The Current Spectre / Meltdown Mitigation Overhead Benchmarks On Linux 5.0

[phoronix]

Phoronix: The Current Spectre / Meltdown Mitigation Overhead Benchmarks On Linux 5.0

With it being a little over one year since Spectre and Meltdown mitigations became public and with the Linux kernel today hitting the big "5.0" release, I decided to run some benchmarks of the current out-of-the-box performance hit as a result of the current default mitigation techniques employed by the Linux kernel. The default vs. unmitigated performance impact for Spectre/Meltdown are tested on an Intel Core i7 and Core i9 systems while there is also an AMD Ryzen 7 box for reference with its Spectre mitigation impact on Linux 5.0.

http://www.phoronix.com/vr.php?view=27658

[blueweb]

Are the stock and not mitigated values switched for the netperf benchmark? EDIT: Nevermind, it's fine. I'm just tired.

P.S. If I try to include a screenshot image, the forum says I'm not authorized to create or remove attachments.

[schmidtbag]

Some of those stock results are shockingly terrible on the Intel side. In some cases, I'm not sure worrying about the mitigations are even worth it, at least for home users.

[Michael]

Are the stock and not mitigated values switched for the netperf benchmark?

P.S. If I try to include a screenshot image, the forum says I'm not authorized to create or remove attachments.

Hmm? Netperf is "higher is better".

FYI, all the graphs / data management from start to finish is managed by PTS, so not possible for cases of switching to wrong values. But when checking out the Netperf results, they favor the non mitigated results.

[andyprough]

birdie wrote this on Aug. 25, 2018, and still true to this day:

To this date, and those vulnerabilities are now eight months old, there hasn't been a single virus detected or an intrusion incident occured which involved them yet kernel developers have slowed down our PCs without giving an option to have the lost performance back.

Six months later, still no exploits in the wild: https://searchsecurity.techtarget.co...ctre-variant-2

Although researchers have discovered over a hundred samples of malware that exploit Spectre and Meltdown, there's still not a single viable exploit that's been found in the wild.

[blueweb]

Hmm? Netperf is "higher is better".

FYI, all the graphs / data management from start to finish is managed by PTS, so not possible for cases of switching to wrong values. But when checking out the Netperf results, they favor the non mitigated results.

Sorry, you're right, it seems fine. Not sure what confused me...

[Licaon]

The Current Spectre / Meltdown Mitigation Overhead Benchmarks On Linux 5.0 Written by Michael Larabel in Software on 6 March 2019. Page 1 of 6. 5 Comments

Written in the future? Michael

[Michael]

Written in the future? Michael

Fixed, thanks. Will be away from my benchmarking systems for a few days and have a number of articles in advance and pegged that display date in for the wrong day.

[milkylainen]

Some of those stock results are shockingly terrible on the Intel side. In some cases, I'm not sure worrying about the mitigations are even worth it, at least for home users.

And you'd be right about that. But side-channel attacks are still pretty uncommon in the wild, regardless if it is a CPU side channel or something else. Most public joe attacks are still in the realm of basic exploiting. Side channels are more for the avid government facility hacks and general spying.

[starshipeleven]

Six months later, still no confirmed exploits in the wild: https://searchsecurity.techtarget.co...ctre-variant-2

fixed

Lack of proof is not proof of lack.