Strange, when one goes to www.spec.org to get the officially obtained
spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
using Apple's definition

In article <no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>,
no <no[at]no.com> wrote:

> Strange, when one goes to www.spec.org to get the officially obtained
> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
> the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
> using Apple's definition

The higher specbenchs on the spec site are probably due to the intel
platform making use of a different compiler than gcc-3.3..

Vareck Bostrom <v.bostrom[at]attbi.com> wrote:
> no <no[at]no.com> wrote:
>> Strange, when one goes to www.spec.org to get the officially obtained
>> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
>> the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
>> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
>> using Apple's definition
>
>The higher specbenchs on the spec site are probably due to the intel
>platform making use of a different compiler than gcc-3.3..

That and a commercial malloc library costing about a thousand dollars.

-- Erick

Vareck Bostrom <v.bostrom[at]attbi.com> wrote:
> erick[at]sfu.ca (Erick Bryce Wong) wrote:
>> Vareck Bostrom <v.bostrom[at]attbi.com> wrote:
>> > no <no[at]no.com> wrote:
>> >> Strange, when one goes to www.spec.org to get the officially obtained
>> >> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
>> >> the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
>> >> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
>> >> using Apple's definition
>> >
>> >The higher specbenchs on the spec site are probably due to the intel
>> >platform making use of a different compiler than gcc-3.3..
>>
>> That and a commercial malloc library costing about a thousand dollars.
>
>Which library is that?

Something called MicroQuill SmartHeap library. Here's the config file
from the SPEC web site:

http://www.spec.org/cpu2000/results/res2003q2/cpu2000-20030407-02062.cfg

-- Erick

no <no[at]no.com> wrote in message news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
> Strange, when one goes to www.spec.org to get the officially obtained
> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
> the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
> using Apple's definition

Please. SPEC is for penis measurement ("see! I've got 1/8" on you!
neener neener"). Watching the keynote, especially the audio app,
showed Dell's dual Xeon getting down on all fours and barking like the
dog that it is.

The only thing in Wintel defense is that, since Intel is
compartmentally lame, their dual systems ("workstation") are lagging
their P4 systems at the moment.

But the dual 3.06 vs. dual 2.0 on that Photoshop test went pretty
unexpectedly bad for Intel. Megahertz myth indeed. Why is Intel so
shitty?

=Heywood=

no <no[at]no.com> wrote in message news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
> Strange, when one goes to www.spec.org to get the officially obtained
> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
> the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
> using Apple's definition

That's because the scores on SPECs website are using some sort of
different scale, I did earlier see officiall SPEC scores for a single
1.8Ghz G5 which scored ~1250/~1350 for Int/FP. I'll try and find the
URL.

Bob

In article <v.bostrom-BBB252.04405124062003[at]netnews.attbi.com>,
Vareck Bostrom <v.bostrom[at]attbi.com> wrote:

> In article <dd5de929.0306240143.70cd5376[at]posting.google.com>,
> imouttahere[at]mac.com (Heywood Mogroot) wrote:
>
> > no <no[at]no.com> wrote in message
> > news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
> > > Strange, when one goes to www.spec.org to get the officially obtained
> > > spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
> > > the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
> > > 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
> > > using Apple's definition
> >
> > Please. SPEC is for penis measurement ("see! I've got 1/8" on you!
> > neener neener"). Watching the keynote, especially the audio app,
> > showed Dell's dual Xeon getting down on all fours and barking like the
> > dog that it is.
> >
> > The only thing in Wintel defense is that, since Intel is
> > compartmentally lame, their dual systems ("workstation") are lagging
> > their P4 systems at the moment.
> >
> > But the dual 3.06 vs. dual 2.0 on that Photoshop test went pretty
> > unexpectedly bad for Intel. Megahertz myth indeed. Why is Intel so
> > shitty?
> >
> > =Heywood=
>
> Perhaps we're seeing the advantage of an (effective) 1 GHz FSB vs an
> (effective) 533 MHz FSB? Not to mention the PPC970 is, after all, a
> pretty good chip.
>
> What kind of cache is on the 970s anyway?

IIRC, it's 512 kb L2 and 2 MB L3 - at least on the 2 GHz model.

One thing that hasn't gotten much comment is the fact that nothing seems
to share a bus anymore (slightly exaggerated). For example, there are
TWO buses from CPU to memory and nothing else shares those buses.

That alone must help real world performance.

flip <flippo[at]mac.com> wrote:
> In article <v.bostrom-BBB252.04405124062003[at]netnews.attbi.com>,
> Vareck Bostrom <v.bostrom[at]attbi.com> wrote:
>
>> In article <dd5de929.0306240143.70cd5376[at]posting.google.com>,
>> imouttahere[at]mac.com (Heywood Mogroot) wrote:
>>
>>> no <no[at]no.com> wrote in message
>>> news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
>>>> Strange, when one goes to www.spec.org to get the officially obtained
>>>> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
>>>> the int test) the scores for the dell 650 ARE NOT LOWER than the dual
>>>> g5 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC,
>>>> even using Apple's definition
>>>
>>> Please. SPEC is for penis measurement ("see! I've got 1/8" on you!
>>> neener neener"). Watching the keynote, especially the audio app,
>>> showed Dell's dual Xeon getting down on all fours and barking like the
>>> dog that it is.
>>>
>>> The only thing in Wintel defense is that, since Intel is
>>> compartmentally lame, their dual systems ("workstation") are lagging
>>> their P4 systems at the moment.
>>>
>>> But the dual 3.06 vs. dual 2.0 on that Photoshop test went pretty
>>> unexpectedly bad for Intel. Megahertz myth indeed. Why is Intel so
>>> shitty?
>>>
>>> =Heywood=
>>
>> Perhaps we're seeing the advantage of an (effective) 1 GHz FSB vs an
>> (effective) 533 MHz FSB? Not to mention the PPC970 is, after all, a
>> pretty good chip.
>>
>> What kind of cache is on the 970s anyway?
>
> IIRC, it's 512 kb L2 and 2 MB L3 - at least on the 2 GHz model.

