This link has just been posted up on Hexus, thought it would find interest here too. https://www.theregister.co.uk/2018/01/02/intel_cpu_design_flaw/ Super eh? Short and sweet as i'm on my phone
5-30% performance hit? Ryzen 2 could actually pass Coffelake IPC it it's near the top of that number... Time to buy some AMD shares?!
IF that proves to be true, I think I'll be laughing my b*llocks off somewhat We'll just have to wait and see eh - facts over fiction chaps, the proof is in the pudding and all that...
This, AMD drivers breaking a load of DX9 ****, yet another macOS/OSX security hole... ...tis the day for **** being broken it seems. Also, given how long the flaws have been around, 10-ish years for intel.... 15 years [reportedly] for the macOS security hole... i know it's all very complicated, but you'd hope **** like this would get spotted sooner...
Is this a trillion dollar mistake? The skeptic in me also wonders if it's just a ploy to get everyone to upgrade in a languishing CPU market <me>lovingly pats 2500k of nearly 7 years </me>
Title amended to have the '?' It was supposed to have. Anyway, i'm still trying to get my head around all this (I ain't no IT ninja master). If there is such a fundamental problem how can an update to the OS remedy a hardware flaw and be secure? I want to see what comes out in the wash, this just seems too much of a massive fubar to take at face value. If true though, my money is on W10 being, coincidentally, the least affected....
I'm writing this up this morning, and I can confirm it's genuine: the Linux patch is already in place and marks pretty much any Intel chip as 'insecure.' (It also does the same for AMD chips, which is a result of the severity of the flaw causing the devs to err on the side of caution - there's no evidence AMD parts are affected, and AMD's working on its own patch which will remove the insecure flag.) Same way Intel 'fixed' the F00F bug: when you know what triggers the problem, you just have a bit of the OS that says "don't do that." Imagine you've got a bike which falls over every time you turn left: you'll soon learn to only turn right. You'll get to your destination eventually, but more slowly than if the bike didn't have the flaw.
Whether or not AMD is affected, from what I understood, the remedy to this design flaw changes the fundamental way user space and kernel space communicate, meaning it'll probably be applied for all chips regardless. Or am I wrongsville?
From my understanding, the performance-sapping change need only be applied to insecure parts (i.e. everything Intel), and doesn't need to be applied to AMD parts but currently is being applied to AMD parts until AMD gets its own patch saying "don't apply this to us" merged.
Knew you'd be on the case I get the premise that you just patch the OS to say "don't do that" but a software update surely could be worked around by some hoodlum with bad intentions? Or am I just exposing my lack of software knowledge which is more than possible Tell you what, why don't I just read your article when you write it.
Looks like phoronix has done some early performance testing on how the patch effects performance on Linux. In light of this security flaw that could see Intel CPUs facing a 5-30% performance hit does anyone think Brian Krzanich selling all the shares he could legal off load rather suspicious?
Technically it would be trivial to have the patch only apply for specific CPUs (provided the manufacturer is forthcoming with information which ones exactly are affected), however if AMD did not rely on the old way to do things then it won't matter if the patch is applied to all CPUs. In other words: Road a is 1 mile long but has a landmine hidden somewhere. Road b is 1.3 miles long but has no landmine hidden. So forcing everyone to take Road b for safety reasons won't slow down anyone who used Road b to start with.
Piece is live now, and includes a link to the patch which prevents PTI (the performance-sapping please-don't-hack-my-Intel-chip patch) from being applied to AMD processors.
I feel like this is going to push some of our customers, unhappily, to new servers if they're already running theirs on the ragged edge..
Windows patch is coming soon apparently. Again, won't be Intel specific so AMD need to unpatch the patch. Hoping to see plenty of articles comparing gaming performance etc.
This is a huge problem for Intel. All the major cloud providers will be affected by this and customers of theirs will all be closing monitoring performance over the next couple weeks as well as scaling services up. I can easily see a class action lawsuit hitting Intel as well. They have more than enough money to deal with it but they will surely lose some significant income, especially as AMD actually has a viable alternative now.
