It will always depend on one's own CPU and usage of the PC. One actually knows it when testing the final release on the own machine, not before. To speak generally of retpoline or any other measure just as IBRS or to compare retpoline to IBRS makes also only limited sense. There will be different combinations and configurations and even variants of retpoline such as: Code: SPECTRE_V2_NONE, + SPECTRE_V2_RETPOLINE_MINIMAL, + SPECTRE_V2_RETPOLINE_MINIMAL_AMD, + SPECTRE_V2_RETPOLINE_GENERIC, + SPECTRE_V2_RETPOLINE_AMD, + SPECTRE_V2_IBRS, For me it is still unclear why retpoline should be less expensive than IBRS. I do not know when exactly there is a restriction....depends on too many factors... Retpoline will have an impact on structure of programms (global change)..... Indirect branches are popular and speculation on these is responsible for performance gain. So now it seems we try to avoid situations where predictions are made and we are changing the code by recompiling (replace indirect callings)... So we are moving into a direction where we try to prevent events while leaving the cause untouched (we're waiting for new CPUs there is no other way).... At the same time the CPU devs actually would have to fight the cause which comes with a complete revision of speculative execution. The right way is to work out a concept with more resources/metadata just like the process context ID. Or bit-setting or flags / tags....to introduce more ways to differentiate. It must be very clear how things are related and which processes can do what...and there must be real barriers and clear privileges. To have the page tables together is not the reason for meltdown (see AMD, KPTI can be turned off on AMD CPUs) and speculation on indirect branches to gain performance can also be realized in a safe way.... But for now -it seems- many people have to do lots of efforts to prevent special situations and when once the cause vanishes they can be all reverted...except Intel does not really want to go after the cause and remains at fighting events where prediction can be abused....means restricting predictions instead of to make them safe. I guess it will take years....
I concur, I have lost much respect for Intel... All things considered, PC's that I build will have High-End AMD CPU's... Intel can "Stick it where the sun never shines"...
You liked the reliability and performance of intel, the better performance part is because of the specific design that now is found to be exploitable and now called a flaw.
dell has suspended many firmware updates that they released in dec/jan as it is causing some rebooting issues so if you installed and working is ok otherwise they posted older versions to revert back to until newer ones posted; they mentioned stick with the microsoft OS patches for now
@MrG: Like Yen said. He's only providing a way to set the registry key that enables / disables Spectre2 protection. Steve Gibson is really smart, but I doubt he'd want to bite into this rotten apple.
My last "high-performance" processor was a first-generation i5 750 after that I started investing in more modest PCs. I currently have a Pentium Dual Core G3240 Haswell and I'm glad I did not spend a fortune on an i5/i7 now I'm needing a quad (or more) core I will certainly go to AMD
Thank you. I'll wait, the upgrade is not so urgent. Even if Intel fixes the bugs in the next generations I see a lot of arrogance on its part. I like the power efficiency of Intel processors but AMD's have not been bad lately.
