just got info ticket was already closed.

I ignore this - I have both email confirmation they get my chip, DHL confirmation says package was picked by AMD and additionally there was info about RMA number in it.

Journeyman III

Re: gcc segmentation faults on Ryzen / Linux

I am wondering if anybody has experimented about "MultiCore Enhancement (MCE)" stuff in relation to the Ryzen problems on Linux platform.

Please check the link below or do google search for more info:

MultiCore Enhancement: The Debate About Free MHz

I did not get my hand on Ryzen 7, but I feel like there might be some relationship between MCE and Ryzen problems.

Especially, as far as I read, once you choose XMP profile then you loose the speed control over cores, which may cause some

synchronization issues. Can anybody enlighten us?