digg

Join DiggAbout

Discover the best of the web!
Learn more about Digg by taking the tour.

64-bit: More than just the RAM

bit-tech.net — "Time and time again, self-proclaimed gurus determine that the only real difference between 32-bit computing and 64-bit computing is the memory limit. Are they right that RAM is a reason? Definitely - but that's missing about 99 percent of the true differences."

Bury
show profanity settings
expand all
  • Error601, on 10/18/2007, -12/+55My favorite line: "And aside from 'nix operating systems, nothing supported true 64-bit computing until Microsoft's lacklustre attempt at re-doing Windows XP in 2005."

    So, if you exclude everything but Windows, nothing supported it until Windows XP in 2005.
    • aliengoods, on 10/17/2007, -18/+4Where's the love for 64-bit OS X?!
    • Rekutyn, on 10/17/2007, -1/+13*hits head*
    • VAXcat, on 10/17/2007, -3/+15And what about 64 bit VMS on the DEC Alpha, years earlier...
      • rootstyle, on 10/17/2007, -1/+10VAX, It's dead Jim.
        • andywebb95, on 10/16/2007, -1/+2VAX architecture is dead... VMS lives on (Alpha and Itanium).
      • fluxion, on 10/17/2007, -2/+6stop digging him down you 'tards
      • andywebb95, on 10/17/2007, -1/+4Was going to say the exact same thing.
        And by years earlier it was QUITE a few years earlier (1992 I believe when the 64 bit Alpha and its supported build of VMS came out).
        I don't use VMS at all these days but it is a splendid operating system and a joy to develop for.
    • av4rice, on 10/17/2007, -2/+19"So, if you exclude everything but Windows, nothing supported it until Windows XP in 2005."

      So if you exclude everything except for Windows, all you're left with is Windows? NO WAY
    • andywebb95, on 10/16/2007, -1/+4Yes it was a pretty bonehead line on the authors part considering there were at least two 64 bit operating systems out there (VMS and BSD?) that I am aware of available prior to 2005.

      Ironic also because he was pointing out other peoples bonehead comments.

      But this article was written for the masses..... and not everyone uses VMS or BSD..... so I could see the point of the exclusion for simplicity purposes.

      • init100, on 10/17/2007, -1/+4"there were at least two 64 bit operating systems out there (VMS and BSD?) that I am aware of available prior to 2005."

        I'm pretty sure that Linux was also 64-bit capable before 2005.
        • devikwolf, on 10/17/2007, -0/+4BSD = *NIX
          Linux = *NIX
        • andywebb95, on 10/17/2007, -0/+1I am not sure....
          If there were builds available for 64 bit processors prior to 2005 I would have to say it was available.
    • mindframe, on 10/17/2007, -1/+5Sun had Solaris on UltraSPARC that was 64bit in 1995. Solaris 2.5.1 I think.
      • daridave, on 10/17/2007, -1/+2yes, and he excludes OSX oses from his analogy. I guess we're getting dugg down by the comment poster, probably refusing to accept his comment was wrong.
      • MacSuxWindozSux, on 10/17/2007, -3/+2Yes he absolutely knows OSX is so much more awesome and he can't stand it...

        Get over yourself.
    • daridave, on 10/17/2007, -4/+7Idiot. It states "aside from 'nix operating systems, nothing that the average consumer used supported true 64-bit computing". Note the "nothing that the average consumer used " . Very important difference and that DOES make sense. Learn to read, smart ass.
      • Error601, on 10/17/2007, -0/+3There's always the ***** in the crowd. The OS doesn't have to be installed on the computer in front of you for you to be using it. We have this fancy concept called networks.
    • daza, on 10/17/2007, -0/+1In fact, Windows XP 64-bit edition came out in 2001. This was the IA64 edition designed to run only on Itanium processors. In 2005, Windows XP Professional x64 Edition came out that supported EM64T and AMD64. I have no idea how I remember this, given I was 11 in 2001, but it is true.
    • Myztry, on 10/17/2007, -0/+1That's like saying:
      "aside from the 1985 Amiga, no mainstream Operating System provided a true 32bit multitasking GUI environment, until Windows 95"
  • Dokument, on 10/17/2007, -5/+24ha. it has 32 diggs currently..
    • toast1226, on 10/16/2007, -2/+10lol when i looked at it, it was at 64 diggs. what a coincidence xD
      • santa7, on 10/17/2007, -2/+7Well I saw 128 diggs, I'm looking into the future
        • TomHellier, on 10/17/2007, -2/+0It's 241 now (lost the point)

