It's a good day to be on an AMD kernel

Subject: General Tech | January 3, 2018 - 01:12 PM |
Tagged: security, patch, kernel, Intel

Intel is having a lousy day after revealing a fundamental flaw in their architecture design; one not shared by AMD chips.  It turns out that many common programs are able to read the protected memory sections of the chips kernel, something commonly known as a very bad thing.  The flaw exists in both Linux and Windows and is serious enough that a patch has been released, which you should install.

However initial reports show the patch has a negative effect on performance, with a worst case scenario showing quite a performance delta.  The thing to keep in mind is that we do not have many data points yet, more testing needs to be done to determine exactly how much performance degradation will be experienced.  We will conduct our own testing here, with a focus on storage which could see the most degradation, especially the newly released Intel Optane.  You can also expect that Amazon, Azure and other hosting companies will be releasing information on the effect this has on their systems, which will give us a good overall view of what this patch will do.

The easiest way to ensure you are not going to experience this issue is to pick up a Ryzen or Threadripper, of course.  The Inquirer offers more insight here.

View Full Size

"A fundamental design flaw in Intel's processor chips has forced a significant redesign of the Linux and Windows kernels to defang the chip-level security bug."

Here is some more Tech News from around the web:

Tech Talk

Source: The Register

January 3, 2018 | 01:19 PM - Posted by anonpcper (not verified)

one shared by AMD chips.

Missing a "not" there.

January 3, 2018 | 01:29 PM - Posted by Jeremy Hellstrom

Good catch, thanks.  That kind of ruins the story doesn't it?

January 3, 2018 | 01:34 PM - Posted by JohnGR

"The easiest way to ensure you are not going to experience this issue is to pick up a Ryzen or Threadripper, of course."

Probably Epyc suits better here. But what I would like to comment is that someone - we don't say names - probably influenced the developers to treat ALL x86 CPUs as having this bug.

It's funny really. All the time people promoting Intel try to convince others that AMD CPUs are incompatible and now some developers out there try to make AMD processors as much compatible as possibly, with Intel processors, in the... bug level.

January 3, 2018 | 01:41 PM - Posted by ButMonopoliesAreGood (not verified)

Intel BORKED INSIDE!

All those billions Intel wasted on Contra Revenue(Est $14 billion) could have been spent on fixing it's CPUs over these past few years that Intel Knew about these CPU inadequacies but Intel did nothing of the sort.

Now AMD does NOT have these problems on its Zen x86 CPUs but Intel's programmers are busily pushing out such linux patches that will force even AMD's Zen based CPUs to suffer also!

See where Monopolies get you in the end and now everyone with an Intel branded CPU gets some performance loss because Intel paid so much illegal kickbacks over the years to keep AMD's competition outside.

All the security issues and most everybody with Intel forced inside their PC/Laptops over the years are now having to pay the piper, and that's the big new year's slowdown complements of the big CPU monopoly.

January 3, 2018 | 01:41 PM - Posted by Rocky123 (not verified)

I think they said Mac Systems are affected as well you might want to add that if it's not already.

January 3, 2018 | 01:56 PM - Posted by ButMonopoliesAreGood (not verified)

All the more reason for Apple to go with Zen/Vega Raven Ridge inside their Macbooks and have no performance issues and better graphics.

We have all had Intel's dog food graphics foisted on us for so many many years now. I'd like to see a Macbook Air with Raven Ridge inside!

Just look at the Intel CEO get rid of all his Intel stock but the required Intel CEO minimum just a short while ago and of course Intel's CEO gets access to the golden ripcord to pull and unfurl that golden parachute to See You Later Suckers riches in some tropical paradise gated community.

January 3, 2018 | 02:04 PM - Posted by Dave Kelly (not verified)

Totally agree . Intel graphics are pure rubbish and it beggars belief why apple still put this crap into their laptops while Ryzen is available now .

January 3, 2018 | 03:28 PM - Posted by BLT (not verified)

ARM64 as well!

https://lwn.net/Articles/740393/

January 3, 2018 | 04:10 PM - Posted by Jeremy Hellstrom

Yes, which makes it unclear if this is actually going to include some AMD chips.  The story is still developing.

January 3, 2018 | 04:21 PM - Posted by mAxius

the amd patch is in and it has been reviewed and accepted https://lkml.org/lkml/2017/12/27/2 this proves intels statement https://newsroom.intel.com/news/intel-responds-to-security-research-find... is fud

January 3, 2018 | 06:45 PM - Posted by NotSoFastUntilTheChipsDesignersSpeak (not verified)

It's funny that Intel is even mentioning any other chip makers names in an attempt at deflecting and spinning the issue to involve any other CPU core maker in the issue.

