Why Does FBSD Run Faster Under i386 On a 64-Bit Computer? - BSD

This is a discussion on Why Does FBSD Run Faster Under i386 On a 64-Bit Computer? - BSD ; Hi! I got an amd64 processor and was googling to see if I could find a hack to run Nvidia i386 graphics card on an amd64 box. (No luck there) Anyway, in the process of my googling this stuff, I ...

+ Reply to Thread
Results 1 to 20 of 20

Thread: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

  1. Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    Hi!

    I got an amd64 processor and was googling to see if I could find a hack to
    run Nvidia i386 graphics card on an amd64 box. (No luck there)

    Anyway, in the process of my googling this stuff, I was reading that
    numerous people with amd64 processors were running FBSD-i386 O/S - instead
    of going with the amd64 version of FBSD.

    So I downloaded the i386 ISO's, had mom send/donate $20 for the bandwith to
    the FBSD team.

    I installed FBSD-i386-6.2-RELEASE and it seems to run faster than the amd64
    FBSD O/S. Why is that?

    I figured I would regret the decision when I went to compile software. NOT,
    I installed firefox under amd64 it took 25 minutes to build and install, it
    took the same amount of time to build and install under the i386 version of
    FBSD.

    Don't know if its true or not, but some of the stuff I read while googling
    said that you won't see much of an increase in amd64 unless you're running
    more than 4-gig of ram, why is that?

    Also, I read some of the post on freebsd-forum and some people said that for
    a desktop system (That's what I use) i386 is better, that amd64 would be
    better for servers.

    It seems strange to me that running i386 on a amd64 bit box would run faster
    than running amd64 on an amd64 box..

  2. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    On Mon, 22 Jan 2007 03:29:38 -0500, Timmy wrote:
    > Hi!
    >
    > I got an amd64 processor and was googling to see if I could
    > find a hack to run Nvidia i386 graphics card on an amd64
    > box. (No luck there)


    This is, unfortunately, something that only NVIDIA can fix.
    Since they only release an unfree, 32-bit only version of their
    driver, the best thing to do is to let them know how bad this is
    by contacting them and letting them know that you would feel much
    better if a driver for your FreeBSD/amd64 system was available.

    > Anyway, in the process of my googling this stuff, I was reading
    > that numerous people with amd64 processors were running
    > FBSD-i386 O/S - instead of going with the amd64 version of
    > FBSD.
    >
    > So I downloaded the i386 ISO's, had mom send/donate $20 for the
    > bandwith to the FBSD team.
    >
    > I installed FBSD-i386-6.2-RELEASE and it seems to run faster
    > than the amd64 FBSD O/S. Why is that?


    Because being able to use a 64-bit virtual-address space is *NOT*
    only about speed. It's a false assumption to being with that a
    larger address range is magically going to make everything run
    obviously faster.

    Even if it does, you will have to run laborious and careful
    benchmarks before you can reason with any amount of confidence
    that there *is* some speed gain by using a 64-bit system. Even
    then, the speed gain may be limited to the very particular set of
    benchmarks you have tested, so it still doesn't mean that
    *everything* will be faster.

    > I figured I would regret the decision when I went to compile
    > software. NOT, I installed firefox under amd64 it took 25
    > minutes to build and install, it took the same amount of time
    > to build and install under the i386 version of FBSD.


    Unfortunately, this only means that compiling the particular
    piece of software isn't a very good benchmark about the
    differences between i386 and amd64 ports

    > Don't know if its true or not, but some of the stuff I read
    > while googling said that you won't see much of an increase in
    > amd64 unless you're running more than 4-gig of ram, why is
    > that?


    Because that's when you will start using physical memory amounts
    which are only possible with amd64 64-bit binaries

    > Also, I read some of the post on freebsd-forum and some people
    > said that for a desktop system (That's what I use) i386 is
    > better, that amd64 would be better for servers.


    This may be true, or it may be untrue. It depends very highly on
    what qualifies as a 'desktop system'.

    Regards,
    Giorgos


  3. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    In article <873b62epec.fsf@kobe.laptop>,
    Giorgos Keramidas writes:
    > On Mon, 22 Jan 2007 03:29:38 -0500, Timmy wrote:
    >> Hi!
    >>
    >> I got an amd64 processor and was googling to see if I could
    >> find a hack to run Nvidia i386 graphics card on an amd64
    >> box. (No luck there)

    >
    > This is, unfortunately, something that only NVIDIA can fix.
    > Since they only release an unfree, 32-bit only version of their
    > driver, the best thing to do is to let them know how bad this is
    > by contacting them and letting them know that you would feel much
    > better if a driver for your FreeBSD/amd64 system was available.


