5 ms·
Intel's Downfall Mitigations Drop Performance Up to 39%, Tests Show
- pie420 3y agoShame, guess you have to buy the next gen CPU from Intel, which will surely not have any exploits.
- j45 3y agoYikes. Is AMD affected by anything like this at the moment?
- mey 3y agoYes, in addition to the prior zenbleed/spectre series of attacks, inception is the catchy name for the latest round of AMD focused speculative execution exploits https://www.theregister.com/2023/08/09/amd_inception/ https://www.theregister.com/2023/08/09/amd_inception/
- steve1977 3y agoAnd what about ARM architectures? (Apple Silicon etc.)
- porkbeer 3y agoSecurity through obscurity. And mac/arm users are a less valuable demographic to the writers of these exploits (for now).
- belter 3y agoWhat you want to know is ...Who gets the least drop in performance when mitigation's are on. That is a tech blog article I am willing to pay for...
- j45 3y agoFor sure. Also I remember hearing the spectre patches eventually improved (which I should check)
- rkwasny 3y agoUh wait a sec, this is a microcode update? that will just install without any notice?
- JorgeGT 3y agoIf I'm reading this correctly, a Linux kernel parameter `gather_data_sampling=off` will be implemented to disable the mitigation: https://www.intel.com/content/www/us/en/developer/articles/technical/software-security-guidance/technical-documentation/gather-data-sampling.html#inpage-nav-2-2-undefined https://www.intel.com/content/www/us/en/developer/articles/t...
- sznio 3y agoWell, CPU's aren't getting much faster generation by generation, so they've gotta find a new way to make the new ones more enticing.
- sph 3y agoOnce again I ask: 1) are all these vulnerabilities being exploited in the wild today? 2) since everyone is running mitigations, no one is spending much time to exploit them, so why shouldn't I turn them off myself? Every time there's a new vulnerability out, I ask the same question and I still don't have a valid reason why I should keep them on. Herd immunity does not only apply to viruses. In fact, the slower our CPUs get because of these fiascos, the more performance I might squeeze out of something no one is exploiting in the wild, unless I have proof of the contrary. Of course, this works only if I disregard anyone's opinion and do it. If everybody turns off mitigations, then I'm unprotected.
- fswd 3y agoshouldn't we all get a refund check for 39% of the price we paid for a CPU?
- BizarroLand 3y agoSo after 7 generations of improving performance by 3-5% a generation nearly all of those gains have been wiped away by a single bug, and other than the newly supported architecture, instruction sets, core count and gains in raw clock speed, there is essentially no compute difference between a 6700k and a 13700k?
- lamp987 3y agoIntel and AMD need to be broken up by anti-trust. We hit a dead end.
- rasz 3y agoMight be time for per virtual cpu core speculation only cache.
- nightshift1 3y agoProbably should have linked to the source that made the efforts to produce the benchmarks: https://www.phoronix.com/review/intel-downfall-benchmarks https://www.phoronix.com/review/intel-downfall-benchmarks