          Hey! Close enough to 256
        • ubergeek09, on 10/17/2007, -2/+0A 128 bit processor and Operating System? Now that would be awesome >:).
    • BigglesPiP, on 10/16/2007, -1/+2wheras I actually did see 128, it says 129 now, but only because I dugg.
    • Jakerzon, on 10/17/2007, -0/+1haha, now its 512 diggs. Data from Star Trek uses a 512 bit positronic brain-processor. lol jk
    • sfrederiksen, on 10/17/2007, -0/+1And now 619! Ha!
  • ravan46, on 10/16/2007, -9/+2Looks interesting, digging as a bookmark to read later...
  • wilsonthecat, on 10/17/2007, -0/+4264 bit machines have bigger registers. There saved you 5 pages of article. What's a register? Ok you might need to read another 5 page article to find that out.

    • AJH16, on 10/17/2007, -4/+16A register is a small piece of memory located on the actual CPU that is very small and useually holds a single word (64 bits in the case of a 64bit processor). Having larger/more registers means that the system can move faster because it spends less time waiting for data to enter the processor. There, saved you another 5 page article.
      • codehkr77, on 10/25/2007, -5/+13pheww...for a minute there... i thought register was where i paid for *****
    • fluxion, on 10/16/2007, -1/+1and relative pointers? lost me on that one.

      i would've thought 32-bit would need relative pointers do to lack of addressing bits in a 32bit instruction (but im thinkin in terms of JMP instructions and whatnot, which i believe do actually use relative addressing...but those arent actually pointers), and that 64-bit would allow for absolute. any fundamental reason 64-bit is required for relative pointers?
      • SJKat, on 10/17/2007, -4/+0You can has pointer in a separate word? lolz.

      • init100, on 10/16/2007, -0/+2I was also thinking about this. I'm 100% sure that x86 can use relative addressing. It is a prerequisite for position-independent code, which is used in Linux (and probably other *nix) shared libraries.
        • geminitojanus, on 10/17/2007, -0/+4You can actually fake relative addressing with absolute addressing at the architecture level, all relative addressing at the processor level does it remove those extra "faking" steps, which speeds up code execution for PIE/PIC.
      • Urusai, on 10/16/2007, -2/+28086 and onward all had segment registers (called "selectors" in 80386 and up). These set the base for your code and data segments. Apparently, nobody was using these selectors once things moved to 32-bit and was just setting them to 0. I don't know, but it sounds like they added some new registers with the exact same function as the old registers, except now the expect people to use them.
    • TeacherOfHeroes, on 10/17/2007, -0/+2Not only would the registers be 64 bit, but when AMD created the standard, they upped the x86's 8 registers to 16, which means less processor stalling while waiting for data to be fetched from cache.

      32-bit code would be compiled so as not to rely on these extra registers.
  • felderado, on 10/16/2007, -13/+2x86_64 processors are not TRUE 64bit processors

    64bit has lower performance vs 32bit under certain circumstances, but generally should not be noticable by any end users

    64bit has lots of nice new features though... the memory limit is one of them. I have 4GB in my machine right now. It has 8 memory slots and can hold 16GB. Someday I'll buy the rest of the ram.
    • santa7, on 10/16/2007, -0/+4My computer supports 8GB of ram and has 2 slots, drives me insane trying to find this 4gb RAM at a cheap price
    • maninblac1, on 10/16/2007, -0/+2x86_64 processors are true 64bit processors, or else you couldn't run 64bit code on them, usually the definition of being 64bit.

      64bit's perfromance drop is not at the processor level but at the system level. Simply because of primitive data bloat. Obviously an operating system that uses 500MB of memory in 32 bit and 900MB of memory at 64bit will perform differently on a 1GB or 2GB system.

      Just tell yourself, my computer hardware can move Data rate X (usually 50MB/s from your HDD) when using 32bit values, it will move half of that data if that data is 64bit. That's assuming you have to move data cold, the processor is fetching data from cache at 98-99% success, from memory 70% of that 1-2% remaining and HDD the rest of the time, regardless of data size when warm, so the amount of time you're really running "half speed" is very very small.
    • trogdor282, on 10/16/2007, -0/+2Uuuuuuh.. Yes they are true 64 bit. The reason 64 bit is slower in some cases is a 64 bit value takes up twice as much memory bandwidth as a 32 bit value. 8 bytes instead of 4. Luckily AMD was able to throw more registers into the new ISA to balance out this effect. A 64 bit processor with the same number of registers would be much slower than its 32 bit counterpart, most of the time.
      • felderado, on 10/16/2007, -0/+1They are not true 64bit processors. They do not let you have an address space of 64bits. Only 40 for AMD64 and 36 for Intel with PAE! Therefore they are not TRUELY 64bit.

