Hardware Will upcoming CPUs be launched pre-patched for Meltdown?

[Futurdreamz]

I'm thinking I'll want to upgrade to a Ryzen 2, but I'm wondering if it's going to launched patched or if they will push back the launch? Does anyone know?

 

[Deleted-394630]

I hope they do, but the way it's looking right now, they don't have much time. Only time will tell. Such a shame that I'm building a PC right now while Ryzen 2 is about to come out, but I have no choice but to build before May, and have Intel-Phobia. Not to mention GPU Prices...


~distant sobbing~

 

[gnmmarechal]

I'm thinking I'll want to upgrade to a Ryzen 2, but I'm wondering if it's going to launched patched or if they will push back the launch? Does anyone know?

AMD processors aren't vulnerable to Meltdown.

 

[Deleted-394630]

AMD processors aren't vulnerable to Meltdown.

What about Spectre?

 

[WiiUBricker]

You can’t know for sure if they will be pre-patched if there is no announcement about it. What we know, however, is the fact that there have been rumors that Intel brought Coffee Lake to the market despite them knowing of Spectre and Meltdown.

 

[Deleted-394630]

You can’t know for sure if they will be pre-patched if there is no announcement about it. What we know, however, is the fact that there have been rumors that Intel brought Coffee Lake to the market despite them knowing of Spectre and Meltdown.

That's part of my Intel-Phobia, quite a few people share it.

 

[kuwanger]

AMD is vulnerable to at least one variant of Spectre. Depending on your OS, you might be lucky to have a kernel that compiles with Retpoline to mitigate the attack in kernel space. Meanwhile, Chrome/Firefox/Edge all are introducing their own software mitigation.

The thing is, it's unclear that it's even meaningful exploitable except in limited circumstances (trying to get crypto keys) from what I understand about how it works. There's also the point that it's highly likely there's a whole lot of other attacks that are similar to Spectre that will be found. I say this because if you're watched any CCC videos from the last few years, it's clear that there's tons of cache-based side-channel timing-based attacks possible with all sorts of clever approaches. So, unless there's some effort to actually isolate all the various possible timing variations within the chip design, we're likely just stuck with various workarounds in software and having to put out the next fire that comes along. Or we have to give up speculative/OOO execution and start having hard limits on cache along with even further isolation of things like flash/javascript.

 

[Originality]

Intel have announced that they will work on making sure future products are not vulnerable to Meltdown and Spectre. How they do this is a matter of waiting and seeing.

 

[Deleted-394630]

Intel have announced that they will work on making sure future products are not vulnerable to Meltdown and Spectre. How they do this is a matter of waiting and seeing.

Wasn't meltdown a firmware issue? I bet they'll just rewrite some of the firmware. Easier said than done of course.

 

[kuwanger]

Wasn't meltdown a firmware issue?

I'd say not really. It requires reordering the logic of when memory protection faults should kick in. That's a pretty non-trivial amount of work--mostly in catching all the cases so we're not left with a partial Meltdown patch.

Intel have announced that they will work on making sure future products are not vulnerable to Meltdown and Spectre. How they do this is a matter of waiting and seeing.

In the end, yes. I'm not particularly positive about Intel at this point, though, given how much Intel has tried to make Meltdown and Spectre synonymous, claim that "many" CPUs are vulnerable to Meltdown, and generally paper over the performance costs. I presume that's a major reason Linus' had such a bad reaction to Intel's patch to the Linux kernel--they imply their plans are for the fixes to be opt-in, which implies there actually will be substantial performance costs. *sigh*

PS - Still wondering if MS is ever going to switch from using INVPCID to just PCID to reduce the performance costs for the Meltdown workaround. My guess is no. Thankfully, at least a lot of benchmarks imply the Meltdown workaround isn't really that bad with games. Not seen any real comprehensive tests, though.

 

[Psionic Roshambo]

https://arstechnica.com/gadgets/201...-fixes-coming-2018-3d-xpoint-ram-not-so-much/

And that concludes all my information on this subject lol

 

[osaka35]

My understanding is it's more about how they do prefetching, where they assumed certain parts were secure when they actually aren'weren't. I can't imagine you could fix something like this with a simple hotfix, or even a simple hardware revision, without impacting speed or abilities. Even if the next batch "fixes" the problem, it might make them run less efficiently...though hopefully not terribly less so. At least, this is my surface understanding.

 

[Futurdreamz]

http://www.tomshardware.com/news/amd-patches-spectre-flaw-variants,36323.html

…It sounds like AMD is fairly certain they can cover it in firmware/OS/software updates, so it may not be necessary to delay Ryzen 2. Which is good, because I just have a Ryzen 5 1500x and am hoping to upgrade to a Ryzen 7 2800x when it launches next month.