Hacker News Reader

Comment by mtgx | Hacker News Reader

They will update 90% of the chips 5 years old or newer, which I assume includes Ivy Bridge and later. But the bug affects older chips, too. Important to keep that in mind.

View on HN

I don't know why you're being downvoted; it's crucial that Intel patches this deficiency for all CPU models which are currently in use, which is pretty much all of them, since some systems work nearly forever. We're talking about going back 15 years or so?

I mean this is like the Y2K bug, in that every system needs to have its CPU verified to ensure that it's covered by the patches, otherwise the system should be shut down and a new CPU provided. Since many of these CPUs have sockets for which CPUs are not produced anymore, whole new systems will have to be purchased.

Of course this might the result Intel wants to achieve.

I imagine most laptop and motherboard vendors have precisely zero interest in locating/dusting-off whatever toolchains are required to mint new BIOSes for long discontinued products.

The actual quote is:

> Intel has already issued updates for the majority of processor products introduced within the past five years. By the end of next week, Intel expects to have issued updates for more than 90 percent of processor products introduced within the past five years.

The post is silent on what will be done with older chips, it didn't say that they are not going to be updated. It makes sense that Intel prioritizes the newer chips.

It makes sense but 5-10 year old laptops aren't uncommon nowadays.

Had this occurred 20 years ago nobody would have cared because computers would get replaced super fast anyway. Nowadays, even a geeky household can have 5+ year old CPUs lying around and still used.

Anecdotally: I game, compile, transcode, and do any other number of CPU-intensive tasks on my machine. My household is definitely "geeky".

I still run an i7 2600k, which is a Sandy Bridge CPU introduced 7 years ago.

I don't (didn't? may not?) have any pressing reason to upgrade, although I was considering it. Funnily enough, even before now I was considering the AMD Zen architecture, and although I'm still undecided I'm being pushed further and further towards AMD through ethical considerations alone.

You won't need a microcode update. The microcode update is only needed for Broadwell and newer (for those, retpolines don't work). Also, hello, fellow 2600k owner :D

Some careful wording there. I'm guessing that "introduced" means when they initially went on sale or something like that. I'd like to know what percentage of systems sold to consumers 2-3 years ago had processors that were "introduced" 5 or more years ago.

It's similar to the absurdly short supported lifetime of many phones, they support them for "2 years after release" but in many cases people are still buying them brand new 2 years after release.

https://newsroom.intel.com/wp-content/uploads/sites/11/2018/... - processors older than 5 years don't need a microcode update, and probably won't even benefit. The new instructions for preventing variant 2 is inferior to retpolines, even according to Intel...

> Note that the insertion of LFENCE must be done judiciously; if it is used too liberally, performance may be significantly compromised.

Wow, as if we could press some magic button and it would pinpoint all the places where a branch condition involves user-controlled data. Of course Intel is downplaying this one to the max, the performance implications for a thorough fix are devastating.

It's okay, retpolines are similarly so for C++: https://reviews.llvm.org/D41723

Again, obviously, not all indirect jumps need this, but still, Intel's fix might be even worse.

It's not the indirect jumps that concern me but all the conditional branches, such as loop conditions.

Reply on news.ycombinator.com