        They ARE slower in some examples. Case in point:
        "If you want to see what happens when you go from 32 bit to 64 bit, use SPARC v9 as your model, not x86. The 64 bit part of amd64 is largely irrelevant; amd64 CPUs are massively faster despite being 64 bit because of all the other additions made by AMD."
        "Take, for example, SPARC v9 or many of the MIPS CPUs. 64 bit is slower for two reasons. Firstly, loading immediate addresses is over twice as slow, since more load and shift instructions are required to get the address into a register. Secondly, the code is larger, which leads to worse cache performance."

        Example of how amd64 processors are faster:
        -AMD64 CPUs have intigrated memory controllers which greatly improve memory bandwidth.
        -AMD64 CPUs have added features which give them an advantage over standard 32bit x86 processors.

        The x86 processors should have never become mainstream. We need competition in our PCs. We need a processor that doesn't have such crappy low register counts.
        • mtekk, on 10/17/2007, -0/+2Intel tried to fix this with the Itanium, but it went over budget, was released too late, and has not made enough strives (ran hot and was slow and is difficult to write compilers for) to make it worth while to adopt in consumer PCs.
        • trogdor282, on 10/17/2007, -0/+1Look up "superscalar processor". Processors haven't had a fixed Instructions Per Clock count in years. Modern processors keep a window of instructions that are ready to go. Each clock cycle it may execute no instructions, one instruction or maybe even 8 instructions. It doesn't even execute them in order, although it does some elaborate ballet to make it look like that. Furthermore, the processor maps dozens of virtual registers to each visible register. So that's not even much of an issue. x86 may be very ugly to look at, but it doesn't cause any performance problems that haven't been solved years ago. I wish people would stop complaining about it.

          If you were right, PowerPC would kick x86's ass all over the place. But it doesn't. The reason is that choice of instruction set isn't actually all that important anymore.
    • SteveMax, on 10/16/2007, -0/+1The code you are running is either more appropriate to 32 bit computing (like trogdor282 explained) or has an optimized ASM path for x86 (and not for x86_64).
      For what I run (basically, particle physics simulations and analysis), s86_64 is a blessing. A simple operation, such as a large matrix diagonalization, is at least twice as fast compared to x86 (on the same computer), due to the extra registers. The extra precision is a nice bonus.
      Summing it up, just because x86_64 gave you lower performance under some load, don't assume that it's always the case.
    • felderado, on 10/16/2007, -0/+1I'm not saying the amd64 processors are totally junk, I'm just pointing out that because they are 64bit doesn't mean they are faster in every single way. There are situations where there are disadvantages when processors move from 32bit to 64bit. AMD64 has some, but it has a lot of improvements that make them unnoticable or make up for it.
  • Section1, on 10/16/2007, -14/+1 Awesome story IMO. I can't tell if it is over my head or not, I will reread when I don't have a headache. I can't wait to understand 64 bit better!

    Hail Digg!

    S`1
    • Section1, on 10/16/2007, -6/+1Why The hell do I keep getting thumbs down whenever I post a comment?! sorry I am off topic, but wt eff? am I doing something wrong, or are people just being stupid?

      S`1
      • rootstyle, on 10/16/2007, -0/+3"or are people just being stupid?"

        Not an accusation I would make while at the same time putting a pseudo signature in all my Digg comments. We all see your user name at the top of all your posts, its ok. That and your original post has absolutely no substance, it is essentially 'awesome article!!!11' which could have better been expressed by simply digging the article and moving on.
        • Section1, on 10/17/2007, -3/+3Fair enough. I was excited by the story, and wanted to be encouraging to the poster, and S`1 is my name in every online thing I do, and it is a habit for me to tag all my posts with it, but I see how it come across as affectation . I suppose I am trying to learn how to use digg properly still.
          "or are people just being stupid?" was probably not the best way to pose a question, but I was aggravated at the time. so sorry for that. I will stop signing to the best of my ability, it sounds like bs, but it is honestly habit.
          thanks for the response rootstyle I see how the the S`1 is off putting and I will try to keep my comments more on the substantial side.
          • Section1, on 10/17/2007, -0/+1hmm. getting dugg down still . .. lol
    • chrismgtis, on 10/16/2007, -1/+1It could be the way you type?
  • soulpunisher, on 10/17/2007, -4/+19now if only we can get people to write software to work with 64bit code instead of 32bit like most still are...
    • StuffOfInterest, on 10/16/2007, -1/+5It is all a matter of crtical mass in the market and external drivers like the memory crunch. The 80386, the first 32-bit processor in the X86 line, came out in 1987. Windows NT didn't come along four 4 years and Windows 95 didn't come out for 8 years. At some point we'll likely see a full scale shift over to 64-bit for all major software and in the space of one or two years you will have more trouble finding 32-bit drivers for new hardware than finding 64-bit drivers. Just try and find a 16-bit driver for Windows software today.
      • archlich, on 10/16/2007, -0/+5I don't want to sound like a fanboy.... The critical mass is for windows. There already is full 64bit support in all the major linux distributions.
        • maninblac1, on 10/26/2007, -1/+2Except drivers.
        • init100, on 10/26/2007, -0/+3@maninblac1