there is no L3

-JB

In article <bd9g73$cje$1[at]sparta.btinternet.com>,
"James Boswell" <JamesBoswell[at]Btopenworld.com> wrote:

> flip <flippo[at]mac.com> wrote:
> > In article <v.bostrom-BBB252.04405124062003[at]netnews.attbi.com>,
> > Vareck Bostrom <v.bostrom[at]attbi.com> wrote:
> >
> >> In article <dd5de929.0306240143.70cd5376[at]posting.google.com>,
> >> imouttahere[at]mac.com (Heywood Mogroot) wrote:
> >>
> >>> no <no[at]no.com> wrote in message
> >>> news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
> >>>> Strange, when one goes to www.spec.org to get the officially obtained
> >>>> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
> >>>> the int test) the scores for the dell 650 ARE NOT LOWER than the dual
> >>>> g5 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC,
> >>>> even using Apple's definition
> >>>
> >>> Please. SPEC is for penis measurement ("see! I've got 1/8" on you!
> >>> neener neener"). Watching the keynote, especially the audio app,
> >>> showed Dell's dual Xeon getting down on all fours and barking like the
> >>> dog that it is.
> >>>
> >>> The only thing in Wintel defense is that, since Intel is
> >>> compartmentally lame, their dual systems ("workstation") are lagging
> >>> their P4 systems at the moment.
> >>>
> >>> But the dual 3.06 vs. dual 2.0 on that Photoshop test went pretty
> >>> unexpectedly bad for Intel. Megahertz myth indeed. Why is Intel so
> >>> shitty?
> >>>
> >>> =Heywood=
> >>
> >> Perhaps we're seeing the advantage of an (effective) 1 GHz FSB vs an
> >> (effective) 533 MHz FSB? Not to mention the PPC970 is, after all, a
> >> pretty good chip.
> >>
> >> What kind of cache is on the 970s anyway?
> >
> > IIRC, it's 512 kb L2 and 2 MB L3 - at least on the 2 GHz model.
>
> there is no L3

I guess my memory failed me. I must have read that the 970 was capable
of supporting 2 MB L3 or something.

Perhaps they decided that with 1 GHz frontside bus per processor that L3
wouldn't make enough difference to justify the cost.

In comp.sys.mac.system Beelsebob <tatd100[at]cs.york.ac.uk> wrote:
> no <no[at]no.com> wrote in message news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
>> Strange, when one goes to www.spec.org to get the officially obtained
>> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
>> the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
>> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
>> using Apple's definition
>
> That's because the scores on SPECs website are using some sort of
> different scale, I did earlier see officiall SPEC scores for a single
> 1.8Ghz G5 which scored ~1250/~1350 for Int/FP. I'll try and find the
> URL.

No, it is simply because they used a different compiler. Intel's own
optimising compiler produces faster code in a lot of situations than
GCC, and especially in SPEC. In real life, you'll see a lot of speed
difference when the application can use the ram, like photoshop
or a datawarehouse.

Stephan.

In article <dd5de929.0306240143.70cd5376[at]posting.google.com>, Heywood
Mogroot <imouttahere[at]mac.com> wrote:

> no <no[at]no.com> wrote in message
> news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
> > Strange, when one goes to www.spec.org to get the officially obtained
> > spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
> > the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
> > 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
> > using Apple's definition
>
> Please. SPEC is for penis measurement ("see! I've got 1/8" on you!
> neener neener"). Watching the keynote, especially the audio app,
> showed Dell's dual Xeon getting down on all fours and barking like the
> dog that it is.

Seriously, although you can rig "real-world" apps to some extent (Logic
vs. Cubase on G5 vs. Xeon is a pretty loose comparison - there *is*
Cubase for Mac, but you don't get to promote an Apple company that
way), the bigger thing in all this is the Mac is no longer Wintel's
speed bitch, after a 4-year, post-roasted-bunny-men hiatus.

And the same anti-Mac whining happened back then. Just as funny then as
it is now. The scramble is back on...

Plus, if Jobs and IBM are true to their word about the 3GHz machine in
12 months (will there be a 4.5GHz Intel/AMD competitor next June?) then
even if the G5 only achieved parity, the Windows world has good reason
to sweat.

Unlike four years ago, the Motorola millstone is no longer around
Apple's neck. With this architecture and IBM's commitment, no one's
going to toss water on this fire any time soon.

AFAIK, this is the first Apple computer completely Motorola-free. Ever.

Wow.

One last thing about the 2/3GHz thing. I can't help but wonder if Apple
could have released a much faster machine today without thoroughly
disrupting the channel and its current inventory of G4's (and G3
iBooks). I kinda don't want to see a 1.42GHz G4 tossed into that
bake-off to see just how crappy our current machines (like the one
sitting under my work desk right now) really are. I'm having a tough
enough time decrying all the swoopy lucite on my G4's - what was that
Ive's guy *thinking*???

Maybe if I wrap it in aluminum foil ... yeah, that's the ticket.

I don't know if Apple's really ever going to shoot that far out
SPEC-wise ahead of whatever Intel/AMD is offering. Why should they?
Having all that Power4/5 headroom tucked away makes it easy to
perpetually spank Wintel for years to come without breaking a sweat.
This leaves them - for the first time ever, really - free to innovate
in many other ways, the real qualitative innovation that speed-whiz fan
boys can't wrap their adolescent minds around.

So now PC's are cheaper and have more apps and games. They're no longer
faster ... once again. And it looks like this time it's going to stay
that way for a while.

Yeah, that leaves me with a big, sh*t-eating grin...

Anyway - back for a brief visit, given this momentous occasion. And
Heywood, ya gotta resist calling the Winboy's "dipwad's." Makes it seem
like they're gettin' too you. We all know that's impossible.

Out.

