All Intel CPUs in last 10 yrs have critical bug to show protected kernel memory areas

D

defaultluser

[H]F Junkie
Joined
Jan 14, 2006
Messages
14,398
https://www.theregister.co.uk/2018/01/02/intel_cpu_design_flaw/

MUCH More Detailed Link On This Issue From Python Sweetnes

Linux Kernel Patch Page

This means real trouble for anything and anyone who uses the internet. With proper exploits, a simple JavaScript program from a web page could read into your kernel space!

Performance impact is hard to assess, because the performance hit would depend on how many times your application hits kernel memory. But I would expect the biggest impact from things like databases, and other I/O heavy apps.

We won't have official performance impact numbers on Windows for a month, when Microsoft makes the patch official. We might get some Linux benches earlier than that.

But that could be enough to close the gap between AMD's Zen and Kaby Lake. Because this bug and corresponding software patch only applies to Intel.

And you all remember what the performance impact happened with AMD's TLB bug on Phenom version 1.0 :(
 
Last edited:
Wow.....ouch 30-35% performance hit? That's pretty big fuck up if this is true. I am sure we will see results.

But damn....
 
Is there a list of affected CPU's? Does this include my crusty old i7-920?
 
No official list until there's an official announcement from Intel. That could be as early as Thursday (rumor I saw elsewhere), or could be longer wait.

WE just found this out about this early because the Linux commits are not secrets.
 
I would like them to shed more light on this and give more information on this bug and its deep lying effects.
 
I wonder if this only affects specific enterprise workloads like the TLB bug, or whether the performance loss is more pervasive.
My understanding is that the current workaround requires flushing the caches every time a context switch (transition from user to kernel mode) happens; this would impact I/O heavy operations. The database guys are screwed for sure, since their workloads are full of small I/O requests, but the 15% loss in performance on the compiler benchmark is a little concerning...
 
DrLobotomy said:
Is there a list of affected CPU's? Does this include my crusty old i7-920?
Click to expand...

Seems like everything from Pentium Pro to current gen. (Yikes)

If that's the case it's crazy to think that this bug has existed for ~23 YEARS and only just now discovered.
 
Egg ===> Intel Face

Wonder if it'll have a cutesy bios/app switch like the TLB bug?
 
Shintai said:
There are benches already. Gaming, encoding etc got 0-1% impact.
Click to expand...
Show me .1% and 1% lows, frametimes, and loading times. Make sure the frame testing is done at 720p. THEN we will talk about "0-1% impact". Average fps means nada.

I believe we will see 5-10% based on what I've been reading. Most people will cite GPU bottlenecked benchmarks and say things like "You won't notice the difference" ... Like I said, 5% or more in any game/app I use, and mine is going back to Newegg. I'll take Ryzen2 at 95% performance and half the cost.

Intel costs a premium and we are paying for that premium, no excuses. End of story.
 
i am sure gaming is not the issue with this kind of bug, it is the kind of bug that could be responsible for billions in losses, not how many frames they get on tomb raider.
 
Well this is not good especially for the older CPU's as i understand it, the patch to fix the security bug is going to be rougher on some systems more then others, I worry this is going to have some harsh effects on my 2nd and 3rd Gen i7's and i5's that my brothers still use for gaming.
This is not good for us the end users but we are not the ones intel is going to be more concerned with, that would be there server and data center business and if the patch hit the bulk of there systems with a sever performance losses of the rumored "up to 30%" it could really hurt there reputation, with that said i still only really care how this is going to effect me, I'm really hoping it's not going to rek the older core CPU's they still had much life left to give
 
Last edited:
Kato1144 said:
Well this is not good especially for the older CPU's as i understand it, the patch to fix the security bug is going to be rougher on some systems more then others, I worry this is going to have some harsh effects on my 2nd and 3rd Gen i7's and i5's that my brothers still use for gaming.
This is not good for us the end users but we are not the ones intel is going to be more concerned with, that would be there server and data center business and if the patch hit the bulk of there systems with a sever performance losses of the rumored "up to 30%" it could really hurt there reputation, with that said i still only really care how this is going to effect me, I'm really hoping it's not going to rek the older core CPU's they still had much life left to give
Click to expand...

You make a good point in that the performance impact will not be the same across all generations and until someone shows testing for each one we simply won't know how much or how little this affects the hardware. I have 2 X58 systems and 1 X79 system still going strong. I hope they aren't impacted severely.
 
https://www.mail-archive.com/[email protected]/msg1553070.html

:LOL:

Code: 
2) Namespace

      Several people including Linus requested to change the KAISER name.

