“Apple released iOS 11.2.2 update to address Spectre security issues,” Melv1n reports. “I was curious about the actual performance change. So I’ve benchmarked my iPhone before and after the security fix.”
“The benchmarks were performed before/after updating iOS in the exact same scenario: no apps running (including background). So the running benchmark basically had the full CPU capacity and all other iPhone resources to its disposal,” Melv1n reports. “All numbers point to the same conclusion: it took a serious hit in performance at every possible level. A lot of benchmark levels show a significant decrease in performance on the iPhone 6 up to 50% on some benchmark levels.”
“Some have fairly pointed out the results could be influenced by the battery throttling that was exposed last month and confirmed by Apple,” Melv1n reports. “That may be a technica[ly] correct argument (which has not been proven by vendor benchmark numbers correlated with the Spectre patch). Several other users and reporters mentioned fluctuating benchmarks with some showing no loss of performance but others did (which were already throttled, so Spectre specific). The whole thing seems to be all over the place right now with no 100% answer what’s right.”
Read more in the full article here.
MacDailyNews Take: Of course, we require more confirmation than just a single, rather old iPhone 6. We await more results for independent parties to see how much these software bandaids negatively affect these and many other inherently defective CPUs. (Note: iPhone 6 is powered by Apple’s A8 SoC.)
CERT: The only way to fix the Meltdown and Spectre vulnerabilities is to replace the CPU. Intel et al. are going to try to sell us on a software bandaid instead of really fixing the problem properly. Watch and see. https://t.co/OeC2AoPdlK #Intel #AMD #ARM
— MacDailyNews (@MacDailyNews) January 4, 2018
SEE ALSO:
Intel’s Spectre patch is causing reboot problems – January 12, 2018
In wake of Spectre and Meltdown, Intel CEO offers open letter, looks to restore confidence in Intel CPU security – January 11, 2018
Apple releases iOS and macOS updates with a mitigation for Spectre CPU flaw – January 8, 2018
Meltdown and Spectre: What Apple users need to know – January 8, 2018
How Apple product users can protect themselves against Spectre and Meltdown CPU flaws – January 5, 2018
Intel’s CEO Brian Krzanich sold off the majority of his shares after finding out about the irreparable chip flaws – January 4, 2018
Apple: All Mac systems and iOS devices are affected by Meltdown and Spectre security flaws – January 4, 2018
CERT: Only way to fix Meltdown and Spectre vulnerabilities is to replace CPU – January 4, 2018
Security flaws put nearly every modern computing device containing chips from Intel, AMD and ARM at risk – January 4, 2018
Apple has already partially implemented fix in macOS for ‘KPTI’ Intel CPU security flaw – January 3, 2018
Intel’s massive chip flaw could hit Mac where it hurts – January 3, 2018
So a random benchmark without attribution shows the same numbers as before on an iPhone with an old battery? WHAT A SCANDAL!!!! Jesus fucking Christ… get a new battery, then test it. You can’t attribute those numbers, on an unnamed benchmark test, to the new security update. I’d like to slap all of these “reporters” for their lack of understanding on the subject they report on… Idiots.
Maybe, maybe not. I wouldn’t be so quick to dismiss those results.Those are big speed differences and although definitely not scientific should be enough to at the very least make you curious as to what may be going on, no?
No, because those number are EXACTLY the same as the numbers from a few weeks ago before this update happened. The author doesn’t specify the benchmark, doesn’t specify when it was performed, and doesn’t specify if the device has a new battery. In other words? COMPLETE AND TOTAL NONSENSE.
Maybe, I prefer TO KEEP AN OPEN MIND, especially when I feel obliged to shout back.
Yes, but if these numbers are even remotely correct we will soon see an avalanche of these reports, so no need worry right now about the validity of this particular one.
The bug exists because of vulnerabilities inherent to every processor that uses speculative execution. The only possible fix is to prevent some instances of that from occurring. That will carry a performance penalty, whether the fix is in software, firmware, or hardware. New chips may mask the penalty by running faster, but it will still exist.
Nevertheless, the fixed computers will run just as fast as they should have run safely in the first place. The extra speed before the fix was gained through a procedure that has now (20 years later) been shown to be vulnerable to exploitation.
Champaign corks popping all over the PC and smartphone world as people with slower hardware head to Best Buy to replace tgat ancient iPhone 6 or laptop (assume something as fundamental as a “CPU nanny” will be needed EVERYWHERE they are sole.
Other thought is the thresholds for the dreaded “battery need replacing” warning may have been “adjusted” (as i would expect a CPU nanny would generate some serious heat leading to faster shutdowns)
Decrease performance was expected.
I would guess that the first patch is a rush job to close the security hole. I would expect that over several more OS updates/iterations, the code will be massaged and efficiencies will increase to decrease the performance hit.
Just stop. MDN you can do better.
That ridiculous “test” was on reddit a couple days ago. The hit apparently was a result of an OLD BATTERY with the changes Apple made with 11.2 onwards.
Numerous other users with newer phones or newer batteries could not ducliplate the result. In fact, the security patch appears to have a minimal impact.
Find it very hard to believe the battery throttling is THAT bad. Those are some low scores!
If it is not the battery, then it’s either falsified clickbait or “sponsored” FUD. (Hi, Samsung!)
In any case, MDN does not need to feed the trolls traffic.
Benchmarks on my iPhone 6s, 9.7” iPad Pro and latest iPad were less than 1% below previous results.
iPhone X iPad Pro 10.5 iMac 27 i7 process, I feel no difference, seems the same to me. I am not going to think about it…
Downloaded the patch to my iPhone X today and I can’t see any lag. I don’t have any benchmark tests but the patch passes the eyeball test for routine us – no see um any slow down.
“ Designed by Apple in California” I think I don’t have to say something else.When somebody becomes big everyone is jealous of him I think that’s the case over here with Apple because they always were a trustid brand. and I hope that they won’t lose their trust unlike uther faulty brands! … that want your money that’s only what they want money money money And they are even ready to play it dirty.
Just ran Geekbench. No Speed differences… neither in CPU nor in Benchmark tests. Benchmark actually scored 8% faster. Battery was half full and the iPhone not connected to power.
Ok, just did the test with iPhone 8 Plus with Geekbench 4.2.1
iOS 11.2.1
Single-Core 4244
Multi-Core 10493
Compute 15240
iOS 11.2.2
Single-Core 4158
Multi-Core 9154
Compute 17777
Not sure why Metal did so much better after the update but performance hit on CPU is well within acceptable if it makes my phone more secure.
Exactly. Thank you so much.
So the throttling is that bad…
The single core difference in davidabrant’s benchmark is 2 percent and the multi core is 12.8 percent. Those numbers will likely improve as code is optimized in later updates. The original article test which is probably bogus saw performance decreases of 41 and 39 percent.
The throttling due to battery…
That is thw difference between Melv1n’s test (old battery) and davidabrant (new battery).
put much stock in the test shown in the article. There are many problems with it.
Okay, I’ll wait for your data…
Isn’t enough tests either.
Here’s mine on iPad Pro 12.9” 2nd Gen
iOS 11.2.1
Single 3935
Multi 9487
Compute 29385
iOS 11.2.2
Single 3918
Multi 9406
Compute 29272
Seems like little to no impact, unless there’s something else going on. This is nowhere near the losses being discussed as possible.
Geekbench 4.2.1
Aww, that’s not good.