Heywood Mogroot wrote in
<dd5de929.0306240143.70cd5376[at]posting.google.com>:

>no <no[at]no.com> wrote in message news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
>> Strange, when one goes to www.spec.org to get the officially obtained
>> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
>> the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
>> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
>> using Apple's definition
>
>Please. SPEC is for penis measurement ("see! I've got 1/8" on you!
>neener neener"). Watching the keynote, especially the audio app,
>showed Dell's dual Xeon getting down on all fours and barking like the
>dog that it is.
>
>The only thing in Wintel defense is that, since Intel is
>compartmentally lame, their dual systems ("workstation") are lagging
>their P4 systems at the moment.
>
>But the dual 3.06 vs. dual 2.0 on that Photoshop test went pretty
>unexpectedly bad for Intel. Megahertz myth indeed. Why is Intel so
>shitty?

Only an idiot would point to a carefully prepared keynote presentation
as if it showed a performance test under lab conditions. It would
appear you are that idiot.

In article <pbpgfvgqhgar9iam48uep2lr5ue2v09qdo[at]4ax.com>,
Steve Hanson <icustomercare[at]usps.com> wrote:

> Heywood Mogroot wrote in
> <dd5de929.0306240143.70cd5376[at]posting.google.com>:
>
> >no <no[at]no.com> wrote in message
> >news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
> >> Strange, when one goes to www.spec.org to get the officially obtained
> >> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
> >> the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
> >> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
> >> using Apple's definition
> >
> >Please. SPEC is for penis measurement ("see! I've got 1/8" on you!
> >neener neener"). Watching the keynote, especially the audio app,
> >showed Dell's dual Xeon getting down on all fours and barking like the
> >dog that it is.
> >
> >The only thing in Wintel defense is that, since Intel is
> >compartmentally lame, their dual systems ("workstation") are lagging
> >their P4 systems at the moment.
> >
> >But the dual 3.06 vs. dual 2.0 on that Photoshop test went pretty
> >unexpectedly bad for Intel. Megahertz myth indeed. Why is Intel so
> >shitty?
>
> Only an idiot would point to a carefully prepared keynote presentation
> as if it showed a performance test under lab conditions. It would
> appear you are that idiot.

So you're saying that the Presidents of Wolfram, Adobe, and a couple
other companies were lying when they said that the Mac was faster than
the PCs?

In article <flip-437A63.10093524062003[at]nnrp03.earthlink.net>,
Flip <flip[at]flippo.com> wrote:

> In article <pbpgfvgqhgar9iam48uep2lr5ue2v09qdo[at]4ax.com>,
> Steve Hanson <icustomercare[at]usps.com> wrote:
>
> > Heywood Mogroot wrote in
> > <dd5de929.0306240143.70cd5376[at]posting.google.com>:
> >
> > >no <no[at]no.com> wrote in message
> > >news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
> > >> Strange, when one goes to www.spec.org to get the officially obtained
> > >> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
> > >> the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
> > >> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
> > >> using Apple's definition
> > >
> > >Please. SPEC is for penis measurement ("see! I've got 1/8" on you!
> > >neener neener"). Watching the keynote, especially the audio app,
> > >showed Dell's dual Xeon getting down on all fours and barking like the
> > >dog that it is.
> > >
> > >The only thing in Wintel defense is that, since Intel is
> > >compartmentally lame, their dual systems ("workstation") are lagging
> > >their P4 systems at the moment.
> > >
> > >But the dual 3.06 vs. dual 2.0 on that Photoshop test went pretty
> > >unexpectedly bad for Intel. Megahertz myth indeed. Why is Intel so
> > >shitty?
> >
> > Only an idiot would point to a carefully prepared keynote presentation
> > as if it showed a performance test under lab conditions. It would
> > appear you are that idiot.
>
> So you're saying that the Presidents of Wolfram, Adobe, and a couple
> other companies were lying when they said that the Mac was faster than
> the PCs?

Musta' been PAID by Apple to lie. It's the only logical explanation. ;->

--
George Graves

Stephan Eggermont <stephan[at]turtle.stack.nl> writes:

> No, it is simply because they used a different compiler. Intel's
> own optimising compiler produces faster code in a lot of situations
> than GCC, and especially in SPEC. In real life, you'll see a lot of
> speed difference when the application can use the ram, like
> photoshop or a datawarehouse.

Keep in mind that the GCC crew may not have access to a lot of
proprietary algor/optimization ithms. Intel keeps the source code and
optimizations secret so people will shell out cash for it. Though I'm
sure Apple has probably changed some code (which they're obliged to
release under the GPL).

There's also the fact that the Intel compiler only has to run on two
architectures x86 and IA-64. How may architectures does GCC run on?
There probably are compromises done in the name of portability. Heck,
you can get a GCC port for the 6811 (not 68011!), a 8-bit
microcontroller.

For anyone interested, GCC is at: http://gcc.gnu.org/

--
David Magda <dmagda at ee.ryerson.ca>, http://www.magda.ca/
Because the innovator has for enemies all those who have done well under
the old conditions, and lukewarm defenders in those who may do well
under the new. -- Niccolo Machiavelli, _The Prince_, Chapter VI

David Magda <dmagda+trace030624[at]ee.ryerson.ca> writes:
[...]
> proprietary algor/optimization ithms. Intel keeps the source code and
[...] ^^^^^^^^^^^^^^^^^^^^^^

Oops. That should probably be "optimization algorithms".

Tim Smith <reply_in_group[at]mouse-potato.com> writes:

> In article <bd936i$1ls$1[at]morgoth.sfu.ca>, Erick Bryce Wong wrote:
> >>> >The higher specbenchs on the spec site are probably due to the
> >>> >intel platform making use of a different compiler than
> >>> >gcc-3.3..
> >>>
> >>> That and a commercial malloc library costing about a thousand dollars.
> >>
> >>Which library is that?
> >
> > Something called MicroQuill SmartHeap library. Here's the config file
> > from the SPEC web site:
> >
> > http://www.spec.org/cpu2000/results/res2003q2/cpu2000-20030407-02062.cfg
>
> That's for Dell. Check the results for Intel. They are using the Microsoft
> libraries, and get similar or better performance, so it is doubtful that the
> MicroQuill library is responsible.

Intel has their OWN compiler that they have developed to work
extraordinarily well with SPEC marks. Of course, this is not
"tweaking" or "cheating" despite the fact few people actually use the
Intel compiler and there have been cases where Intel chips have seen
a 10% gain in SPEC for the SAME chips just because of an update to
this compiler.