          What drivers? I guess that you must mean proprietary drivers, since the ones distributed with the kernel work equally well in 64-bit as in 32-bit. But even some proprietary drivers are available in 64-bit versions, such as the nVidia drivers.
      • maninblac1, on 10/16/2007, -0/+4Exactly, just like DDR and DDR2, AM2 moved to DDR2, and DDR2 hit 51% of the market, then DDR basically disappeared.
    • trogdor282, on 10/17/2007, -0/+6Most things just take a recompile to go 64 bits. Probably with the exception of drivers. And native 32 bit programs work fine on a 64 bit OS. Wherever people need 64 bit, they've already moved up to it. Servers and such. When the average desktop is pushing 4-8 gigs of ram, the average desktop will be 64 bit.
      • maninblac1, on 10/16/2007, -0/+2Not really, as long as programmers are using bad coding (in C such as assuming integers are 32bit and using generic pointer math) those programs will crash and burn.
        • init100, on 10/17/2007, -0/+4"in C such as assuming integers are 32bit"

          The size of integers do not necessarily change with a switch to 64-bit. But I agree that one shouldn't assume too much about the sizes of various data types. It's better to use the sizeof() operator.
        • Ademan, on 10/17/2007, -0/+1Hrm, if only they created a keyword for determining the size of a variable in C!
          *cough*
          You'd have to be a HORRIFIC programmer not to use sizeof(), i mean i learned about that in my first ever compsci course...
          • HalfBrian, on 10/17/2007, -0/+2But with many new programmers learning Java in their first CompSci class then thinking they are awsome and going to learn a new language, they gloss over some of the finer points of the new language.
      • Shootfast, on 10/17/2007, -0/+2Well if they weren't using 64 bit with 4-8gb of ram, they wouldn't be addressing half of their RAM
    • jgtg32a, on 10/17/2007, -2/+15My personal opinion is that vista should not have come out in a 32bit version, but thats just me.
      • dsn0wman, on 10/17/2007, -4/+5It probably shouldn't have come out at all. At least for another year or so ;-)
      • MacSuxWindozSux, on 10/17/2007, -0/+4Well Back in February Microsoft offered people to sign up on their website to have free Vista x64 DVD's mailed out to anyone who bought Vista x32 or vice versa.
    • grumpyrain, on 10/17/2007, -0/+2It really doesn't matter whether you are programming 64-bit or 32-bit for most high level code. Wow64 ( http://en.wikipedia.org/wiki/WOW64 ) will take a 32 bit process, use the underlying hardware for AMD (which supports both instruction sets) or emulate 64 bit calls in software for Intel. Low level and driver code is the main slowdown, but is generally available for 64 bit. High level software will not commit to solely to 64 bit unless it needs that much RAM or the last 32 bit only processors are 5+ years old.
  • codyman, on 10/16/2007, -3/+6I use xp x64 sp2 on my laptop and it works great (amd64 3400+). I've been using it as it's daily OS for 2 years now, and haven't run into any problems (except for the mentioned driver issues... which aren't all that bad because my laptop has drivers for everything fortunately.)

    Overall, the system is a lot more responsive and even though it is an older machine, I find it faster than a lot of these new laptops (especially vista crippled ones...) in even things such as boot up times / desktop response / app loading times / etc.
    • Abomonog, on 10/16/2007, -0/+1Hmm. I had stability issues on an amd64 3000+ early on. Been running 32 bit since. Maybe better support now.
    • maninblac1, on 10/26/2007, -3/+6Don't even talk about laptop hardware. The hardware that is out right now for laptops, is only incrementally better than it was 3-4 years ago. Pentium M's can still run circles around most apps, because laptops ares a crippled platform to begin with.
      • rootnik, on 10/26/2007, -0/+5You are kidding, right?