      We came up with a list of technically correct acronyms:

          User Address Space Separation, prefix uass_

          Forcefully Unmap Complete Kernel With Interrupt Trampolines, prefix fuckwit_

      but we are politically correct people so we settled for

        Kernel Page Table Isolation, prefix kpti_

      Linus, your call :)
 
I wonder if it's time to upgrade my 4770k/e3-1240v3 systems to threadripper.
 
Wow! Intel better have spare capacity to ramp production. They could charge more too if there is a rush. Win win! (for Intel and AMD)
 
nomas said:
Wow! Intel better have spare capacity to ramp production. They could charge more too if there is a rush. Win win! (for Intel and AMD)
Click to expand...
3 months from now ... "Intel CPU shipments increased by 30% in Q1 2018."
What a coincidence.
 
  • Like
Reactions: Sufu
like this
I don't do much gaming inside Virtual Machines on my rig, so not bothered in the slightest. :cool:
 
Sure am glad I ponied up big bucks on a SL 8700k now :( My 5.1ghz will suddenly become (effectively) 3.57ghz.
 
arnemetis said:
Sure am glad I ponied up big bucks on a SL 8700k now :( My 5.1ghz will suddenly become (effectively) 3.57ghz.
Click to expand...

Can’t tell if serious or just not capable of reading. Windows benches have been posted, there is almost no user impact for gaming, rendering, or disc access.
 
Ocellaris said:
Can’t tell if serious or just not capable of reading. Windows benches have been posted, there is almost no user impact for gaming, rendering, or disc access.
Click to expand...
Sorry, forgot the /s. Half tongue in cheek there. The reality is until we see a [H] writeup, we don't know the realistic impact this will have.
 
This is brutal. And makes me real happy i went with AMD this time around. This isnt just a few % this is huge.
 
Ocellaris said:
Can’t tell if serious or just not capable of reading. Windows benches have been posted, there is almost no user impact for gaming, rendering, or disc access.
Click to expand...

Bit early to come to that conclusion. We seen a couple of benches that showed only a minor hit to performance, what we have not seen are frame times as all this swapping back and forth could introduce stuttering. Way to early yet to say for sure one way or the other. Patches should be out soon for many to play with tho and will see then what impact it makes.
 
Anyone can sign up for Windows Insider program and get the changes now if they are interested. Who is going first? :p
 
I think the HardwareLuxx article was updated with a couple more gaming benchmarks which show a slightly larger drop.
 
TaintedSquirrel said:
Show me .1% and 1% lows, frametimes, and loading times. Make sure the frame testing is done at 720p. THEN we will talk about "0-1% impact". Average fps means nada.

I believe we will see 5-10% based on what I've been reading. Most people will cite GPU bottlenecked benchmarks and say things like "You won't notice the difference" ... Like I said, 5% or more in any game/app I use, and mine is going back to Newegg. I'll take Ryzen2 at 95% performance and half the cost.

Intel costs a premium and we are paying for that premium, no excuses. End of story.
Click to expand...

This, nobody gives a damn about their cpu's for average fps at 4k. Frametimes and smoothness (the s[H]it we care about) could take a big hit without affecting dumb fps numbers very much.

The I/O hit seems like it is pretty bad in the enterprise loads, some regular programs can also be I/O bound during various tasks.
 
StormClaw said:
What if i disable Win updates and never install this patch?
Click to expand...
it should go with out saying that not patching this is a mistake unless you plan to keep that PC off the internet, there might not be a program out there that takes advantage of this bug yet but you can bet your ass some one some where is working hard on capitalizing on intels failures to make some serious $$$ and this will be a glaring security hole, infact it already is we just didn't know it existed until recently.
 
Yeah, I never trust Phoronix to do anything in-depth, which was why I didn't initially post the benchmarks. They're useful for a quick read, but lack any sort of depth.

We'll have to wait for someone who actually care about the tests before we'll get anything useful on the real gaming impact.
 
defaultuser is correct, it is too early to hit the panic button, for all we know the performance hit might not even affect the average user, it might only hurt a small percentage of applications and use cases, i personally don't know and im sure the same goes for almost every one here. We will have to collectively wait for the patch to be availability to the masses and then do the testing from there until then this is mostly speculation
 
maxius said:
10 years try 20 last intel cpu that didnt use "speculative execution" was the og pentium
Click to expand...

No, don't update the thread, all rumors point to the speculative memory access added in Core 2 Duo being the cause of this.

It's just 10 years old.
 
Last edited:
You must log in or register to reply here.
Back
Top