--
Edward Dodge

/Confabulation Consulting/

George Graves <gmgraves[at]pacbell.net> writes:


> Don't understand why all this pissing contest stuff is so
> important. To Mac users, it's just not important which computer is
> faster. If it were, most would have jumped ship long since. To
> "potential Mac customers" all that matters (speed-wise) is the clock
> speed. As long as the clock-speed numbers don't match, the one with
> the lower number is seen as not as powerful. Things like machine
> throughput, memory pipeline speeds, and other enhancements don't
> mean beans if the raw MHz on the Mac is still lower than the raw MHz
> on the Wintel side and it is.

George, the way I explain it to people is by asking them a
question: "Are there any other chips IN THE WORLD that run FASTER
than 3GHz?" They have to search their memory banks for a while
before they say, "No." Then I ask them: "Are there any other chips
IN THE WORLD that are more powerful than the Pentium 4?" I then
opine: "Of course, or every chip would be a Pentium 4!" Then I
lead them to the conclusion: "What then do you think all those GHz
are REALLY about?"

--
Edward Dodge

/Confabulation Consulting/

In comp.sys.mac.system Tim Smith <reply_in_group[at]mouse-potato.com> wrote:
> What I think is most interesting about the SPEC numbers is that AMD
> uses the Intel compiler for the SPEC results, both for Athlon and
> for Opteron.

If you dig deeply enough, there are also some GCC SPECint2000 results
for Opteron on http://www.spec.org.

rick jones
--
oxymoron n, commuter in a gas-guzzling luxury SUV with an American flag
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to raj in cup.hp.com but NOT BOTH...

> Strange, when one goes to www.spec.org to get the officially obtained
> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
> the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
> using Apple's definition

There is a very interesting article dealing with this issue:

http://www.haxial.com/spls-soapbox/apple-powermac-G5/

It seems that Apple did cheat on the benchmarks...

so long,
Sven

In article <m1fzlzgk8z.fsf[at]g3.com>, Edward Dodge <someone[at]g3.com>
wrote:

> Tim Smith <reply_in_group[at]mouse-potato.com> writes:
>
> > In article <bd936i$1ls$1[at]morgoth.sfu.ca>, Erick Bryce Wong wrote:
> > >>> >The higher specbenchs on the spec site are probably due to the
> > >>> >intel platform making use of a different compiler than
> > >>> >gcc-3.3..
> > >>>
> > >>> That and a commercial malloc library costing about a thousand dollars.
> > >>
> > >>Which library is that?
> > >
> > > Something called MicroQuill SmartHeap library. Here's the config file
> > > from the SPEC web site:
> > >
> > > http://www.spec.org/cpu2000/results/res2003q2/cpu2000-20030407-02062.cfg
> >
> > That's for Dell. Check the results for Intel. They are using the Microsoft
> > libraries, and get similar or better performance, so it is doubtful that the
> > MicroQuill library is responsible.
>
> Intel has their OWN compiler that they have developed to work
> extraordinarily well with SPEC marks. Of course, this is not
> "tweaking" or "cheating" despite the fact few people actually use the
> Intel compiler and there have been cases where Intel chips have seen
> a 10% gain in SPEC for the SAME chips just because of an update to
> this compiler.
>
> --
> Edward Dodge
>
> /Confabulation Consulting/

garbage in garbage out. It's an old saying.

s

George Graves <gmgraves[at]pacbell.net> wrote in message news:<gmgraves-1CC564.11194724062003[at]newssvr16-ext.news.prodigy.com>...
> Don't understand why all this pissing contest stuff is so important. To
> Mac users, it's just not important which computer is faster.

This is app-specific. For most users, your statement is true.

But not for Lightwave/Maya users, serious Photoshop uses, FPS game
players.

Dual 1.42GHz is still a pretty fine machine for those above tasks, but
the ability to get the best tool available really opens the wallet.

> If it were, most would have jumped ship long since.

The above markets *have* jumped. Isn't Adobe selling 50-50 between the
platforms?

> To "potential Mac customers" all
> that matters (speed-wise) is the clock speed. As long as the clock-speed
> numbers don't match, the one with the lower number is seen as not as
> powerful. Things like machine throughput, memory pipeline speeds, and
> other enhancements don't mean beans if the raw MHz on the Mac is still
> lower than the raw MHz on the Wintel side and it is.

Recently, with Itanium, Pentium M, Opteron, we're finding a market
where clockspeed itself isn't the deciding issue.

=Heywood=

imouttahere[at]mac.com (Heywood Mogroot) writes:

> Vareck Bostrom <v.bostrom[at]attbi.com> wrote in message news:<v.bostrom-BBB252.04405124062003[at]netnews.attbi.com>...
> > In article <dd5de929.0306240143.70cd5376[at]posting.google.com>,
> > > The only thing in Wintel defense is that, since Intel is
> > > compartmentally lame, their dual systems ("workstation") are lagging
> > > their P4 systems at the moment.
> > >
> > > But the dual 3.06 vs. dual 2.0 on that Photoshop test went pretty
> > > unexpectedly bad for Intel. Megahertz myth indeed. Why is Intel so
> > > shitty?
> >
> > Perhaps we're seeing the advantage of an (effective) 1 GHz FSB vs an
> > (effective) 533 MHz FSB?
>
> Yes, but the benchmarking on 800Mhz FSB systems isn't that big a
> difference from 533.
>
> If you watch the keynote, the dual XP system has got one CPU pegged
> and the other flipping to 0->100% when it stutters on the audio (this
> test is towards the end of the show).
>
> > Not to mention the PPC970 is, after all, a
> > pretty good chip.
> >
> > What kind of cache is on the 970s anyway?
>
> developer.apple.com/technotes/tn/tn2087.html lays it all out.
>
> Twice the cacheline size of a G4 (128 bytes)
> Twice the L1 instruction cache (64K, direct-mapped, whatever that
> means)

> Same size L1 data cache (32K) but write-thru, 2-way associative, WTM)
> Twice the L2 cache (512K)

In effect, direct-mapped means if you have two memory
lines that both map to the same cache line, loading one
forces the other out. If by unhappy coincidence those
two memory lines are both in active use, the result is
poor performance.

2-way associative means that the two memory lines can
share the same cache line. If you happen to be actively
using three memory lines that all map to the same cache
line, again you suffer. But this happens less often.

Direct-mapped caches allow faster lookup, so there is
a tradeoff to be made.

-SEan

In article <24fhfv4qsud7tqd8bs76t2hd7qs5binpqj[at]4ax.com>,
Steve Hanson <icustomercare[at]usps.com> wrote:

> George Graves wrote in
> <gmgraves-7798C5.11123624062003[at]newssvr16-ext.news.prodigy.com>:
>
> >In article <flip-437A63.10093524062003[at]nnrp03.earthlink.net>,
> > Flip <flip[at]flippo.com> wrote:
> >
> >> In article <pbpgfvgqhgar9iam48uep2lr5ue2v09qdo[at]4ax.com>,
> >> Steve Hanson <icustomercare[at]usps.com> wrote:
> >>
> >> > Heywood Mogroot wrote in
> >> > <dd5de929.0306240143.70cd5376[at]posting.google.com>:
> >> >
> >> > >no <no[at]no.com> wrote in message
> >> > >news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
> >> > >> Strange, when one goes to www.spec.org to get the officially obtained
> >> > >> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
> >> > >> the int test) the scores for the dell 650 ARE NOT LOWER than the dual
> >> > >> g5
> >> > >> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC,
> >> > >> even
> >> > >> using Apple's definition
> >> > >
> >> > >Please. SPEC is for penis measurement ("see! I've got 1/8" on you!
> >> > >neener neener"). Watching the keynote, especially the audio app,
> >> > >showed Dell's dual Xeon getting down on all fours and barking like the
> >> > >dog that it is.
> >> > >
> >> > >The only thing in Wintel defense is that, since Intel is
> >> > >compartmentally lame, their dual systems ("workstation") are lagging
> >> > >their P4 systems at the moment.
> >> > >
> >> > >But the dual 3.06 vs. dual 2.0 on that Photoshop test went pretty
> >> > >unexpectedly bad for Intel. Megahertz myth indeed. Why is Intel so
> >> > >shitty?
> >> >
> >> > Only an idiot would point to a carefully prepared keynote presentation
> >> > as if it showed a performance test under lab conditions. It would
> >> > appear you are that idiot.
> >>
> >> So you're saying that the Presidents of Wolfram, Adobe, and a couple
> >> other companies were lying when they said that the Mac was faster than
> >> the PCs?
> >
> >Musta' been PAID by Apple to lie. It's the only logical explanation. ;->
>
> Careful, George, that argument has seen a lot of use on csma...by Mac
> advocates wishing to impugn any report that showed off a sluggish Mac.
> You don't want to throw away your only form of mental defense against
> reality, do you?

Really? It has? Care to produce some examples?

--
Alan Baker
Vancouver, British Columbia
"If you raise the ceiling 4 feet, move the fireplace from that wall
to that wall, you'll still only get the full stereophonic effect
if you sit in the bottom of that cupboard."

In article <dd5de929.0306241632.e3d8f6b[at]posting.google.com>,
imouttahere[at]mac.com (Heywood Mogroot) wrote:

> George Graves <gmgraves[at]pacbell.net> wrote in message
> news:<gmgraves-1CC564.11194724062003[at]newssvr16-ext.news.prodigy.com>...
> > Don't understand why all this pissing contest stuff is so important. To
> > Mac users, it's just not important which computer is faster.
>
> This is app-specific. For most users, your statement is true.
>
> But not for Lightwave/Maya users, serious Photoshop uses, FPS game
> players.
>
> Dual 1.42GHz is still a pretty fine machine for those above tasks, but
> the ability to get the best tool available really opens the wallet.
>
> > If it were, most would have jumped ship long since.
>
> The above markets *have* jumped. Isn't Adobe selling 50-50 between the
> platforms?

Think about that product split before you congratulate yourself on
coming up with it. 2% of the computer marketplace responsible for 50%
sales? Hell, if every corporate communications department in the country
bought one copy of Photoshop or Illustrator (or whatever else) it would
MORE than account for the 50% of product sold to the Wintel platform
without ever touching a REAL professional user.

> > To "potential Mac customers" all
> > that matters (speed-wise) is the clock speed. As long as the clock-speed
> > numbers don't match, the one with the lower number is seen as not as
> > powerful. Things like machine throughput, memory pipeline speeds, and
> > other enhancements don't mean beans if the raw MHz on the Mac is still
> > lower than the raw MHz on the Wintel side and it is.
>
> Recently, with Itanium, Pentium M, Opteron, we're finding a market
> where clockspeed itself isn't the deciding issue.

It is with potential converts.

--
George Graves

In article <BB1EB0B7.121C2%snail[at]easytype.ch>, Sven Mumenthaler wrote:
>> Strange, when one goes to www.spec.org to get the officially obtained
>> spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
>> the int test) the scores for the dell 650 ARE NOT LOWER than the dual g5
>> 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
>> using Apple's definition
>
> There is a very interesting article dealing with this issue:
>
> http://www.haxial.com/spls-soapbox/apple-powermac-G5/
>
> It seems that Apple did cheat on the benchmarks...

I read that article, and that guy was full of a lot of crap. Well, I'm
being harsh by saying that, but he didn't say the full story (for
example, on certain SPEC tests, you get better performance by turning
*off* hyperthreading on the P4, which is something both Dell and Apple
did when they did their benchmarking).

Apple's VP has a good followup to that posted on slashdot -- it's worth
a gander to see the other side of the story.

--

-------------------- http://www.techhouse.org/lou ----------------------
"Dragonmaster Lou" | "Searching for a distant star, heading off to
lou at techhouse org | Iscandar, leaving all we love behind, who knows
Tech House Alum | what dangers we'll find..."
-----------------------------------------------------------------------

In article
<gmgraves-1CC564.11194724062003[at]newssvr16-ext.news.prodigy.com>,
George Graves <gmgraves[at]pacbell.net> wrote:

> To "potential Mac customers" all
> that matters (speed-wise) is the clock speed. As long as the clock-speed
> numbers don't match, the one with the lower number is seen as not as
> powerful.

Pretty much analogous to automotive horsepower.


- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
"We really haven't done everything we could to protect our
customers ... Our products just aren't engineered for security."
-- Brian Valentine, Sr VP in charge of MS Windows' Dev Team

In article
<ThomRosario-A9A3BF.11522525062003[at]news-central.giganews.com>,
Thom Rosario <ThomRosario[at]nospam.com> wrote:

> In article
> <gmgraves-1CC564.11194724062003[at]newssvr16-ext.news.prodigy.com>,
> George Graves <gmgraves[at]pacbell.net> wrote:
>
> > To "potential Mac customers" all
> > that matters (speed-wise) is the clock speed. As long as the clock-speed
> > numbers don't match, the one with the lower number is seen as not as
> > powerful.
>
> Pretty much analogous to automotive horsepower.

And like automotive horsepower, it's largely in an of itself a useless
benchmark for much of anything (except in a car engine it might be
indicitive of fuel consumption).

--
George Graves

In article <vfifkjrp71jbf5[at]corp.supernews.com>,
Michael Rice <marice[at]whiteice.com> wrote:

> On 6/24/2003 1:19 PM, George Graves wrote:
> [...]
> >
> > Don't understand why all this pissing contest stuff is so important. To
> > Mac users, it's just not important which computer is faster. If it were,
> > most would have jumped ship long since. To "potential Mac customers" all
> > that matters (speed-wise) is the clock speed. As long as the clock-speed
> > numbers don't match, the one with the lower number is seen as not as
> > powerful. Things like machine throughput, memory pipeline speeds, and
> > other enhancements don't mean beans if the raw MHz on the Mac is still
> > lower than the raw MHz on the Wintel side and it is.
> >
>
> I am a "potential Mac customer", and I've seen that for my apps
> (compiling software, number crunching and OpenGL viz), it's all about
> raw clock speed and video performance. I've done real-world benchmarks
> between Mac and Intel and the systems are about the same, MHz for MHz.
> The Mac falls down a bit in memory bandwidth and OpenGL driver
> performance, but overall they are comparable. Basically, a 3x faster
> Intel chip translates (almost linearly) to increased productivity in
> those specific tasks.
>
> However, operating envrionment is a big factor. Having cleaner, tightly
> integrated, better engineered OS and productivity tools is where Apple
> and Mac really shine. The G5 will narrow the CPU power chasm, which
> should, at least, put Mac back on the playing field.
>

Like I said, to potential Mac customers all that matters is raw clock
speed.

--
George Graves

In article
<gmgraves-239AA6.11541625062003[at]newssvr14-ext.news.prodigy.com>,
George Graves <gmgraves[at]pacbell.net> wrote:

> In article
> <ThomRosario-A9A3BF.11522525062003[at]news-central.giganews.com>,
> Thom Rosario <ThomRosario[at]nospam.com> wrote:
>
> > In article
> > <gmgraves-1CC564.11194724062003[at]newssvr16-ext.news.prodigy.com>,
> > George Graves <gmgraves[at]pacbell.net> wrote:
> >
> > > To "potential Mac customers" all
> > > that matters (speed-wise) is the clock speed. As long as the clock-speed
> > > numbers don't match, the one with the lower number is seen as not as
> > > powerful.
> >
> > Pretty much analogous to automotive horsepower.
>
> And like automotive horsepower, it's largely in an of itself a useless
> benchmark for much of anything (except in a car engine it might be
> indicitive of fuel consumption).

My point, exactly.


- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
"We really haven't done everything we could to protect our
customers ... Our products just aren't engineered for security."
-- Brian Valentine, Sr VP in charge of MS Windows' Dev Team

On 6/25/2003 1:56 PM, George Graves wrote:
> In article <vfifkjrp71jbf5[at]corp.supernews.com>,
> Michael Rice <marice[at]whiteice.com> wrote:
>
>
>>On 6/24/2003 1:19 PM, George Graves wrote:
>>[...]
>>
>>>Don't understand why all this pissing contest stuff is so important. To
>>>Mac users, it's just not important which computer is faster. If it were,
>>>most would have jumped ship long since. To "potential Mac customers" all
>>>that matters (speed-wise) is the clock speed. As long as the clock-speed
>>>numbers don't match, the one with the lower number is seen as not as
>>>powerful. Things like machine throughput, memory pipeline speeds, and
>>>other enhancements don't mean beans if the raw MHz on the Mac is still
>>>lower than the raw MHz on the Wintel side and it is.
>>>
>>
>>I am a "potential Mac customer", and I've seen that for my apps
>>(compiling software, number crunching and OpenGL viz), it's all about
>>raw clock speed and video performance. I've done real-world benchmarks
>>between Mac and Intel and the systems are about the same, MHz for MHz.
>>The Mac falls down a bit in memory bandwidth and OpenGL driver
>>performance, but overall they are comparable. Basically, a 3x faster
>>Intel chip translates (almost linearly) to increased productivity in
>>those specific tasks.
>>
>>However, operating envrionment is a big factor. Having cleaner, tightly
>>integrated, better engineered OS and productivity tools is where Apple
>>and Mac really shine. The G5 will narrow the CPU power chasm, which
>>should, at least, put Mac back on the playing field.
>>
>
>
> Like I said, to potential Mac customers all that matters is raw clock
> speed.
>

Your implication is that clock speed is irrelevant. Guess what - it's
not! I'm not just talking CPU either - FSB and memory are also critical.

Most "potential Mac customers" are currently Intel customers. Take a
peek at the few benchmarks that are available - the current single CPU
high-end Mac gets whipped by high-end PCs (e.g., 1x1.4GHz Mac vs 1x3GHz
P4). Obviously things get closer if you use a dual CPU Mac. But for
memory throughput, look at the STREAM benchmark - the current Mac can't
compete.