    This isn't completely true. A NVIDIA engineer wrote a long detailed
    email to either the hackers or amd64 mailing list with a list of
    needed FreeBSD kernel changes. John Baldwin did some of the work,
    but more needs to be done.


    --
    Steve
    http://troutmask.apl.washington.edu/~kargl/

  4. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    Giorgos Keramidas wrote:

    > On Mon, 22 Jan 2007 03:29:38 -0500, Timmy wrote:
    >> Hi!
    >>
    >> I got an amd64 processor and was googling to see if I could
    >> find a hack to run Nvidia i386 graphics card on an amd64
    >> box. (No luck there)

    >
    > This is, unfortunately, something that only NVIDIA can fix.
    > Since they only release an unfree, 32-bit only version of their
    > driver, the best thing to do is to let them know how bad this is
    > by contacting them and letting them know that you would feel much
    > better if a driver for your FreeBSD/amd64 system was available.


    Good idea, I well make a post on their M/B and hope that the developers read
    those request.

    >> Anyway, in the process of my googling this stuff, I was reading
    >> that numerous people with amd64 processors were running
    >> FBSD-i386 O/S - instead of going with the amd64 version of
    >> FBSD.
    >>
    >> So I downloaded the i386 ISO's, had mom send/donate $20 for the
    >> bandwith to the FBSD team.
    >>
    >> I installed FBSD-i386-6.2-RELEASE and it seems to run faster
    >> than the amd64 FBSD O/S. Why is that?

    >
    > Because being able to use a 64-bit virtual-address space is *NOT*
    > only about speed. It's a false assumption to being with that a
    > larger address range is magically going to make everything run
    > obviously faster.
    >
    > Even if it does, you will have to run laborious and careful
    > benchmarks before you can reason with any amount of confidence
    > that there *is* some speed gain by using a 64-bit system. Even
    > then, the speed gain may be limited to the very particular set of
    > benchmarks you have tested, so it still doesn't mean that
    > *everything* will be faster.


    I haven't run any type of benchmark programs, but everything seems to be
    just as fast or faster when compiling. I figured running amd64 programs on
    an amd64 O/S would blow i386 out of the water.

    >> I figured I would regret the decision when I went to compile
    >> software. NOT, I installed firefox under amd64 it took 25
    >> minutes to build and install, it took the same amount of time
    >> to build and install under the i386 version of FBSD.

    >
    > Unfortunately, this only means that compiling the particular
    > piece of software isn't a very good benchmark about the
    > differences between i386 and amd64 ports


    i386 seems to be upgrade the whole system in the same amount of time using
    portupgrade -arR.

    >> Don't know if its true or not, but some of the stuff I read
    >> while googling said that you won't see much of an increase in
    >> amd64 unless you're running more than 4-gig of ram, why is
    >> that?

    >
    > Because that's when you will start using physical memory amounts
    > which are only possible with amd64 64-bit binaries



    Than my computer is fubar. Why sell any computer with amd64 chips, while the
    M/B is only upgradeable to 2-gig of ram? If you need 4-gigs of ram to get
    the benefit of amd64 - than the M/B should 'come with' or 'be upgradeable'
    to 4-gigs of ram. My M/B is only upgradeable to 2-gigs which seems to make
    the amd64 processor obsolete, yes? I guess I'm a moron for not staying with
    intel chip sets.

    >> Also, I read some of the post on freebsd-forum and some people
    >> said that for a desktop system (That's what I use) i386 is
    >> better, that amd64 would be better for servers.

    >
    > This may be true, or it may be untrue. It depends very highly on
    > what qualifies as a 'desktop system'.


    Desktop:

    Xorg and Windowmaker

    Mail: Mutt, procmail, postfix, spamass-milter/spamassasin, and fetchmail.

    Media: Mplayer, Xine, Xmms & Vlc.

    Browser: Firefox and Opera and other odds and ends.


    > Regards,
    > Giorgos



  5. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    Giorgos Keramidas wrote:
    > > I installed FBSD-i386-6.2-RELEASE and it seems to run faster
    > > than the amd64 FBSD O/S. Why is that?