AMD is only involved in seeing that their hardware does not get slowed down by any Linux/Windows commits that may be uncessary for AMD's brand of x86 ISA based products.

look at old Intel trying to dodge and obsfucate the issue by even attempting to speak for any other CPU maker!

Intel is trying to make it look like an industry wide security issue when it appears to be an Intel brand security issue. Intel wants the that artifically leveled playing field so Intel's mistake becomes everybodys' mistake. Classic Deflection of the Issue BY Intel!

January 3, 2018 | 06:34 PM - Posted by NotSoFastUntilTheChipsDesignersSpeak (not verified)

Not really its just one of those patch first to be safe and wait and see if ARM64 Really has the issue! So Kernel commits need to be taken with a big old grain of salt until the hardware maker can actually confirm that it's an issue on ARM64.

It should also be noted that ARM Holdibngs can only speak for their exact Refrence Designs and the customers who only use the ARM Holdings's exact and unmodified ARM Holdings designed Refrence cores.

So those top tier ARMv8A ISA architectural licensees will have to speak for their custom ARM designs as those designs like the Apple A series designs are fully custom CPU micro-archs that are engineered to run the ARMv8A ISA. So Apple's A Series cores are not made/designed by ARM Holdings as Apple only licenses the ARMv8A ISA from ARM holdings and NOT any of ARM Holdings' Refrence CPU designs!

And any of those ARMv8A ISA architectural licensees may have vastly different underlying hardware that is capable of executing the ARMv8A ISA in their own way and Nvidia's Denver cores are custom as well!

AMD has already stated that their x86 ISA based CPU core designs do not have the issue and ARM Holdings has yet to comment as well as the custom ARM CPU core makers who have their own custom CPU micro-archs that are engineered to run the ARMv8A ISA.

Any Commits issued may just be precautionary and the NDA on the issue still has not expired.

January 3, 2018 | 08:29 PM - Posted by BLT (not verified)

It is as official as it gets since the author of the patch is Will Deacon who is a kernel hacker at ARM Ltd.

January 3, 2018 | 09:42 PM - Posted by AndThisNewYearHasOnlyJustStartedInEarnest (not verified)

Only for ARM Holdings refrence design cores but there are plenty of custom micro-archs out there that are engineered to execute the ARMv8A ISA and those custom designs are not from ARM Holdings.

Will Deacon codes for ARM Ltd. so is that person related to ARM holdings/Soft Bank and was the fix just meant for temporary security until ARM Holdings' chip engineers can suss out any actual problems in its CPU designs. I'll wait for Arm Holdings official statment on the matter and those official statments of custom ARM core designers like Apple, Nvidia/Others. ARM Holdings can not speak for Apple or Nvidia/Others with custom micro-archs of their own that are engineered to excute the ARMv8A ISA.

Programmers can post preemptive fixes into commits even if there may not be any problems with the hardware in order to err on the side of safty/security until the higher ups release the official results.

January 3, 2018 | 02:06 PM - Posted by Axle Grease

All of us will be doing tests before and after the patch.

January 3, 2018 | 02:11 PM - Posted by Anonymous12313dd (not verified)

Honestly this better be optional install if really has any impact on performance

January 3, 2018 | 02:57 PM - Posted by mAxius

the impact of the design flaw is huge. imagine you are an optimized data center with intel cpus that uses "speculative execution" (last one that didn't use it was the og Pentium) your workload was getting done before the patch and after the patch you need 20-65% more resources to do the same workload. yeah you are going to be out for flesh and blood. we don't know the full proformence impacts till all the patches are out between now and friday of next week at least. Linux we can see everything that is happening in real time. For Windows and some other OSes we can not see whats going on behind the scenes. All Operating Systems that use x86 are going to need to patch this design flaw. So the next week and a half is going to be very interesting.

Note to not install this patch is willful stupidity you don't want anything that is not authorized reading kernel memory as it opens your system to a security black hole that you will never escape form.

January 3, 2018 | 03:33 PM - Posted by Randy (not verified)

Considering this patch has been in the fast ring of Windows for 1-2 months and there not being a performance outcry I would hold out to see actual results over speculative (pun intended) assumptions. One would imagine major data centers have already tested these patches to some extent already and as there is no talk of huge data center hardware orders going crazy I'm thinking there is the chance that things are being overblown.

