Re: G4 vs. Intel/AMD Performance

In article <2005112121341877923%micron@invalidnet>,
Michelle Ronn <micron@xxxxxxxxxxx> wrote:

> On 2005-11-21 13:18:34 -0800, TheLetterK <theletterk@xxxxxxxxxxxxxxxxxx>
> said:
>

> >>
> >> Yes, AMD even smokes the Pentium M in performance. The Opteron can run
> >> down at 20 watts,
> > If you clock it really low. At which point your performance is lower.
>
> Duh. This is how most low power processors, even the M, operate. You do
> not get that fabulous battery life by running the proc at full tilt.
>
> >
> >> and the portable AMD parts match the power of the Pentium M.
> > In what dimension are you living?
>
> Earth, come and join me. The Turon clocks in with 25W and 30W versions.
> The Pentium M starts at 27W.
>
> Last time I checked, 25W < 27W. However, that is in the number system
> that I use. Your number system may vary.

The problem is that you want to have it both ways. You want to talk
about the Turon when you're talking about energy usage, but you talk
about the Opteron when you talk about performance.

The Opteron can't compete with Pentium M on energy usage and Turon can't
compete with Pentium M on performance.

The Pentium M offers a solid balance of performance AND energy usage.
THAT is the critical matter here.


>
> The clock in that AMD tends to rate their power draws on the high side,
> and Intel on the low side. (basing this on Opterons that are rated at
> 89W and actually draw 60W at full throttle, against Intel Xeons that
> are rated at 105W and pull 105W)

Hardly relevant - since the Xeon isn't a desktop PC processor, nor is it
one of the chips Apple is considering.

>
> >
> >> It is not really fair for Intel to compare an M to an Opteron. The perf
> >> difference is a bit embarassing for Intel.
> >>
> >> The core performance difference in the Opteron and anything that Intel
> >> offers is that the Opteron puts the memory controller on the processor
> >> complex. Intel does not.
> > On-die memory controllers are nice, but not essential to good
> > performance. Different designs for different needs.
>
> Um, go look that one up. You thinking is a little flawed.

And, yet, when AMD first came out with the on-die memory controllers,
they were bragging about something like a 5% performance gain. It's just
not that big a deal in the scheme of things. Sure, it would be nice, but
there are plenty of other factors which are at least as important.

>
> >
> >> AMD has Intel smoked in memory bandwidth
> > Big deal. Both designs have more bandwidth than can be effectively used
> > with current memory chips.
>
> Nope again. The AMD designs have significantly less memory latency.
> 60ns vs 200ns for Intel. If you are disk bound, then yes... memory does
> not matter in that case so much. I am seeing OS installations taking 40
> mins with Intel, and 17 with AMD. I used to think that OS installations
> were disk bound as well.

If you're seeing a 17 minute vs 40 minute installation of Windows (which
is mostly limited by CD and hard disk speed), then you're not comparing
comparable systems.
.

Relevant Pages

  • Re: Processors
    ... the big difference is that Pentium and Celeron ... AMD is a company that makes processors;-) ... Pentium and Celeron come from the Intel stable. ... Turion (although they make low-power chips like the Geode and many other ...
    (alt.os.linux.suse)
  • Dell, Intels "whore" losing loyalty battle
    ... But Canada's version of Dell, MDG is still an Intel-only whore. ... AMD had already got its foot partially in Dell's ... already selling Opteron and Athlon 64 based systems. ... Intel only shop and AMD continued its rapid growth in the enterprise. ...
    (comp.sys.intel)
  • Amd-Intel
    ... AMD vs. Intel - The eternal debate ... Who is AMD? ... Advanced Micro Devices, Inc. engages in the design, manufacture, ... it was too slow to compete effectively with Intel's Pentium ...
    (comp.sys.ibm.pc.hardware.chips)
  • Re: Amd-Intel
    ... > Who is AMD? ... > Advanced Micro Devices, Inc. engages in the design, manufacture, ... > Who is Intel? ... it was too slow to compete effectively with Intel's Pentium ...
    (comp.sys.ibm.pc.hardware.chips)
  • Re: What is Intel IA-32?
    ... > IV, Celerons, AMD chips, and everything else that acts like a pentium. ... it's what Intel used to call "x86" but they changed to "Intel ...
    (alt.lang.asm)
Loading