[WARNING] Intel Skylake/Kaby Lake processors: broken hyper-threading


Recommended Posts

  • 2 weeks later...
On 10/7/2017 at 6:27 AM, ars92 said:

Any idea if this affects coffee lake?


Sent from my iPhone using Tapatalk

 

Coffee Lake just launched on the 5th, so I highly doubt anyone has a built a system for unRAID based on that platform.

Its also going to be a while before the non Z based chipsets are out. Current wave of boards are targeted towards enthusiasts, gamers and overclockers mainly.

 

If you are planning a new build, you should wait for a board designed for workstation or server use. Also the first little while, BIOS's tend to have minor issues that could really be problematic for server use.  No way i would call day 1 launch BIOS stable for server use.

 

Sorry I dont have a proper answer to your question, but if you were thinking of a new build I would wait a few weeks at the very least to see what errata pops up with that CPU and wait for the inevitable work around.

 

IMHO, Coffee lake is a rushed to market product to stop the momentum Ryzen has.  Not saying it will be a bad product, as early results look good.  Just saying for server use, you don't want any surprises. 

 

Link to comment
 
Coffee Lake just launched on the 5th, so I highly doubt anyone has a built a system for unRAID based on that platform.
Its also going to be a while before the non Z based chipsets are out. Current wave of boards are targeted towards enthusiasts, gamers and overclockers mainly.
 
If you are planning a new build, you should wait for a board designed for workstation or server use. Also the first little while, BIOS's tend to have minor issues that could really be problematic for server use.  No way i would call day 1 launch BIOS stable for server use.
 
Sorry I dont have a proper answer to your question, but if you were thinking of a new build I would wait a few weeks at the very least to see what errata pops up with that CPU and wait for the inevitable work around.
 
IMHO, Coffee lake is a rushed to market product to stop the momentum Ryzen has.  Not saying it will be a bad product, as early results look good.  Just saying for server use, you don't want any surprises. 
 

Agreed. Thanks for your insights.

Sent from my LG-H990 using Tapatalk

Link to comment
  • 2 weeks later...

I also have an Asrock Rack board and am in the same situation. They said they were "working on it" months ago but no update.

 

I also have an issue where the IPMI in my board uses such an old encryption, the latest Java version blocks it as too insecure. So they are supposedly making an update for that too.

 

Not holding my breath.

Link to comment
  • 2 weeks later...

This reads quite disturbing and may impact me (i7 7800X). I do have some issues these days, but I never thought they may be related to the CPU.

 

At the same time I would anticipate that there are many Unraid users running Skylake/Kabylake CPUs. Is this issue for a real and a big deal? This thread is quite short and nobody for sure confirms a major issue.

 

The link to the debian thread is a bit cryptic to me. Can someone advice whether my CPU is impacted and can be fixed via update or otherwise?

Link to comment
  • 1 month later...
On 6/26/2017 at 9:30 PM, jonp said:

We are looking into providing a way to allow users to apply a microcode update as a workaround that allows you to temporarily patch out of this bug on a per-boot basis

 

Hey @jonp, are we getting any closer to being able to apply microcode updates at boot time?

 

Although I do not have *lake processor, my Haswell is listed as having microcode updates:
  https://downloadcenter.intel.com/download/27337/Linux-Processor-Microcode-Data-File
but it isn't clear when the updates came out or whether they are already in my June 2015 BIOS.  It also isn't clear what the updates do, but I don't expect that Intel would release them if they weren't important.

 

Because /lib/firmware is mounted readonly, it isn't possible for us to drop the latest microcode files into /lib/firmware/intel-ucode and test them with iucode_tool as described here:
  https://askubuntu.com/questions/545925/how-to-update-intel-microcode-properly 
Which means there's not a lot we can do without your help.

 

So... can we please get the latest Intel microcode files added to bzfirmware, have iucode_tool added to bzroot, and have the kernal compiled with the appropriate flags to early/late load the microcode?  That should be enough to let us test our individual systems and manually modify syslinux to load the updates. I'm not sure how much of that process you'll be able to automate, but at least those that want to could manually patch their own systems.

Link to comment
50 minutes ago, limetech said:
 
Can you tell if this fixed anything?

 


Supposedly, the microcode that fixes the hyperthreading issue is the 2017-07-07 microcode. There is now a 2017-11-17 release as well. I have a Skylake Xeon in my main server, but, I am running 6.3.5 on that server. I may try updating it to rc16b to see what microcode is applied.

My backup server with a Haswell i5 did get a microcode update to the 2017-01-27 version when I applied rc16b but these later revisions are supposed to apply to it as well according to the Intel web site.

 

EDIT: This post from Limetech in the rc16b thread explains what is going on.

 

1 hour ago, limetech said:

This file contains all the microcode for all Intel CPUs which can have their microcode updated.  When they say the latest microcode for a given CPU has a certain date, I believe they are referring to the release of the microcode file itself, and not necessarily that that CPU has an update.  Make sense?


2017-01-27 is the latest microcode available for the i5 4590, but, if I updated my Skylake Xeon server to rc16b it would likely get a microcode with a different date.

 

Limetech confirmed they downloaded the 2017-11-17 microcode file which is the latest.  Different microcode for different CPUs is contained in that file.
 

 

Edited by Hoopster
Link to comment
  • 3 months later...
On 3/18/2018 at 8:22 AM, HellDiverUK said:

I'm just about to go to an i7 in my unRAID box (i7-8700) - is this HT problem still a thing?

My guess is that this is not a concern now with many processors because of the recently-implemented early loading of microcode during unRAID boot.  My CPU had a microcode fix released last July.  It never made it into a BIOS update from MB manufacturer, but, the Intel microcode being loaded now by unRAID fixes it.

 

Even if Coffee Lake processors were released before the problem/fix were discovered, I assume they are also covered by recent microcode releases.  Since intel has been releasing microcode to address spectre/meltdown I assume those also have the hyperthreading fixes as well.

Edited by Hoopster
  • Like 1
Link to comment
On 3/26/2018 at 10:28 AM, John_M said:

Is the microcode update mechanism specific to Intel processors or does it work with AMD products too? Indeed, does AMD make microcode updates available to the public or are they only distributed via BIOS updates?

 

It does work with AMD too, but AMD typically does not release firmware in a form where it can be loaded by "early update".  Probably there is  technical reason for this and as a result you must wait for bios to be updated instead.

  • Like 1
Link to comment
  • 4 weeks later...

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.