While I do shit on it a lot, it is what we often say: It's not all bad, it's only mostly bad.

After all, x86 is one of the most polished turds on this planet, but, sure, it has some.. Decent parts, right?

It's not like Intel focused, since day 1 (with the 1101 and 3101), on delivery time above all else, which left it with many design flaws, right? Or the fact that Intel copied Sun Microsystem's homework on TSO (memory ordering), and even to this day, refuses to elaborate on the implementation details, right?

Alright, alright, x86 is, at the very least, a usable instruction set and platform. While the R&D hasn't been very high vs. its early 2010 days, let's hope Pat puts a little more effort than the last CEOs.

In addition, Hybrid was a huge hack into the architecture, which left CPUID.1Fh leaves being different on P and E cores... And even then, no SMT and no AVX-512 for the E-cores (Atom), despite both being present on the Atom cores for the Xeon Phi in 2012. (There's at least X86-S coming, and that was overdue. So, yay?) Paying a subscription for SGX (Intel OnDemand)...

Despite all that, and my personal experience (do not write an x86 decoder up to AVX-512 with EVEX and MVEX prefixes, I'm warning you), it still has a place in consumer, commercial, and scientific markets... For compatibility reasons, mainly (many conditions apply).

x86 doesn't deserve to die.. At least just yet.