    >
    > Because being able to use a 64-bit virtual-address space is *NOT*
    > only about speed. It's a false assumption to being with that a
    > larger address range is magically going to make everything run
    > obviously faster.
    >
    > Even if it does, you will have to run laborious and careful
    > benchmarks before you can reason with any amount of confidence
    > that there *is* some speed gain by using a 64-bit system. Even
    > then, the speed gain may be limited to the very particular set of
    > benchmarks you have tested, so it still doesn't mean that
    > *everything* will be faster.


    Not everytning will be faster, true. But some things are considerably
    faster and you don't need laborious benchmarks to see it. I have
    numerical computations which run more than 50% faster in 64 bits mode,
    and notably formal computations with maple run between 50% and 100%
    faster in 64 bits mode (with maple 10 64 bits). The difference is
    absolutely gigantic.

    >
    > This may be true, or it may be untrue. It depends very highly on
    > what qualifies as a 'desktop system'.


    What is a pain in 64 bits mode is to have 64 bits applications requiring
    32 bits plugins such as firefox with acroread or flash plugin. Currently
    i have this combination running fine on Ubuntu, but it was certainly not
    "out of the box". This may restrict the usefulness of 64 bits mode for
    desktops. Also, as far as i can see OpenOffice 64 bits (on Ubuntu) runs
    well except openoffice-base which doesn't see the java installation.
    According to Google it seems that this is a well known problem.
    However all other Java programs (eclipse, netbeans, etc.) run fine in 64
    bits mode.

    --

    Michel TALON


  6. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    Michel Talon wrote:

    > Giorgos Keramidas wrote:
    >> > I installed FBSD-i386-6.2-RELEASE and it seems to run faster
    >> > than the amd64 FBSD O/S. Why is that?

    >>
    >> Because being able to use a 64-bit virtual-address space is *NOT*
    >> only about speed. It's a false assumption to being with that a
    >> larger address range is magically going to make everything run
    >> obviously faster.
    >>
    >> Even if it does, you will have to run laborious and careful
    >> benchmarks before you can reason with any amount of confidence
    >> that there *is* some speed gain by using a 64-bit system. Even
    >> then, the speed gain may be limited to the very particular set of
    >> benchmarks you have tested, so it still doesn't mean that
    >> *everything* will be faster.

    >
    > Not everytning will be faster, true. But some things are considerably
    > faster and you don't need laborious benchmarks to see it. I have
    > numerical computations which run more than 50% faster in 64 bits mode,
    > and notably formal computations with maple run between 50% and 100%
    > faster in 64 bits mode (with maple 10 64 bits). The difference is
    > absolutely gigantic.
    >


    Hi.
    If you don't mind me asking, how much ram are you using when you run these
    benchmarks? Are you using 4-gig or more? From what I understand you will
    need a least 4-gig to see any major speed increase.

  7. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    Timmy wrote:
    >
    > Hi.
    > If you don't mind me asking, how much ram are you using when you run these
    > benchmarks? Are you using 4-gig or more? From what I understand you will
    > need a least 4-gig to see any major speed increase.


    I run with 1 gig. You absolutely not need > 4gig to see a difference. Of
    course with > 4 Gig, 64 bits is almost mandatory to have performance.

    --

    Michel TALON


  8. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    On Mon, 22 Jan 2007 07:09:37 -0500, Timmy wrote:

    > If you don't mind me asking, how much ram are you using when you run these
    > benchmarks? Are you using 4-gig or more? From what I understand you will
    > need a least 4-gig to see any major speed increase.


    Michael can (and probably will) answer for himself, of course, but the
    question is simply much more complicated than this simple generalization.

    On the down-side, 64 bit pointers are larger than 32-bit pointers, so
    they occupy more space in memory, so caches of a particular fixed size are
    less effective, to the extent that memory for a particular application is
    filled with pointers. Applications vary on this.

    On the plus side, it is really quite difficult for 32-bit applications to
    access more than 2 to 4G of memory, and so have to resort to various
    work-arounds. These tend to be significantly slower than the direct
    approach that is possible with 64-bit pointers. This depends on the size
    of your memory and your data set. Usage varies.

    On the plus side, the 64 bit architectecture is actually not just all
    about wider registers. There are also more of them, so the calling
    convention is different, and the opportunities for compiler optimizations
    are a bit different. Position-independent code, as used in shared
    libraries, is significantly more efficient, for example.

    Associated with that last point: compile times are not useful comparison
    points, because the compilers are in a real sense different: they are
    producing code for what are really different processors, so you should not
    expect the execution times to be the same. (Also, compilers tend to be
    pointer-data-structure-rich examples of applications, which could
    exacerbate the first point about cache pressure from lots of long pointers
    in memory structures.)

    Cheers,

    --
    Andrew


  9. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    Andrew Reilly wrote:

    > On Mon, 22 Jan 2007 07:09:37 -0500, Timmy wrote:
    >
    >> If you don't mind me asking, how much ram are you using when you run
    >> these benchmarks? Are you using 4-gig or more? From what I understand you
    >> will need a least 4-gig to see any major speed increase.

    >
    > Michael can (and probably will) answer for himself, of course, but the
    > question is simply much more complicated than this simple generalization.
    >
    > On the down-side, 64 bit pointers are larger than 32-bit pointers, so
    > they occupy more space in memory, so caches of a particular fixed size are
    > less effective, to the extent that memory for a particular application is
    > filled with pointers. Applications vary on this.
    >
    > On the plus side, it is really quite difficult for 32-bit applications to
    > access more than 2 to 4G of memory, and so have to resort to various
    > work-arounds. These tend to be significantly slower than the direct
    > approach that is possible with 64-bit pointers. This depends on the size
    > of your memory and your data set. Usage varies.
    >
    > On the plus side, the 64 bit architectecture is actually not just all
    > about wider registers. There are also more of them, so the calling
    > convention is different, and the opportunities for compiler optimizations
    > are a bit different. Position-independent code, as used in shared
    > libraries, is significantly more efficient, for example.
    >
    > Associated with that last point: compile times are not useful comparison
    > points, because the compilers are in a real sense different: they are
    > producing code for what are really different processors, so you should not
    > expect the execution times to be the same. (Also, compilers tend to be
    > pointer-data-structure-rich examples of applications, which could
    > exacerbate the first point about cache pressure from lots of long pointers
    > in memory structures.)
    >
    > Cheers,
    >


    Thanks for the explanation, I have a general idea of what you're saying. I'm
    not programmer YET ;-) I can make calculators in perl & calculators in C++,
    I'm learing my way around the G++ compiler, Yeah, I know that's lame, but
    you have to start somewhere..

    You seem to comfirm what I've read, -> I think. Is there any reason to use
    64bit-O/S on a box that has 1-gig of ram? Is there any advantage of using
    amd64 on a box with less than 4-gig of ram for a desktop computer?

    TIA



  10. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    >Thanks for the explanation, I have a general idea of what you're saying. I'm
    >not programmer YET ;-) I can make calculators in perl & calculators in C++,
    >I'm learing my way around the G++ compiler, Yeah, I know that's lame, but
    >you have to start somewhere..
    >
    >You seem to comfirm what I've read, -> I think. Is there any reason to use
    >64bit-O/S on a box that has 1-gig of ram? Is there any advantage of using
    >amd64 on a box with less than 4-gig of ram for a desktop computer?


    Yes. It all depends apon what your applications do. For
    some applications there are gains. For others there are
    losses.

    >TIA


  11. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    On Mon, 22 Jan 2007 04:55:51 -0500
    Timmy wrote:

    > Than my computer is fubar. Why sell any computer with amd64 chips, while
    > the M/B is only upgradeable to 2-gig of ram? If you need 4-gigs of ram to
    > get the benefit of amd64 - than the M/B should 'come with' or 'be


    You need 4GB or more to benefit from 64 bit addressing modes, but
    there's more to an AMD64 than 64 bit addressing modes.

    > upgradeable' to 4-gigs of ram. My M/B is only upgradeable to 2-gigs which
    > seems to make the amd64 processor obsolete, yes? I guess I'm a moron for
    > not staying with intel chip sets.


    Not really - an AMD64 is still a very fast i386 implementation,
    I'm not sure what hits the best of the price/performance stakes right now
    but AMD64 has certainly been there and may still be.

    The key question is (as always) does it do what you need it to do
    for a price you are happy to pay - if so smile, if not shop more
    carefully

    --
    C:>WIN | Directable Mirror Arrays
    The computer obeys and wins. | A better way to focus the sun
    You lose and Bill collects. | licences available see
    | http://www.sohara.org/

  12. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    Steve O'Hara-Smith wrote:
    > On Mon, 22 Jan 2007 04:55:51 -0500
    > Timmy wrote:


    > > Than my computer is fubar. Why sell any computer with amd64 chips, while
    > > the M/B is only upgradeable to 2-gig of ram? If you need 4-gigs of ram to
    > > get the benefit of amd64 - than the M/B should 'come with' or 'be


    > You need 4GB or more to benefit from 64 bit addressing modes, but
    > there's more to an AMD64 than 64 bit addressing modes.


    I'd disagree with that -- I have a huge database (which will certainly
    grow to over 4GB in the future) that benefits from 64 bit addressing
    because I simply mmap() the file into my address space; I don't necessarily
    bring all of the database into physical memory -- I just need to be able
    to directly dereference pointers... Works really well, and I couldn't do
    it with >32bit pointers.

    Side question: I know the sizeof() a pointer is 64 bits, but how many
    significant bits are there currently, 48? [AMD64/FreeBSD 6.2]

    Cheers,
    -RK
    --
    Robert Krten, Antique computer collector looking for PDP-8 and PDP-8/S
    minicomputers; check out their "good home" at www.parse.com/~museum
    Email address is valid; greylisting spam filter in effect.

  13. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    Robert Krten wrote:
    > Steve O'Hara-Smith wrote:
    > > On Mon, 22 Jan 2007 04:55:51 -0500
    > > Timmy wrote:


    > > > Than my computer is fubar. Why sell any computer with amd64 chips, while
    > > > the M/B is only upgradeable to 2-gig of ram? If you need 4-gigs of ram to
    > > > get the benefit of amd64 - than the M/B should 'come with' or 'be


    > > You need 4GB or more to benefit from 64 bit addressing modes, but
    > > there's more to an AMD64 than 64 bit addressing modes.


    > I'd disagree with that -- I have a huge database (which will certainly
    > grow to over 4GB in the future) that benefits from 64 bit addressing
    > because I simply mmap() the file into my address space; I don't necessarily
    > bring all of the database into physical memory -- I just need to be able
    > to directly dereference pointers... Works really well, and I couldn't do
    > it with >32bit pointers.


    S/B: "*without* >32 bit pointers" (and that applies now, because it's
    already past the limits of 32 bit addressing). Sorry about the typo...

    --
    Robert Krten, Antique computer collector looking for PDP-8 and PDP-8/S
    minicomputers; check out their "good home" at www.parse.com/~museum
    Email address is valid; greylisting spam filter in effect.

  14. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    On Tue, 23 Jan 2007 07:37:04 -0600
    info2007@parse.com (Robert Krten) wrote:

    > Steve O'Hara-Smith wrote:
    > > On Mon, 22 Jan 2007 04:55:51 -0500
    > > Timmy wrote:

    >
    > > > Than my computer is fubar. Why sell any computer with amd64 chips,
    > > > while the M/B is only upgradeable to 2-gig of ram? If you need 4-gigs
    > > > of ram to get the benefit of amd64 - than the M/B should 'come with'
    > > > or 'be

    >
    > > You need 4GB or more to benefit from 64 bit addressing modes,
    > > but there's more to an AMD64 than 64 bit addressing modes.

    >
    > I'd disagree with that -- I have a huge database (which will certainly
    > grow to over 4GB in the future) that benefits from 64 bit addressing
    > because I simply mmap() the file into my address space; I don't


    Hmm good point - OK I should have said "You need to be addressing
    more than 4GB to benefit from 64 bit addressing modes"

    --
    C:>WIN | Directable Mirror Arrays
    The computer obeys and wins. | A better way to focus the sun
    You lose and Bill collects. | licences available see
    | http://www.sohara.org/

  15. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    On Mon, 22 Jan 2007 12:34:40 -0500, Timmy wrote:

    > You seem to comfirm what I've read, -> I think. Is there any reason to use
    > 64bit-O/S on a box that has 1-gig of ram? Is there any advantage of using
    > amd64 on a box with less than 4-gig of ram for a desktop computer?


    Sure. I do. My dual AMD64 box only has 1G (or maybe 2, I can't
    remember), and I run FreeBSD-amd64. Advantages are: it's more fun and
    it's faster for the 64-bit integer arithmetic that I happen to do a lot of.

    Cheers,

    --
    Andrew


  16. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    Timmy wrote:
    :
    : You seem to comfirm what I've read, -> I think. Is there any reason to use
    : 64bit-O/S on a box that has 1-gig of ram? Is there any advantage of using
    : amd64 on a box with less than 4-gig of ram for a desktop computer?
    :
    I upgraded(?) my firewall to amd64 months ago. It has a
    whole 1GB of memory.

    More to the point, I was able to get my feet wet WRT amd64
    before moving my workstation to amd64. It does have 4GB, but
    the peripheral/bridge info takes up 1.5GB, so the i386 implementation
    only allowed me to use 2.5GB.

    Bruce
    --
    ------------------------------------------------------------------------
    "I like bad!" Bruce Burden Austin, TX.
    - Thuganlitha
    The Power and the Prophet
    Robert Don Hughes


  17. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    brucegb@realtime.net (Bruce Burden) writes:

    > Timmy wrote:
    > :
    > : You seem to comfirm what I've read, -> I think. Is there any reason to use
    > : 64bit-O/S on a box that has 1-gig of ram? Is there any advantage of using
    > : amd64 on a box with less than 4-gig of ram for a desktop computer?
    > :
    > I upgraded(?) my firewall to amd64 months ago. It has a
    > whole 1GB of memory.


    That's some firewall. My firewall is a 486 with 40MB.

    -- Patrick

  18. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    kkt wrote:
    :
    : That's some firewall. My firewall is a 486 with 40MB.
    :
    In a long, sad tale, I was hoping to upgrade my work
    station from dual athalons to dual dual-core opterons over
    the Memorial Day weekend. That finally happened over the
    Labor Day weekend. In the process, I wound up with a dual
    (but not dual-core) Opteron system.

    It didn't seem reasonable to leave the dual 800MHz
    Coppermines in place. Plus I fried a connector on my former
    workstation board in the thrash, so the original plan of
    using the dual athalons as the firewall went up in a bit
    of smoke...

    The basic lesson learned is to ALWAYS use memory the
    mobo manufacturer recommends, and not something "compatible".

    Bruce
    --
    ------------------------------------------------------------------------
    "I like bad!" Bruce Burden Austin, TX.
    - Thuganlitha
    The Power and the Prophet
    Robert Don Hughes


  19. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    Begin
    On 2007-01-22, Michel Talon wrote:
    > What is a pain in 64 bits mode is to have 64 bits applications requiring
    > 32 bits plugins such as firefox with acroread or flash plugin. Currently
    > i have this combination running fine on Ubuntu, but it was certainly not
    > "out of the box". This may restrict the usefulness of 64 bits mode for
    > desktops.


    Note that this is only true because ``we'' insisted, insist,
    and keep insisting, on sticking with x86 long after it was obsolete
    as an architecture[1]. There have been numerous other takes at 64 bit
    architectures, and quite a few of them much better from a technical
    PoV, but before now nobody wanted them. Now, the great unwashed seems
    to think 64bits is a substitute for the clearly bankrupt More MHz More
    Better mantra.


    [1] Some might argue that because it's all RISC-y underneath that makes
    the instruction set somehow less obsolete, but I tend to disagree.
    The amount of expansions and extentions is a bit staggering and
    embracing them all is often as not not feasible; even choosing and
    picking what you need is tricky in the face of all the different
    CPUs Out There.

    --
    j p d (at) d s b (dot) t u d e l f t (dot) n l .
    This message was originally posted on Usenet in plain text.
    Any other representation, additions, or changes do not have my
    consent and may be a violation of international copyright law.

  20. Re: Why Does FBSD Run Faster Under i386 On a 64-Bit Computer?

    Begin
    On 2007-01-25, Bruce Burden wrote:
    > It didn't seem reasonable to leave the dual 800MHz
    > Coppermines in place. Plus I fried a connector on my former
    > workstation board in the thrash, so the original plan of
    > using the dual athalons as the firewall went up in a bit
    > of smoke...


    Dual athlons seems a bit of overkill for a firewall. You could even
    try and calculate whether you can finance a new board plus nx series
    amd geode[1] cpu with the expected savings in power. Some preliminary
    calculations and guesswork indicated I'd break even with replacing my
    750MHz athlon in about two years. Replacing your SMP box'll likely do
    that -- stuff of that generation is quite power hungry. Too bad nx1500's
    are actually more expensive than nx1750s, and the board support for nxes
    is scetchy. Still, with rising energy prices it might soon be viable.


    [1] Which should be quite enough. Or a soekris or pcengines board, of
    course. Also, Wim van de Putte (the belgian obsd on soekris nut) has
    a new shiny low power via board with 4x gige in the works. Still
    x86, and you won't run wire speed on all the gige ports as that'd
    swamp the bus, but fairly nice all the same.

    --
    j p d (at) d s b (dot) t u d e l f t (dot) n l .
    This message was originally posted on Usenet in plain text.
    Any other representation, additions, or changes do not have my
    consent and may be a violation of international copyright law.

+ Reply to Thread