U tweetove putem weba ili aplikacija drugih proizvođača možete dodati podatke o lokaciji, kao što su grad ili točna lokacija. Povijest lokacija tweetova uvijek možete izbrisati. Saznajte više
Intel just gave me permission to release the blog on the #AVX Clock Speed Throttle #Spectre. So as promised, here it is - the way I intended to publish it back in June before #NetSpectre was announced: http://www.numberworld.org/blogs/2018_6_16_avx_spectre/ …https://twitter.com/Mysticial/status/1007884805026013184 …
If anyone is interested, I have a few other (long-shot) ideas for Spectre attacks. One of which eliminates the need for any timers. If I get the time, I may write about these as well. But they don't look serious enough to bother going through Intel again.
Nice work! [Sad that Intel always slows the clock. Many enthusiasts install gargantuan cooling systems on their CPUs, presumably enough to remain stable at full speed? This is like Apple arbitrarily reducing clock on older phones even if owner is using an external battery pack.]
The throttling isn't artificial. It's required to keep the chip from overheating or going unstable. At equal clock speeds, #AVX512 can pull more than 2x the power of non-AVX. So either you run at the lowest common denominator. Or you let non-AVX run faster than AVX.
Very nice write up! Did you get any feedback from Intel w.r.t. the resolution order of mispredicted branches?
My interaction with Intel can be lightly described as "unproductive". They never admitted anything about the exploit and tried their best to drag their feet for as long as possible. It was only after #NetSpectre went public did they finally say, "the concept was already known".
Nice work
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.
