Israel Inside, Insecurities Outside.
Just to think it has been almost four years since the cover had been lifted upon the fact that Intel Core processors are insecure piles of shit, which I’m sure had nothing to do with single core supremacy from cut corners in branch prediction and execution.
Thinking back they were good times, then Intel CEO, Brian Krzanich sold the maximum amount of Intel shares as legally possible, why? Because Intel’s vulnerabilities didn’t just start and end with the illusive Meltdown and Spectre.
Plundervolt rendered it so now even overclocking is insanely vulnerable along with other siide channel attacks that targeted Intel CPUs with Hyper-Threading enabled. When you buy Intel nothing is safe and secure, with software mitigations coming out years ago now having already somewhat impacted the performance of Intel Core processors while they never were truly defeated until Skylake was finally shown the door.
And now? Spectre is back and it’s fucking pissed.
Intel themselves along with “VUSec Security Research Group“ have revealed yet another Spectre-class speculative execution vulnerability dubbed “Branch History Injection” or BHI.
Branch History Injection affects all modern Intel Core processors, from Intel’s 4th generation “Haswell” processors to their more recent 11th and 12th generations of Rocket Lake and Alder Lake CPUs.
Along with specific ARM processors being affected such as the Cortex A15, A57, A72, Neoverse V1, N1 and N2 being vulnerable.
While AMD Ryzen CPUs are nowhere to be mentioned whatsoever, probably because AMD engineers aren’t cheating kikes.
Though this hasn’t stopped Intel from pointing fingers directly at AMD with the unjustified claims that their mitigations aren’t just ineffective but faulty as well.
While in its infancy, Branch History Injection is more or less a proof of concept attack where vulnerable CPUs are now open to Spectre V2 exploits, while affected Intel Core processors have already somewhat been mitigated, BHI avoids the Intel Enhanced Indirect Branch Restricted Speculation (EIBRS) and ARM ID_PFR0_EL1 CSV2 assignment.
VUSec also reports that Branch History Injection enables cross-privilege Spectre V2 exploits allowing kernel-to-kernel exploits allowing attackers to place predictor entries into the global branch prediction history make kernel leak data. The resulting exploit results in the arbitrary leakage of kernel memory which could reveal sensitive data.
Intel of course plans to release security mitigations for the aforementioned processors Intel Core 4000-12000 series of CPUs being those affected, thankfully however Linux users can sleep soundly as its community has already initiated mitigations against Spectre V2 exploits on affected processors.
The recommendation being to enable Retpolines, along with Intel eIBRS (Enhance Indirect Branch Restricted Speculation) to combat against Spectre V2 vulnerabilities.
And it’s such Linux mitigations that Phoronix has put to the test to yet again determine the overall performance impact mitigating against Spectre V2 would otherwise have on the performance capabilities of Intel Core processors, specifically I/O performance.
Just like when exploit mitigations were first deployed on Intel Core based systems Spectre will once again have a rather dramatic effect when it comes to the performance and I/O capabilities of Intel Core processors which amuses me to no end.
Phoronix having tested the latest MSDT wonderchild, the Intel Core i9-12900K along with the Tiger Lake based Intel Core i7-1185G7, the performance impact the aforementioned mitigations actually have is terrifying for those who pride themselves on supporting mega-kike corporations such as the Intel Corporation.
With the overall capacity and execution for the Rocket Lake based Intel Core i9-12900K dropping massively upon activating eIBRS and Retpolines highlights a staggering 26.7% reduction in SockPerf throughput, though only just a 13.8% drop in SockPerf latency.
Other aspects such as KeyDB, RocksDB, SQLite, LevelDB are suffering performance reductions ranging from 10.4% to 4.5%.
With other aspects from I/O being hit with a 14.5% reduction along with Networking suffering a 9.2% performance hit with the likes of DaCapo Java Benchmarking suffering a minimal but still significant performance reduction of 2.1% while even GIMP image “manipulation” program also suffers a 2% performance reduction. Safe to say that even end users will be noticing the performance hit proper mitigations will be causing.
However if we take a look at the performance impact upon the previous generation Tiger Lake of which Rocket Lake is actually based upon, you can see that the performance impacts are much more severe.
With the Intel Core i7-1185G7 mobile CPU suffering a whopping 35.6% reduction in file creation with I/O performance being hit much harder than Alder Lake with a 34.1% impact with eIBRS and Retpolines.
Further performance hits of 28.1% are seen with PostMark D.T.P and 25.6% on Stress-NG Socket Activity performance figures, at least after that the performance reduction does then lessen to a more calm 10.3% in context switching as the performance reduction from mitigations then drop off but are still present losing anywhere from 9.8% to 2% performance across various tests.
Most interestingly there’s actually a 2.2% gain in performance from the patched machine in Selenium, but for the most part when it comes to Intel Core processors you certainly get what you pay for.
And by that I of course mean with bargain bin prices come bottom line security resulting in hilariously cucked performance.