That being said, it is absolute crap and even worse that nobody will hold Intel responsible for it.

January 3, 2018 | 03:58 PM - Posted by Moyenni (not verified)

If disk I/O is severely hit, does it make Optane useless on Intel procs?

January 3, 2018 | 04:16 PM - Posted by mAxius

ok two links read them both and you will be in the loop also don't be afraid to click the links in the storys

https://arstechnica.com/gadgets/2018/01/whats-behind-the-intel-design-fl...

https://www.theregister.co.uk/2018/01/02/intel_cpu_design_flaw/

January 3, 2018 | 04:27 PM - Posted by Vince (not verified)

Phoronix has some initial benchmarks showing the performance loss.

https://www.phoronix.com/scan.php?page=article&item=linux-415-x86pti&num=1

https://www.phoronix.com/scan.php?page=article&item=linux-more-x86pti&num=1

Interesting, games do not seem to affected.

https://www.phoronix.com/scan.php?page=news_item&px=x86-PTI-Initial-Gami...

January 3, 2018 | 04:51 PM - Posted by Vince (not verified)

Note: Phoronix tested Linux gaming performance. The jury is still out on Windows gaming performance. However, I would expect a performance hit on at least one title, due to the way the DRM is run. Assassin's Creed Origins for the PC runs the Denuvo DRM inside VMProtect (a virtual machine) instead of alongside the game.

January 3, 2018 | 04:49 PM - Posted by Drazen (not verified)

Tomorrow when they find some bug in AMD all above Intel haters will ... say nothing!
CPUs today are VERY VERY complicated and during development is not possible to think what might come out in future. Software could be fixed but hardware not (usually).
Today all related to viruses is meaningless. AV simply has too much to check and CPUs are on knees. I even doubt AV is checking very old viruses. And with new ones they first need to be detected.

January 3, 2018 | 05:07 PM - Posted by JohnGR

You are kidding right? Or just started reading the "How to Troll for dummies" guide. If this was an AMD bug, AMD would have been stabbed, burned, hang, butchered, buried by the tech press.
Anyway, don't worry. When people realize that this means more (fixed) CPU sales in the future, intel's share price will go back up.

January 3, 2018 | 05:25 PM - Posted by TheUglyx86UglyISAsoFullOfThatISABloat (not verified)

Intel CPUs from 10+ years have this bug and Intel with all that money to totally waste on Contra Revenue in a failed attempt at getting that power hungry CISC x86 ISA shoehorned down into the low power using domain of the RISC ISA based mobile designs.

The x86 ISA is an ugly mess full of legacy ISA bloat that needs to be removed while the RISC designs are better for low power usage on Tablet/Phone mainstream designs. I hope that AMD will rethink pushing the release of its custom ARMv8A ISA running K12 design to the point that its K12 project is effectively mothballed. K12 is supposed to be like Zen on the inside but K12 was engineered to execute the ARMv8a ISA and AMD better think twice about becoming a little version of Intel and keeping the world dependent in that nasty nasty x86 ISA.

IBM and all those other OpenPower power8/power9(RISC ISA Designs) licensees, including Google, will not have to worry too much about the complexities and erratas that arrise from using that CISC x86 ISA and its legacy ISA bloat so full of curd.

AMD has no issues with this errata/defect currently but Lisa Su better get that custom K12 project restarted because AMD needs to become and ISA indipendent CPU maker that can work with more ISAs than only that ugly x86 ISA.

So many Intel apologists out in force today working that damage control.

Intel is a hated abusive monopoly interest that abused its market prosition(Intel Got Busted doing so) and hampered progress/innovation while forcing millions to suffer from its dog food graphics over the years. So any Intel hate is well deserved among those with an actual clue about what Intel's nefarious actions have done to damage the x86 ISA related market. AMD needs to get onto other CPU ISA options also and not become a Little version of Intel.

January 3, 2018 | 05:11 PM - Posted by mAxius

Shots Fired AMD finally Found its backbone

"AMD Says ‘Near Zero Risk’ to Its Chips"

https://www.barrons.com/articles/amd-says-near-zero-risk-to-its-chips-15...

Intel is Fud master

January 3, 2018 | 08:50 PM - Posted by jedeye1973 (not verified)

Sounds like the makings of a possible class action lawsuit. (Possibly)

January 6, 2018 | 12:55 PM - Posted by Caw (not verified)

This is false for I just had a meltdown on my amd processor AMD FX-4100 Zambezi Quad-Core 3.6GHz last week on December 30 2017.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.