Athlon 64 FX-51

[danscript]

I have been reading about the new Athlon 64 FX-51 and I am very impressed. I was wondering if anyone has had any luck running the processor with linux? I assume it would run fine in 32-bit mode (the cpu). The bad thing is not being able to take full advantage of chips abilities.

[ph34r]

I believe that AMD64 is a valid architecture choice when compiling the kernel.... To get full benefit of course you'd want to compile everything from source, so something like gentoo would be ideal.

[danscript]

Thanks for replying. I am used to Slackware. How does Gentoo compare other than having to compile everything from source?

[MorphiusFaydal]

ive used both, and have the latest install disks for both..

and, simply because my network card does not work in linux, i prefer slackware.. not the new 9.1 version, but the 9.0 version..

gentoo is good, of you have broadband (i do), and a working network card, and you know a bot about setting up linux, and what needs to be in some config files..

still..

its preference..

but do you really want to shell out nearly 800 US Dollars for a processor? and thats the price for the OEM chip.. the retail boxed is even more expensive..

yeah..

Chris

[bandwidth_pig]

As ph34r had mentioned, the Athlon 64 is indeed a option when compiling the 2.6 branch kernels. I don't recall seeing it in the 2.4 flavor though. I saw it when compiling the new 2.6 test yesterday.

[bwkaz]

SuSE has an experimental AMD64 version. I think Mandrake does too, but I don't know for sure.

[bandwidth_pig]

Does gcc have support for the Athlon 64? It must if it's in the kernel. I guess that was a dumb question. But...I guess my next question would be, if gcc does have support for it, and you can pass a optimization flag when compiling source for various software, will the software you are compiling have to be written specifically with instructions that would make use of the 64 bit power?

I have thought of taking a more serious look at this processor myself. All started with that bogomips thread that was on JL not long ago if any of you saw it :D

My P4 is running at 2785 bogomips. Whats interesting is that is higher than what I saw previously. I don't understand why that woudl be.

[bwkaz]

Just about the only thing that changes bogomips is processor clock speed (unless you change processor types, like from P4 to Athlon XP; but that's another whole monster). If you've changed your front-side bus at all, that will change the bogomips too.

Basically all the kernel does is spins in a loop doing nothing other than incrementing a counter. It stops when a certain amount of time has passed, and reports the counter value as bogomips (possibly after doing some math on it). The "official" definition of bogomips is "the number of millions of times per second that your processor can do absolutely nothing" -- except increment the counter, of course.

There is (or used to be) a BogoMIPS HOWTO at www.tldp.org -- might be worth a look. It might have been renamed the "benchmarking HOWTO", as well.

Anyway, yes, gcc supports the Athlon 64 (and Opteron) processors. When you configure it, it tries to guess the architecture (or you can tell it specifically what to compile for), and if that architecture is x86-64, then it enables all the Athlon 64 / Opteron stuff. Like the extra optimization flags described at the end of this page:

http://gcc.gnu.org/onlinedocs/gcc-3.3.2/gcc/i386-and-x86-64-Options.html#i386%20and%20x86-64%20Options

The code does not usually have to be specifically written for 64-bit processors. However, some of the assumptions that people make (i.e., that "long" is always exactly 32 bits, as opposed to "at least 32 bits" like the C spec says) are incorrect on a 64 bit platform. So it is sometimes necessary to change some code.

[bandwidth_pig]

Thanks for the info!