Originally Posted by azanimefan
the belief now, is intel will either have to scrap HT, or redesign it from the ground up. and if they scrap HT entirely, they won't be able to compete with AMD at all in the server, desktop or even laptop markets.
-BTW: replacing/redesigning hyper-threading is something they should have been doing since Ryzen released and it was clear AMD's version of SMT was more efficient and produced better results then hyperthreading, a tech that intel hasn't changed since the P4. When HT came out on the P4 it gave 15-25% performance uplift in HT friendly environments. Which is identical to it's current, modern day performance. AMD's SMT gives somewhere around 20%-35% performance uplift in SMT friendly environments. Intel is getting screwed hard for their lack of innovation. I mean look at these vulnerabilities, they're persistent across over a decade worth of CPUs due to design similarities in their core i lineup. some of these vulnerabilities were identified by Intel's own engineers when the first design concepts for the core chip design were debated and (ultimately) scrapped in the 90's. Not enough innovation Intel!
Why would they need to do that? The problem is in the cache and register file no? They simply need to implement proper checks in those, not re-design hyper threading itself. Though yes it should be done for a performance boost, but not for mitigating these security issues. Its possible that with a redesign of the register file and cache that we would get the performance boost in HT though is they also increase size and associativity for the core.