Linux live usb drive: "Live" vs "Live 686"?

[jaydee]

I was wondering what the actual difference is between the options "Live" and "Live 686" when booting up to a live linux USB drive (Debian 6 for instance).

I don't have anything to really go by, but my assumption is that "Live" is designed for CPU architectures up until x586 (Pentium II or K6 class), and "Live 686" is designed for Pentium III/Athlon or newer CPUs, is this correct?

 

[jaydee]

bump

 

[lxskllr]

I assume you're correct. I didn't answer cause I don't know for sure, and I've never seen that. I'm guessing it gives you the choice so time isn't wasted loading obsolete instruction sets.

 

[Cerb]

The 486 is basically just there for older and/or incompatible CPUs or chipsets. I'm pretty sure it lacks SMP and PAE support, along with huge pages.

While many older CPUs actually run the 686 kernels fine, it's becoming an optimization issue, since more modern CPUs benefit from different instruction mixes and orderings, and while they support the old instructions, they can run them slower directly, microcode them, etc.. Overall, it probably doesn't matter for >99% of computers and users, but if the <1% it does matter for can show compatibility or performance differences...

(Pentium II or K6 class)

Not quite: 486, Pentium, K6, and VIA/Cyrix 6x86 type CPUs, for the now-common i486 kernels. 386 and up for i386, obviously (other 386s were direct clones, while some 486-class CPUs were separately developed, much like later CPUs)

The Pentium II is the very definition of a 686. The Pentium Pro, Pentium II, Pentium III, and Celerons prior to Netburst, along with the Pentium-M, Core Solo/Duo, and their derived Celerons, are all definitively 686 processors, being based directly on the P6 core, and all supporting only minor ISA revisions over the years, from the PPro (the Core 2 still had P6 bits and pieces, but I think x86-64 is a good cut-off). Also, AMD's Athlon and Duron lines, through the Athlon XP/MP, would be considered 686-class, along with P4s.

 

[jaydee]

Cerb, can you be a little more clear? In your first sentence, did you mean to say "486" or "686"?

My understanding was, that we are still in the "686" generation of CPU architecture, is that not the case?

 

[Cerb]

Cerb, can you be a little more clear? In your first sentence, did you mean to say "486" or "686"?

Both. 386, too. 486 are pretty common for compatibility kernels, and are the standard going forward. The Linux kernel itself has dropped official 386 support, either with 3.8 or 3.9 (fairly recently, so Debian could get another 386-supporting version out ).

My understanding was, that we are still in the "686" generation of CPU architecture, is that not the case?

Hardware-wise, Nehalem pretty much got rid of the P6; and IIRC, somebody from Intel said that SB truly had nothing left of the PPro's design, when it was coming out. The P4 supported all the same instructions, but was nothing like it.

Software-wise 686 is still basically the latest as far as 32-bit goes, because nobody wanted to optimize for P4s, especially once saner CPUs started coming from Intel, again.

x86-64, however, isn't, software-wise. It has 8 true GPRs, requires SSE2 and PAE, and loses the ability to use real-mode addressing.