Spectre/Meltdown/L1TF/MDS Mitigation Costs On An Intel Dual Core + HT Laptop

[ArchLinux]

Once again great stuff by Michael

Really weirdo stuff by the commenters:

Intel is a mess. I hope in a class action against this scammer!

This is one of the few situations where I've actually been able to notice the performance losses without needing benchmarks to confirm my observations. I've noticed my i3 Haswell laptop getting slower and my overall CPU usage going up the past few months.

I'm actually a little surprised Intel hasn't been more heavily hit yet. I don't think anyone could sue them for the slew of security risks they put so many millions of people into, but, they could be sued for selling products that don't perform to the specs they expected. Not only has Intel not lowered their prices, they actually brought them up, due to the shortage

Intel isn't under obligation to have vulnerability-free products, geniuses.

[coder]

I don't understand why they keep patching holes on Intel CPUs. As long as you don't kill Intel Management Engine IME, you are actually exposed as a naked fly.

Don't be silly - they have different exposure profiles.

You can be exposed to Spectre/Meltdown vulnerabilities even from Javascript code running in a web browser, whereas IME would not. AFAIK, IME vulnerabilities require local code execution, or at least being on an unsecured network.

[coder]

Phoronix: Spectre/Meltdown/L1TF/MDS Mitigation Costs On An Intel Dual Core + HT Laptop

Timed Kernel Compilation would've been nice to see, for those of use that do software development on our laptops.

[noangel]

Wow, the difference is huge. I'll think twice before buying Intel CPUs again. And they are still more expensive than CPUs from AMD!

[bepvte]

Any idea which protections are the most important? I would like to regain some speed but I do believe I need some migitations. Which is the easiest or least easiest to execute?

[ermo]

As someone who has a penchant for using older hardware -- albeit most of it AMD -- this has my attention. The newest piece of hardware I operate is a Skylake equipped dual core laptop with SMT.

Here's hoping Zen 2 isn't too vulnerable and errata ridden.

[willmore]

First of all: thanks Michael for performing these benchmarks!

It is not the uArch it's the way Intel implemented their processors, they did some unsafe optimizations, which do not clear all shared resources after use. This could happen with arm as well, as we might have deep pipelines and implementations of the uArch might perform unsafe speculative execution of branches to increase performance.. they just don't do it, as ARM is mostly used on power sensitive devices and executing something "just in case" may waste power.

"the way Intel implemented their processors" is the uArch. That's the part they broke. If you're trying to say that x86 isn't the fault, just the way that Intel implemented it, then that's the uArch, not the arch.

[Spacefish]

First of all: thanks Michael for performing these benchmarks!

It is not the uArch it's the way Intel implemented their processors, they did some unsafe optimizations, which do not clear all shared resources after use. This could happen with arm as well, as we might have deep pipelines and implementations of the uArch might perform unsafe speculative execution of branches to increase performance.. they just don't do it, as ARM is mostly used on power sensitive devices and executing something "just in case" may waste power.

[Bsdisbetter]

Intel is a mess. I hope in a class action against this scammer!

Actually it's the architecture that's the problem:

https://secure64.com/not-vulnerable-intel-itanium-secure64-sourcet/

No problems with openvms on itaniums

[cfeck]

I don't understand why they keep patching holes on Intel CPUs. As long as you don't kill Intel Management Engine IME, you are actually exposed as a naked fly.