        I saw a laptop the other day that has 1600MHz FSB. Mobile dual core CPUs make today's laptop faster than most desktops 3-4 years ago, let alone laptops. Lets not begin to talk about the speed of ddr2 ram compared to the pc2100 that was in most of your laptops 3-4 years ago.
    • codyman, on 10/16/2007, -0/+1I am using it on a Compaq R3000 w/ Geforce4 440Go 64mb version... I am only reporting my own experiences and so far they are great...
  • BigglesPiP, on 10/17/2007, -5/+19Theory is only theory, real world applications show a negligible improvement running 64bit over running 32bit, discussion over.

    The 1st noticeable advantage of 64bit to the end user will be... the ability to use more than ~3GB of RAM.
    • lambda, on 10/17/2007, -10/+5You're just wrong. Discussion not over.
      • maninblac1, on 10/17/2007, -1/+11He's not wholely wrong, but let's face the facts, the average user isn't doing 64bit complex vector floating point calculations, doing scientific calculations in matlab or things like that. 64 bit's performance advantage over 32 bit is in math computation and then if and only if the values can not be contained in the 32bit number space.

        64bit will come into light not because of a performance advantage, but as a necessity as programs get bigger. The only apps now that are 64 bit are (games big memory users), scientific/CAD apps (matlab etc) (big memory users) etc, as you see it's more necessity over anything else.
        • init100, on 10/26/2007, -0/+8I'd guess media processing applications (media encoders, media decoders, image processing software, etc) could take advantage of 64-bit processing.
    • fluxion, on 10/17/2007, -1/+6and relative pointers! man those are awesome!
  • CATSCEO, on 10/17/2007, -6/+164-bit? Wheres my 128-bit?
  • lambda, on 10/22/2007, -3/+8A 64-bit mode OS lets you address 2^64 addresses, e.g. terrabytes of RAM
    A 64-bit mode process can use all 16 general purpose registers (only 8 can be used in 32-bit mode), improving CPU performance for most workloads
    • geminitojanus, on 10/18/2007, -3/+5Except a huge amount of the performance gain is lost to having to carry around 64-bit pointers to everything.

      While it's definitely faster, unless you're running a cluster or super-computer there's not much reason for it. As Microsoft has proven twice now, they're incompetent in retooling their OS to use x86-64 effectively, and the Linux guys simply aren't interested (while there are 64-bit builds of lots of pieces of software, thanks to the portability-by-design aspect of their code, it just doesn't show the performance gains enough for people to actually focus on hardening 64-bit software specifically).

      x86-64 might have gone over better if they extended it into a "short" VLIW architecture (being able to do packed, mixed operations in general registers), but in its "let's just stretch everything out" form, it's currently a non-starter.
    • grumpyrain, on 10/17/2007, -0/+1> A 64-bit mode OS lets you address 2^64 addresses, e.g. terrabytes of RAM
      Actually Exabytes of RAM. (over 18 million terabytes).
    • noname202, on 10/17/2007, -0/+0Actually, the implementation used today only utilizes 48 bits for memory addressing. But I guess 256TB RAM should be enough for anyone ;P
      The architecture does however in theory support 64 bit memory addressing, so the missing 16 bits can safely be added in future CPUs.
  • whitesaint, on 10/17/2007, -2/+7Correct me if I'm wrong, but isn't Mac OS X Leopard going to be 64-bit native on systems that support it, and then switch back to 32-bit mode when it runs 32-bit apps?
    • lambda, on 10/17/2007, -0/+14Correct, except it doesn't have to "switch back" anything. x86_64 CPUs can execute both 32-bit and 64-bit processes in 64-bit mode (e.g. when running a 64-bit OS).
      • whitesaint, on 10/17/2007, -1/+3Sweet. Thanks bro. So probably all the main applications that ship with Leopard are going to be 64-bit as well. Can't wait. :]
      • hoyanf, on 10/17/2007, -1/+0Guess lambda you're only seeing x86 platform and fails to realize the ppc platform... Leopard will be one of the most amazing OS ever as 32/64 & ppc/x86 platform... If you see in depth with current Tiger its already ppc32/ppc64/x86/x86_64 thats 4 architectures in a single install...
    • JoeSobieski, on 10/21/2007, -2/+1No. The kernel in Mac OS X Leopard is only a 32-bit process. In fact on x86 the lower 4 GB of an applications address space is where the kernel is mapped in.
      • SpacePirate, on 10/21/2007, -0/+1Last I checked, the kernel in _tiger_ is already 64 bit. Now, however, all of the Cocoa libraries will have native 64 _and_ 32 bit support. In short, recompile, see benefit.
    • fr34k5h0w, on 10/17/2007, -0/+1What all processors will be considered 64 bit? Core2 Duo's and Xeons? Or just Xeons, as the Leopard preview site only mentions Xeon processors.
  • TheSwashbuckler, on 10/17/2007, -1/+7"Well, historically, the x86 architecture was finalised with a 16-bit word length for the 8086. This meant that every instruction, every piece of data, and every memory address had to be exactly 16 bits."