The common rebuttal is: "well thost are just benchmarks and they don't
mean anything in the real world". That's why I ran my own tests (CPU
intensive apps that test both in-cache and out-of-cache, OpenGL fps
tests, etc.) using my real-world applications on a PowerBook 1GHz. My
results were similar to the "standard" benchmarks. With those kinds of
results, the current Mac is just not a viable solution for CPU or memory
intensive applications, which is a much larger market than most people
believe. That's why I am, like everyone else, so elated that Apple
finally made a move.

On 6/25/2003 8:22 PM, George Graves wrote:
> In article <vfkf2vrdor1id1[at]corp.supernews.com>,
> Michael Rice <marice[at]whiteice.com> wrote:
>
>
>>On 6/25/2003 1:56 PM, George Graves wrote:
>>
>>>In article <vfifkjrp71jbf5[at]corp.supernews.com>,
>>> Michael Rice <marice[at]whiteice.com> wrote:
>>>
>>>
>>>
>>>>On 6/24/2003 1:19 PM, George Graves wrote:
>>>>[...]
>>>>
>>>>
>>>>>Don't understand why all this pissing contest stuff is so important. To
>>>>>Mac users, it's just not important which computer is faster. If it were,
>>>>>most would have jumped ship long since. To "potential Mac customers" all
>>>>>that matters (speed-wise) is the clock speed. As long as the clock-speed
>>>>>numbers don't match, the one with the lower number is seen as not as
>>>>>powerful. Things like machine throughput, memory pipeline speeds, and
>>>>>other enhancements don't mean beans if the raw MHz on the Mac is still
>>>>>lower than the raw MHz on the Wintel side and it is.
>>>>>
>>>>
>>>>I am a "potential Mac customer", and I've seen that for my apps
>>>>(compiling software, number crunching and OpenGL viz), it's all about
>>>>raw clock speed and video performance. I've done real-world benchmarks
>>>>between Mac and Intel and the systems are about the same, MHz for MHz.
>>>>The Mac falls down a bit in memory bandwidth and OpenGL driver
>>>>performance, but overall they are comparable. Basically, a 3x faster
>>>>Intel chip translates (almost linearly) to increased productivity in
>>>>those specific tasks.
>>>>
>>>>However, operating envrionment is a big factor. Having cleaner, tightly
>>>>integrated, better engineered OS and productivity tools is where Apple
>>>>and Mac really shine. The G5 will narrow the CPU power chasm, which
>>>>should, at least, put Mac back on the playing field.
>>>>
>>>
>>>
>>>Like I said, to potential Mac customers all that matters is raw clock
>>>speed.
>>>
>>
>>Your implication is that clock speed is irrelevant. Guess what - it's
>>not! I'm not just talking CPU either - FSB and memory are also critical.
>
>
> I have implied nothing. What I SAID was that most Mac users don't care
> that much about speed, if they did, they'd have jumped platforms by now.
>
>
>>Most "potential Mac customers" are currently Intel customers. Take a
>>peek at the few benchmarks that are available - the current single CPU
>>high-end Mac gets whipped by high-end PCs (e.g., 1x1.4GHz Mac vs 1x3GHz
>>P4). Obviously things get closer if you use a dual CPU Mac. But for
>>memory throughput, look at the STREAM benchmark - the current Mac can't
>>compete.
>
>
> And most of us Mac users don't really care. But Wintel users do. That's
> why I said that:
> "To "potential Mac customers*" all that matters (speed-wise) is the
> clock speed. As long as the clock-speed numbers don't match, the one
> with the lower number is seen as not as powerful." Was I wrong?
>
>
>>The common rebuttal is: "well thost are just benchmarks and they don't
>>mean anything in the real world".
>
>
> You seem to be missing my point. If Mac users cared that much about
> those benchmarks, I should think that they would no longer be Mac users.
>
>
>>That's why I ran my own tests (CPU
>>intensive apps that test both in-cache and out-of-cache, OpenGL fps
>>tests, etc.) using my real-world applications on a PowerBook 1GHz. My
>>results were similar to the "standard" benchmarks. With those kinds of
>>results, the current Mac is just not a viable solution for CPU or memory
>>intensive applications, which is a much larger market than most people
>>believe. That's why I am, like everyone else, so elated that Apple
>>finally made a move.
>
>
> To many of us, the fact that the new Macs are faster than the last
> generation of Macs is good news. That they are are faster or slower than
> Wintel computers is completely irrelevant for many of us.
>
>
> *Current Wintellers
>

We agree on that point - I also couldn't care less what the GHz rating
of the CPU is. The only thing I care about is achieved throughput; for
a long time the Intel chips have been far in front of the Mac. The G5
should change that.

In article <vfkjvu773348b7[at]corp.supernews.com>,
Michael Rice <marice[at]whiteice.com> wrote:

