Do Apple's Security Updates for Meltdown and Spectre Slow Things Down?

Apple has security updates out now for most of their OSes and I am hesitant to apply them as I have heard that any potential fix for Meltdown and Spectre have a negative impact on CPU performance... a hit I'm not willing to take. I don't believe the risk is great especially since no exploit yet exists. And with most of the population fixing their stuff, there would be no incentive to write a malicious script that could be used on the few people who did not update. Real hackers and want as big a target as possible. Script kiddies won't be able to exploit it.

But on the other side, I REALLY hate seeing that red notification on my App Store icon....

I installed the fix for Sierra 10.12.6 today and before/after Black Magic and GeekBench only indicates 1% slowdown, within the margin of error. This was on a 2017 iMac 27 (which uses a Kaby Lake i7-7700K CPU).

It's a tricky area because supposedly older CPUs are subject to more performance impact. The CPU feature which reduces slowdown from Meltdown patches is called PCID which apparently on Haswell or newer CPUs: arstechnica.com/gadgets/2018/01/heres-ho...ll-hurt-performance/

I believe the 2013 Mac Pro uses Ivy Bridge Xeons, which is older than that. 2013 and later iMac 27s have Haswell or later CPUs which supposedly have PCID.

Based on my measurements, macOS 10.13.2 Supplemental Update and macOS 10.13.3 have no effect, compared to macOS 10.13.2 (the earliest version of High Sierra I have.) I used GeekBench 4. running it three times on macOS 10.13.2; both the Single-Core and Multi-Core scores were within less than 1% between the three runs. After installing 10.13.2 Supplemental Update and again after installing 10.13.3, GeekBench scores were virtually identical to those of 10.13.2, also being within less than 1%.
I have an iMac Pro, 10 core, Vega 64.

Even though I used Geekbench, it's not a good test for Meltdown patch performance impact. The Meltdown "page table isolation" patch affects system call overhead not compute performance. Benchmarks or regular apps like Geekbench tend to spend long periods in compute-bound code within user mode. They don't make that many system calls. By contrast I/O-bound apps do more system calls.
However an I/O-intensive app which does large I/Os will be less affected than an app which does many small I/Os (hence higher syscall frequency).

FCPX does a mix of large I/Os for video files and small I/Os for database I/O when managing plist files, and building/maintaining thumbnails in the event browser. If there was any FCPX scenario that might see a marked slowdown from Meltdown patches it would be if when doing many small I/Os on an older (ie before Haswell) machine which lacks PCID. E.g, with the event browser in filmstrip mode set to one thumbnail per clip and when scrolling down through a large event. That's when it does lots of small I/Os.

Here is an interesting article on the topic. // link deleted by Mod //

I updated to 10.13.2 and have noticed no difference in performance during my daily editing and graphics work. There's the nanosecond on paper, there's the end user experience. Technically there could be a drop in performance, but it's nanoseconds an end user would never notice in real world work, is my conclusion.

dmb wrote: Here is an interesting article on the topic. link deleted by Mod

That link is being reported as very suspicious by all browsers I have and they're refusing to connect.

Redifer wrote: …That link is being reported as very suspicious…

apologizes Joe to mess in your post above, deleting that doubtful link…
Not only not-trusted, but - tweaking it a bit - not suited to minors … :ohmy: