Re: [OT] Re: [gentoo-user] Intel Core Duo Processor - Anyone?

[Lord Sauron]

On 3/30/06, Mike Myers <fluffymikey@×××××.com> wrote:
> Just my .02c, but it seems like the 64-bit processors come with more
> hype than benefits.  Not that the 64-bit move is a bad thing at all,
> but I mean it just seems like people tend to expect much more out of
> them than what they should.

You're very close to the mark, actually.

> It would seem like a more accurate, but oversimplified explaination
> would be that it simply allows for other improvements within the
> computer, but it does not improve anything on it's own.  For instance,

Yes, you're very close.  It does allow for one major thing OTTOH. 
With a 64-bit wide word, more precise calculations take half the time
they would on a 32-bit chip.  They don't give any other real major
64-bit exclusive benefits though.  The advantages of a 64-bit variable
isn't really relevant for most uses though.  Things like MatLab are
greatly benefitted, however, normal desktop use isn't.  Some video
games are now being made 64-bit, so they'll benefit from more precise
gameplay at higher speeds, however, you are right: 64-bit en se
doesn't give any other amazing benefit.

> allowing >4GB ram, which in turn gives better performance.  From what
> I've read, there are improvements in certain things that are specific
> to number crunching, like a database with mathematical formulas.

Yup.

> However, for a desktop processor, the difference is going to be barely
> noticeable, if any, especially since most desktops don't use more than
> 4 gigs of ram.

True.  However, sticking to 32-bit for the rest of forever isn't a
terribly good idea, now is it?

> It definitely seems to be a difficult thing to explain though due to
> the nature of the processor.  Most people think simply 'more numbers =
> more speed', but that's not really case, and surely not the point.
> Since around the mid 90's, processor speeds have steadily increased,
> but in the last couple of years, that increase has halted.

Not really.  AMD is still making their chips more efficient and
faster, though the new fad is to add more cores.  However, eventually
this will still limit threads to the speed of one core, which'll
prompt more and more rapid speed increases.  Just be patient; you
don't need all that number-crunching power right now, do you? ; )

> Supposedly, the speeds have been maxed out for the size of the
> processors, so that's why the manufacturers are trying different
> routes, like hyperthreading, dual core, multi-core, and 64-bit.  None

Well, they also need to make the thing smaller.  We're still on what? 
95 nanometre?  Smaller means more transistors in the same area.

> of these features directly improve performance, but they do increase
> it's capabilities.  More specifically, they allow the computer to do
> MORE tasks better, instead of focusing on speeding up tasks.  That's
> not a bad thing really, because it's nice to be able to do multiple
> things simultaneously, like burning a cd while listening to mp3s and
> playing games on a LAMP server that's running emerge -u world without
> any degradation in performance in any of the processes.

People who do that scare me.

> That kind of performance seems to be what is intended with these
> different avenues that the chip makers are taking.  That is not to say
> that single tasks will perform any better, and I think the lack of
> discerning the difference is causing a lot of confusion for most
> people, especially when they aren't familiar with low level
> programming.

In the end this might degenerate to a "programmer's rating" thing. 
IE: one standardised benchmark.

> On 3/29/06, Richard Fish <bigfish@××××××××××.org> wrote:
> > On 3/29/06, Lord Sauron <lordsauronthegreat@×××××.com> wrote:
> > > www.alienware.com  I beg to differ.  I could have sworn I saw a laptop
> > > with more than 2G...  where was it... wow!  You appear to be right!
> > > Darn.. I could have SWORN I saw something with > 2G...
> >
> > Actually, you are right.  I neglected the monstrous Clevo laptop.  Its
> > an AMD X2 with capacity for 2 optical drives plus 2 hard drives, up to
> > 3G of memory, and a 200W power adapter.  Weighs 12-15 lbs, _not_
> > counting the power adapter!  This is acutally a Clevo design, sold by
> > Sager, AGearnotebooks, and many others.  Alienware got it with a
> > customized case.  All of the reviews I read on it basically said
> > "incredible performance, excellent display, but heavy, noisy, and
> > really hard to describe how large it really is".
> >
> > I was actually considering purchasing this beast...but the noise
> > factor scared me off.  Not really appropriate for a shared office or
> > conference room.
> >
> > > compiler helps with the 64-bit part.  It gets a bit technical, but
> > > there is a big difference between something made from the ground up as
> > > 64-bit versus something that was made 32-bit and just recompiled
> > > 64-bit.
> >
> > For most applications, this is not true.  The vast majority of C/C++
> > code that runs on a desktop system couldn't care less whether longs
> > and pointers are 32-bits or 64-bits in size.  It is a compiler
> > function to deal with that.  And it is also a compiler function to
> > determine whether 64-bit or 32-bit registers should be used for a
> > particular operation.  FYI, gcc has supported non-x86 64-bit CPUs for
> > a long time, so gcc's 64-bit support is probably more mature than you
> > think.  So are the applications...many open source applications were
> > ported and adapted (if necessary) to 64-bit sparc and alpha processors
> > back in the late 90s.
> >
> > There are opportunities for some programs to take advantage of special
> > processor operations through assembly instructions.  This is very
> > similar to how 3Dnow, MMX, SSE, et. al. make programs faster.  So
> > there may be some specific optimizations for some operations that can
> > be improved over time.
> >
> > An example of an application domain that could benefit from 64-bit is
> > encryption, because for key setups you need to calculate very large
> > numbers.  Such numbers could be calculated about twice as fast with
> > 64-bit operations vs 32-bit.  *BUT*, this does almost nothing for the
> > actual data encryption itself.
> >
> > A good resource on the 64-bit vs 32-bit issues is to look at AMDs
> > optimization guide for software developers.  Chapter 3 is particularly
> > relevant:
> >
> > http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/25112.PDF
> >
> > -Richard
> >
> > --
> > gentoo-user@g.o mailing list
> >
> >
>
>
> --
> Mike Myers
> mike@××××.us
> http://www.yaay.us
>
> --
> gentoo-user@g.o mailing list
>
>


--
========== GCv3.12 ==========
GCS d-(++) s+: a? C++ UL+>++++ P+
L++ E--- W+(+++) N++ o? K? w--- O? M+
V? PS- PE+ Y-(--) PGP- t+++ 5? X R tv-- b+
                DI+++ D+ G e* h- !r !y
========= END GCv3.12 ========

-- 
gentoo-user@g.o mailing list