Overclock.net › Forums › Industry News › Hardware News › [Phoronix] Segmentation Faults On Zen CPUs Under Heavy Workloads
New Posts  All Forums:Forum Nav:

[Phoronix] Segmentation Faults On Zen CPUs Under Heavy Workloads - Page 3

post #21 of 402
There is a hyperthreading bug in Skylake and Kabylake processors that causes crashes and potential corruption/data loss. Intel fixed the microcode in April and distributed it to board makers. HOWEVER...


AsRock, for instance, hasn't bothered to update the BIOS of the "Fatality" ITX Skylake board I got for a colleague's VR machine. Some in this forum have tossed around a variety of fallacies to attempt to defend this unacceptably low level of support. Apparently, BIOS patches to fix severe bugs are no longer a requirement for basic support in the eyes of many here. Just throw your Skylake rig into the trash and buy whatever a company released in the last week or two.


Like this Ryzen bug, the Skylake/Kabylake bug appears under heavy loads.


There are kludges to work around the bug in Linux and Windows reportedly but proper BIOS patches are necessary for a basic level of support. People should not have to be locked out of other operating systems just because vendor like AsRock don't understand what proper support is.

Here's a list of the fallacies I remember:


1) I'm immature in the way I'm handling the problem. (Blame me, not AsRock and/or other board makers' lack of adequate support.)

2) I'm ineffective in the way I'm handling the problem. (Blame me, not AsRock and/or other board makers' lack of adequate support.)

3) The bug can't be discussed in AMD-related subforums because it's not relevant. (It was brought up in the context of someone considering buying an AsRock board, a topic that, obviously, involves the issue of the brand's support quality.)

4) It's only one board so it doesn't matter. (Wrong and wrong — 1: It's not just one board. 2: It still matters if it's one board.)

5) It's adequate policy to only update the very latest boards, maybe not ever patching the rest. (Wrong. This person tried to compare Skylake with Sandy Bridge, in 2017, too.)

6) It doesn't interest me because I am only interested in AMD stuff. (As if this guy is the entire forum, or is the only person who reads and participates in that topic.)

7) Other board makers do the same thing. (Good tu quoque—pot/kettle fallacy.)

8) The patch might trickle down someday, after the company has updated the higher-end boards. (The patch was released by Intel to board makers in April. Not good enough.)

9) Using word-of-mouth is wrong. (Yeah, right. Reputation via word-of-mouth is huge in business. Always has been. Always will be.)

10) Complaining about inadequate support and warning others about it is "slandering" a business. (Blame me. It's my fault the BIOS hasn't been patched.)

11) If you annoy me I'll leave. (Literally tried to threaten to quit the forum because I criticized AsRock.)

12) "We" aren't interested. (Gotta love the royal we in an open Internet forum.)


I probably forgot some of them.


So, it's interesting to hear about a potential hardware bug in Zen that may require a BIOS patch. I wonder which fallacies are going to be tossed around this time?
post #22 of 402
Quote:
Originally Posted by superstition222 View Post

I probably forgot some of them.

"Moral equivalence" is one in particular that stands out.
post #23 of 402
I use GCC in windows without issue.
Upgrayedd
(8 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen 1700 @ 3.95GHz 1.387v Asus Prime B350 Plus  RX480 XFX RS 1350/2200 Corsair Vengeance LPX 16GB 2933Mhz cas 14 
Hard DriveOSMonitorPower
Samsung 950 Pro M.2 Windows 10 pro x64 Viotek curved FHD 144Hz EVGA 750w 80 Bronze 
  hide details  
Reply
Upgrayedd
(8 items)
 
  
CPUMotherboardGraphicsRAM
Ryzen 1700 @ 3.95GHz 1.387v Asus Prime B350 Plus  RX480 XFX RS 1350/2200 Corsair Vengeance LPX 16GB 2933Mhz cas 14 
Hard DriveOSMonitorPower
Samsung 950 Pro M.2 Windows 10 pro x64 Viotek curved FHD 144Hz EVGA 750w 80 Bronze 
  hide details  
Reply
post #24 of 402
Quote:
Originally Posted by zGunBLADEz View Post

thats interesting thanks for the share

@gupsterg
you may look at this

Remembered something. It was either the same workload or another workload that causes errors. The guy said the Ryzen is recognized as a NUMA device and if one thread from a complex throws an inter-process interrupt to a thread working on another complex it'll stall forever. Or something like that. Not familiar with HPC.
New and Shiny
(18 items)
 
  
CPUMotherboardGraphicsRAM
Rynze 7 1700 ASUS ROG Crosshair VI Hero Sapphire RX Vega 56 G.SKILL TridentZ F4-3200C14D 
Hard DriveHard DriveHard DriveHard Drive
Kingston HyperX 3K Crucial MX300 Western Digital Black Western Digital Green 
Hard DriveCoolingOSMonitor
Western Digital Red Noctua NH-U14S Windows 10 Pro Dell U2414H 
MonitorKeyboardPowerCase
Dell P2414H Ducky One Corsair RM650x NZXT H440 White 
MouseAudio
Logitech G502 Proteus Spectrum Schiit Modi2U & Magni2U 
  hide details  
Reply
New and Shiny
(18 items)
 
  
CPUMotherboardGraphicsRAM
Rynze 7 1700 ASUS ROG Crosshair VI Hero Sapphire RX Vega 56 G.SKILL TridentZ F4-3200C14D 
Hard DriveHard DriveHard DriveHard Drive
Kingston HyperX 3K Crucial MX300 Western Digital Black Western Digital Green 
Hard DriveCoolingOSMonitor
Western Digital Red Noctua NH-U14S Windows 10 Pro Dell U2414H 
MonitorKeyboardPowerCase
Dell P2414H Ducky One Corsair RM650x NZXT H440 White 
MouseAudio
Logitech G502 Proteus Spectrum Schiit Modi2U & Magni2U 
  hide details  
Reply
post #25 of 402
Quote:
Originally Posted by stahlhart View Post

"Moral equivalence" is one in particular that stands out.
Corporations have no morals. They exist to get as much money as possible for as little product/service as possible. However, consumers do have morals. So, there is an inherent conflict between the amorality of the corporation and the morality of the consumer.
post #26 of 402
Does anyone know the bug happen with CPLEX/Gurobi or any of the DL frameworks?
post #27 of 402
If you look around a bit some people claim this could be a bug with old versions of GCC. Some Linux distros don't always have the most up-to-date repositories and you get old GCC versions.
post #28 of 402
Quote:
Originally Posted by superstition222 View Post

Corporations have no morals. They exist to get as much money as possible for as little product/service as possible. However, consumers do have morals. So, there is an inherent conflict between the amorality of the corporation and the morality of the consumer.

I'm not talking about the corporations.
post #29 of 402
ohh look, another issue that haters can use against Ryzen.

though on a side note, is it even a critical bug? from what i could understand it only happens on very specific workloads.
i don't think this is something that most of the common users should be worried about.
post #30 of 402
Same can be said for Intel haters.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [Phoronix] Segmentation Faults On Zen CPUs Under Heavy Workloads