    Well, that's a load of crap. Plenty of instructions were a byte.
    • init100, on 10/16/2007, -0/+2There were quite some crap in the article. Other crappy points were the stuff about "relative pointers", the talk about registers as a data size instead of as a memory, the bull about NX being controversial for some alleged use as DRM, etc.
      • geminitojanus, on 10/16/2007, -0/+1The latter bit actually was a concern; if someone implemented a Trusted Computing system with x86, the NX bit would virtually assure that it couldn't be broken (as the paramount way to exploit a system is by overrunning a buffer somewhere that gives you the ability to execute code). Of course, there are a lot of steps down the road that would have to be put in place, but around the time the NX bit was added was around the time of Microsoft's Palladium scare, and people were nervous thinking that this was the reasoning for it. So yeah, it's FUD if said today, but in a historical context, it's sensible.
  • zephyrprime, on 10/17/2007, -0/+20Let me condense the whole article for people: 64bits is a little faster. It supports more memory. Some other minor niceties. Also, this guy is wrong about programs running in their own virtual space only being possible with with 64bit tech. Separate virtual address spaces have existed for a long time on 32bit machines. This guy really is not qualified to write this article if he doesn't know something simple like this. Also, ""Well, historically, the x86 architecture was finalised with a 16-bit word length for the 8086. This meant that every instruction, every piece of data, and every memory address had to be exactly 16 bits." Wrong. Many many instructions and data items on x86 are more than 16bits big.
  • anarchytv, on 10/22/2007, -1/+364bit Firefox build: http://www.mozilla-x86-64.com/download.html
    • Trixrox, on 10/16/2007, -0/+2Which is sweet, I use it, but it is weird living without flash...
      • nontitle, on 10/16/2007, -0/+1Gnash, i'm not sure if it's available for windows though.
        Also, I was able to run flash on a 64bit linux computer, it had to go though a ton of emulation things though.
    • TnTBass, on 10/16/2007, -0/+1Trixrox nailed it perfectly. 64 bit browser, no 64 bit flash. Which I don't notice unless I'm looking for youtube videos, or certain apps within Facebook.
      I actually much prefer the 64 bit Firefox (for windows) from this site: http://www.vector64.com/WindowsBuilds.html
    • brownr21, on 10/17/2007, -6/+1Vista 64 comes with 64-bit IE, so no need to use that rubbish.
      • ubergeek09, on 10/17/2007, -2/+2Are you joking? You use the sad excuse for a browser that is Internet explorer and then say Firefox is rubbish? Ignorance is bliss I guess..
  • iwblj, on 10/17/2007, -4/+2Lame. The main reason to prefer a 64-bit system is that individual files are now commonly larger than 1GB. Large files can be accessed much more efficiently if you can memory map them into the process address space instead of using traditional read/write. This approach to I/O requires having a big enough address space that you can comfortably work with the available files. The 4GB RAM limit was a red herring thrown out by vendors that did not have the best interest of their customers at heart.
    • bobbles, on 10/17/2007, -0/+1Yes I'm sure it had nothing to do with 90% of the worlds computers being 32-bit systems.
      • iwblj, on 10/18/2007, -0/+2lol. Pretty gullible, huh? There's no law that says you need to use a 32-bit processor when you have less than 3GB of memory installed. You could use a 64-bit processor with only 1GB of physical memory and still potentially outperform a 32-bit system with much more memory in certain scenarios (for example the memory mapped file i/o). The dominance of 32-bit systems was fine with microsoft and intel because it allowed them to partition the server and desktop markets and maintain a higher price differential from the resulting market segmentation. If AMD hadn't come along it would be 32-bit on the desktop and Itanium on the server.
  • Wosat, on 10/17/2007, -3/+5This article fails because its hypothesis is faulty. When someone asks on a forum what the differences are between 32-bit and 64-bit, they're wanting to know what the differences are *to them*. Do they care that they'll be able to do operations on 64-bit words with a single instruction? No! If I ask what the differences are between front wheel drive and rear wheel drive on Car Talk, I'm not wanting to know about the history of drive systems or the differences in constructing them, I want to know how it affects me AS A DRIVER.

    If the article description and introduction weren't so misleading, it'd be an ok article, but instead it starts with the idiotic statement that 99% of the difference is NOT RAM, then talks about some history, some inconsequential speed advantages of 64-bit, some inconsequential speed disadvantages of 64-bit, the problem with drivers and support (already mentioned by their "self proclaimed guru"), and RAM. All that, and they didn't even mention memory mapped file access.