> On 6/25/2003 8:22 PM, George Graves wrote:
> > In article <vfkf2vrdor1id1[at]corp.supernews.com>,
> > Michael Rice <marice[at]whiteice.com> wrote:
> >
> >
> >>On 6/25/2003 1:56 PM, George Graves wrote:
> >>
> >>>In article <vfifkjrp71jbf5[at]corp.supernews.com>,
> >>> Michael Rice <marice[at]whiteice.com> wrote:
> >>>
> >>>
> >>>
> >>>>On 6/24/2003 1:19 PM, George Graves wrote:
> >>>>[...]
> >>>>
> >>>>
> >>>>>Don't understand why all this pissing contest stuff is so important. To
> >>>>>Mac users, it's just not important which computer is faster. If it were,
> >>>>>most would have jumped ship long since. To "potential Mac customers" all
> >>>>>that matters (speed-wise) is the clock speed. As long as the clock-speed
> >>>>>numbers don't match, the one with the lower number is seen as not as
> >>>>>powerful. Things like machine throughput, memory pipeline speeds, and
> >>>>>other enhancements don't mean beans if the raw MHz on the Mac is still
> >>>>>lower than the raw MHz on the Wintel side and it is.
> >>>>>
> >>>>
> >>>>I am a "potential Mac customer", and I've seen that for my apps
> >>>>(compiling software, number crunching and OpenGL viz), it's all about
> >>>>raw clock speed and video performance. I've done real-world benchmarks
> >>>>between Mac and Intel and the systems are about the same, MHz for MHz.
> >>>>The Mac falls down a bit in memory bandwidth and OpenGL driver
> >>>>performance, but overall they are comparable. Basically, a 3x faster
> >>>>Intel chip translates (almost linearly) to increased productivity in
> >>>>those specific tasks.
> >>>>
> >>>>However, operating envrionment is a big factor. Having cleaner, tightly
> >>>>integrated, better engineered OS and productivity tools is where Apple
> >>>>and Mac really shine. The G5 will narrow the CPU power chasm, which
> >>>>should, at least, put Mac back on the playing field.
> >>>>
> >>>
> >>>
> >>>Like I said, to potential Mac customers all that matters is raw clock
> >>>speed.
> >>>
> >>
> >>Your implication is that clock speed is irrelevant. Guess what - it's
> >>not! I'm not just talking CPU either - FSB and memory are also critical.
> >
> >
> > I have implied nothing. What I SAID was that most Mac users don't care
> > that much about speed, if they did, they'd have jumped platforms by now.
> >
> >
> >>Most "potential Mac customers" are currently Intel customers. Take a
> >>peek at the few benchmarks that are available - the current single CPU
> >>high-end Mac gets whipped by high-end PCs (e.g., 1x1.4GHz Mac vs 1x3GHz
> >>P4). Obviously things get closer if you use a dual CPU Mac. But for
> >>memory throughput, look at the STREAM benchmark - the current Mac can't
> >>compete.
> >
> >
> > And most of us Mac users don't really care. But Wintel users do. That's
> > why I said that:
> > "To "potential Mac customers*" all that matters (speed-wise) is the
> > clock speed. As long as the clock-speed numbers don't match, the one
> > with the lower number is seen as not as powerful." Was I wrong?
> >
> >
> >>The common rebuttal is: "well thost are just benchmarks and they don't
> >>mean anything in the real world".
> >
> >
> > You seem to be missing my point. If Mac users cared that much about
> > those benchmarks, I should think that they would no longer be Mac users.
> >
> >
> >>That's why I ran my own tests (CPU
> >>intensive apps that test both in-cache and out-of-cache, OpenGL fps
> >>tests, etc.) using my real-world applications on a PowerBook 1GHz. My
> >>results were similar to the "standard" benchmarks. With those kinds of
> >>results, the current Mac is just not a viable solution for CPU or memory
> >>intensive applications, which is a much larger market than most people
> >>believe. That's why I am, like everyone else, so elated that Apple
> >>finally made a move.
> >
> >
> > To many of us, the fact that the new Macs are faster than the last
> > generation of Macs is good news. That they are are faster or slower than
> > Wintel computers is completely irrelevant for many of us.
> >
> >
> > *Current Wintellers
> >
>
> We agree on that point - I also couldn't care less what the GHz rating
> of the CPU is. The only thing I care about is achieved throughput; for
> a long time the Intel chips have been far in front of the Mac. The G5
> should change that.
>
So, This new Mac is faster than last year's? Great. Nice to see that
the science of personal computers is advancing -its about time.
Computers have been stuck with the G4 at [at]1.5 GHz for long enough.

--
George Graves

In article <flippo-E21D27.07070024062003[at]news.central.cox.net>,
flip <flippo[at]mac.com> wrote:

> In article <v.bostrom-BBB252.04405124062003[at]netnews.attbi.com>,
> Vareck Bostrom <v.bostrom[at]attbi.com> wrote:
>
> > In article <dd5de929.0306240143.70cd5376[at]posting.google.com>,
> > imouttahere[at]mac.com (Heywood Mogroot) wrote:
> >
> > > no <no[at]no.com> wrote in message
> > > news:<no-FA05F2.23593823062003[at]newssvr19-ext.news.prodigy.com>...
> > > > Strange, when one goes to www.spec.org to get the officially obtained
> > > > spec numbers, one sees that for SPECfp_rate_base2000 (the mac lost in
> > > > the int test) the scores for the dell 650 ARE NOT LOWER than the dual
> > > > g5
> > > > 2 gig. As far as SPEC is concerned, the g5 IS NOT the fastest PC, even
> > > > using Apple's definition
> > >
> > > Please. SPEC is for penis measurement ("see! I've got 1/8" on you!
> > > neener neener"). Watching the keynote, especially the audio app,
> > > showed Dell's dual Xeon getting down on all fours and barking like the
> > > dog that it is.
> > >
> > > The only thing in Wintel defense is that, since Intel is
> > > compartmentally lame, their dual systems ("workstation") are lagging
> > > their P4 systems at the moment.
> > >
> > > But the dual 3.06 vs. dual 2.0 on that Photoshop test went pretty
> > > unexpectedly bad for Intel. Megahertz myth indeed. Why is Intel so
> > > shitty?
> > >
> > > =Heywood=
> >
> > Perhaps we're seeing the advantage of an (effective) 1 GHz FSB vs an
> > (effective) 533 MHz FSB? Not to mention the PPC970 is, after all, a
> > pretty good chip.
> >
> > What kind of cache is on the 970s anyway?
>
> IIRC, it's 512 kb L2 and 2 MB L3 - at least on the 2 GHz model.

No L3 cache on the 970s.

I don't think there is any technology that is significantly faster than
2 400MHz DDRAM chips running in parallel and that is affordable at the
same time. The big iron POWER4 systems use DRAM for L3 cache (each bit
of memory made of six transistors or so, really fast and really
expensive).

The L3 cache on a current G4 Mac runs significantly slower than main
memory on the G5.

Christian Bau <christian.bau[at]cbau.freeserve.co.uk> wrote in message news:<christian.bau-6EF359.00365830062003[at]slb-newsm1.svr.pol.co.uk>...
> The L3 cache on a current G4 Mac runs significantly slower than main
> memory on the G5.

but with much higher latencies.

The on-die DDR-333 memory controllers on the Opteron have much less
latency than the G5's FSB <-> DDR-400 path.

I forget where I read this or I'd post the link.

=Heywood=