MDS / Zombieload Mitigations Come At A Real Cost, Even If Keeping Hyper Threading On

[perpetually high]

Yes, but if you do please recompile the kernel without retpolines, without this, you're still comparing apples to oranges.

My understanding is you can leave CONFIG_RETPOLINE=y when compiling the kernel, and just disable it via a boot parameter; spectre_v2=off does essentially that. That parameter disables any and all spectre v2 mitigations (which includes retpoline)

[torsionbar28]

The sad part is that more mitigations are inevitable, and CPUs will get even slower. It's just a matter of time.

Now, imagine the possibility that this is Intel's attempt at planned obsolescence, and that they're paying people to scrutinize their current chips so they can release expensive chips which are supposed to be "secure". Sprinkle in some FUD about AMD, and you have a perfect recipe to bump the prices up and get ahead of the competition. This is just my attempt to consider the wildest possible conspiracy theory, but I wouldn't be surprised if that turned out to be true, just like that landfill of E.T. cartridges in New Mexico.

It is certainly a plausible scenario. Current chips faster than the competition via unscrupulous short-cuts, then forced obsolescence by "discovery" of critical security flaws. This is a win-win for intel, and a lose-lose for intel customers. I have a feeling EPYC is going to become very popular in certain market segments...

[starshipeleven]

My idea is that they could announce such chips in the future, and it doesn't even matter whether they'll actually be secure. The point is that they could be trying to pull off such a marketing stunt.

Eeeh, that would involve throwing a whole truck of shit over themselves (their previous selves anyway), I still don't see that happening.

If they pushed so hard that their older processors were flawed, then this would be a PR hit for Intel as a whole, and it would orient more people to buy AMD instead than whatever new and "secure" chip Intel is selling later.

I mean, with what kind of straight face can they claim their new CPU is "secure" after their older processors were such a cesspool clusterfuck of hacks for the sake of speed while they still adamantly claimed "all is fine" and paid/bribed/NDAed companies to keep any bug or issue covered up?

[starshipeleven]

I have a feeling EPYC is going to become very popular in certain market segments...

Yes it is. AMD is on a roll for server market and high-end stuff. https://pcper.com/2019/05/frontier-f...deon-instinct/

[angrypie]

Oh wow, at this rate Intel CPUs will perform slower than a stock FX. No sight of hope inside the Pandora's box.

[Mel Spektor]

if you are still buying intel on 2019 you should feel ashame.

[GreenReaper]

My understanding is you can leave CONFIG_RETPOLINE=y when compiling the kernel, and just disable it via a boot parameter; spectre_v2=off does essentially that. That parameter disables any and all spectre v2 mitigations (which includes retpoline)

It is still likely to mean it can no longer compile in a direct call. I don't have figures on whether it has a significant impact, though.