    Anyone curious about this kindof stuff would be better off reading ars technica.
    • jtatum, on 10/16/2007, -0/+2Exactly. And, continuing with that line of thinking, the main driver for 64-bit uptake in the consumer space is going to be the increased demand for RAM. So, when someone posts on a forum that 64-bit is for additional RAM support, their answer is pretty valid to the consumer that asked. Additionally, if the same poster recommended avoiding it, that recommendation is also supported by the article itself (see page 4 about driver issues).
    • MacSuxWindozSux, on 10/16/2007, -0/+2"Do they care that they'll be able to do operations on 64-bit words with a single instruction? No!"

      I was always told, and hoping for a "Performance Increase". Details were light and I was patient.
      Seems multi-core was the real performance boost.

      Breaking the 4GB ram limit will effectively push the world into 64bit processing. Windows Vista and Windows XP x64 are the solution for 9/10 people.
      Although some Linux migration would certainly be welcome.
  • jazzmichael, on 10/17/2007, -2/+1A 64-bit CPU can access far more system RAM (memory), as much as 18 exabytes (EB). That is, 18 quintillion bytes of data. This can't happen to 32-bit.
    • combatchuck, on 10/16/2007, -0/+1From what I've read, there isn't enough matter in the solar system to make 18 exabytes of memory.
  • Topher06, on 10/17/2007, -3/+2Yes, this is a lame invalid article. For the average enduser, 64 bit is unecessary. Consider that there is still not a lot of support for 64bit platforms from various software and hardware vendors, the only thing the average end user will experience more of with the 64bit platform is headaches. 64bit solutions are clearly targeted for the server market and high end workstations where they process or handle files and memory larger then the 32bit barrier allows, its not intended as an upgrade over 32bit, but to target a different market.
    • ubergeek09, on 10/16/2007, -0/+064 bit operating systems should be more supported by the geeks. We need to buy or use 64 bit operating systems so that developers will write more applications that take advantage of a 64 bit operating system and hardware. If developers did this it could help give a big boost in speed with hardware people already have but aren't taking advantage of.
    • mooninite, on 10/16/2007, -0/+1The computer industry is meant to EXPAND all the time. Why is it always "unnecessary" when something new comes out? Yes we *NEED* 64-bit on Joe Smoe's desktop. Otherwise... how are we going to invent AI? :P
    • MacSuxWindozSux, on 10/16/2007, -0/+1By 2010 you are probably going to see a lot more 64Bit systems, by 2015 they will be very common.
      All you need to do is check ram usage on average PC's. If you follow the curve the average PC will hit 4GB in a few years.
  • Wosat, on 10/17/2007, -0/+7in case anyone wants to swim in the deep end:
    http://arstechnica.com/articles/paedia/cpu/x86-64. ...
  • ortucis, on 10/17/2007, -3/+1Here's the summary for those who don't want to RTFA or just don't care to know > 64 bit is better. Now go home or do something else.
  • RicktheBrick, on 10/17/2007, -2/+1The nineties saw memory increase from 1 Mega bytes to 1 giga bytes. There has been very little increase in memory since. If the pace had continued there would be a 256 Giga bit chip by now. If a home computer had 256 giga bytes of ram memory people would demand software that used that memory but since system with more than 4 Giga bytes of memory are rare there is very limited demand for that software. I do not understand why this has happened since manufactures have reduced size from 140 nanometers to 65 nanometers today and Intel will produce 45 nanometers this year and further reduce that to 32 nanometers in 2009. 64 bit computing will not catch on until there is a large increase in the size of memory sticks.
    • MacSuxWindozSux, on 10/16/2007, -0/+1Average memory allotment in 2000 was only about 256 mb. Maybe still 128 mb.
  • axisds, on 10/18/2007, -4/+2digg needs more of quality articles like this. This is what digg is about... minus the corny subject lines (in the article)
  • ubergeek09, on 10/16/2007, -0/+0Good article, so many people always seem to think the only thing 64 bit is good for is RAM.. But if more programmers would write games and applications that would take advantage of 64 bit hardware, then it would really be more noticeable.. I hear that Microsoft will make it's next OS 64 bit only which is really a good move, it forces programmers to write applications with 64 bit hardware in mind, which forces programmers to finally catch up with hardware..
    • Wosat, on 10/17/2007, -0/+2I beg to differ on a few points. First off, a 64-bit OS does not force developers to do anything (unless they write drivers). Second, games have far more to gain from exploiting multiple cores and SIMD operations than they stand to gain from moving to x86-64. The reason for this is that so much of game code is made up of branches and operations on single-precision floats and integer values that fit in a 32-bit word that much of the speed increases realized in the move to 64-bit are offset by the bloat that results. The benchmarks bare this out -- not much speed increase to be had.

      I do agree with you that 64-bit is about more than RAM, and that application writers should write platform independent code. I just don't see the move to 64-bit as a major issue affecting the user experience. The main software obstacle, at this point, is multi-core support and scalability. Besides, compared to the move from 16-bit to 32-bit, the move from 32- to 64-bit is cake.
  • quleap, on 10/16/2007, -0/+1The author needs some basic training in computer architecture. What the hell is "relative pointer"? I thought he was referring to offset addressing. Has he ever heard of virtual memory? At least 386 already supported it.
    • eatmorgnome, on 10/17/2007, -0/+1Maybe you should learn2research before firing off shots with your vast Computer Engineering 101 knowledge. Start here, http://www.x86-64.org/documentation/assembly.html
    • programmega, on 10/17/2007, -0/+1seriously, if you don't know what "relative pointer" is and manage to confuse it with virtual memory, you REALLY shouldn't be trying to school someone who's obviously got enough knowledge to write for a pro site. If eatmorgnome's reference isn't enough for you, you may wish to try something smaller and easier to digest, maybe wikipedia's article:
      http://en.wikipedia.org/wiki/X86-64
      Then again, you seem to not be the only one out of these 120+ comments that seems to think they know more than AMD, Intel, OR the author.
  • digitallysick, on 10/16/2007, -0/+1want to see the difference in the 2? install vista on a 32 bit machine, and then on a 64 bit, same with Linux (32 bit and 64 bit) then work with graphics in adobe/gimp and you can see the difference when loading filters/plugins for sure.
  • dougmc, on 10/17/2007, -0/+4`However, neither should be confused with IA-64, which is the Intel Itanium instruction set. This is currently the only true 64-bit instruction set in general use today'

    They must have a creative version of `only', `true' or `in general use today'.
  • brundlefly76, on 10/17/2007, -0/+2Its funny, the first UNIX computer I ever used (and the first internet-connected computer I ever used) was a 64-bit computer - a DEC Alpha.
    Its so weird to think that 13 years later we are still struggling with 64-bit adoption on the desktop.

    I have used x64 chips since they came out, but stopped using them on the desktop almost immediately, as it is completely true - for desktop use, the hardware and software compatibility problems are not worth it. I even run Ubuntu in 32-bit just to save an inevitable headache down the road (examples being the firefox/flash 32/64 bit issues under linux and the fact that the extended mplayer codecs I need to do my video work are 32-bit only).

    However, I have 2 cabs full of 64-bit linux servers, because in my experience both MySQL (especially) and Apache run significantly better under 64-bit.
    • iwblj, on 10/17/2007, -0/+0I'm using 64-bit Kubuntu (edgy eft) on a quad-core Opteron. My only substantive complaint is the situation with the flash plugin for firefox. I could install a 32-bit firefox instead, but in the end I really don't care enough to bother. In my case the system has 4GB of RAM so if I'd installed the normal OS I would have only had access to about 3GB because of reserved space for the graphics card and other i/o devices. On a home system I probably would just install the 32-bit firefox and keep everything else native.
  • seanof, on 10/17/2007, -0/+1"Unlike today's computers, these more rudimentary processors did not allow for data to be spanned over multiple words, limiting the functions that could be done over the already sluggish clock speeds. If any data was larger than 16 bits it would end up truncated, and smaller would end up crashing (and so would need to be "padded" before processing)."

    This is nonsense. You can perform operations on any data if you have enough memory to hold the result. It may take more then one machine instruction, but it can be done. You can also manipulate data as small as 8 bits in any x86 processor.

    "By making use of the relative pointers, each program is capable of running in its own "virtual space" rather than in an absolute position within the CPU and memory."

    This has been possible since the 286
  • LordofShadows, on 10/22/2007, -1/+1Article is wrong, it's just the ram.
  • cmbdon, on 11/05/2007, -0/+0got myself a 10000 bit computer - dropped it from 5th floor of my office - didnt improve the speed of anything though - well apart from the passers by walking speed when they saw it plummet towards them !!!! ROFL !
  •  

    Login or register to add a comment

    Create a new account or login to join in the conversation on Digg. You'll also be able to Digg stories to help promote things you like.


© Digg Inc. 2008 — Content posted by Digg users is dedicated to the public domain.
DIGG, DIGG IT, DUGG, DIGG THIS, Digg graphics, logos, designs, page headers, button icons, scripts, and other service names are the trademarks of Digg Inc.