>Firstly, the consumer doesnt need altivec if it means compromising speed
in non-altivec appz.
So you'd advocate a 5% gain in performance in integer apps (say) so the
consumer has a 50% longer wait in iMovie? In any case you seem to pull
this from nowhere. Where is a G3 faster than a G4 of equal clockspeed?
> In altivec enabled appz then yes its fine, but in
normal appz you might aswell have a G3.
... so the question posed is `is there a significant enough mass of
altivec-enabled apps now/in the near future'. As I pointed out already,
yes there is.
> and if you put a G4 instead of a G3 you will indeed be compromising
performance.
Where?
> Aslong as you consider that IBMs G3 is at 1Ghz and has a 400Mhz bus, and
going by the record of what the latest G4s are like the 7460 will be a
weak excuse for upping the clock speed.
Pfffht. A 2.5x multiplier? The CPU would be the bottleneck. 266 would do
just fine.
>You are right the average consumer is a tight ass, so putting a G3 in an
imac will cut on costs as opposed to putting a G4 in.
Apple saves what, $20 bucks per unit? I've already pointed out the
benefits of stronger float and altivec. Don't tell me that's not worth the
extra ~$20.
> aswell as the fact that the G3 if anything else will sound faster to the
average consumer.
... assuming the G3 can ramp up to higher clockspeeds than the G4.
>From what you "hear" of the 7460??? From what i know of the past G4s/
Which apple no longer uses.
>the current G4 design, i can tell you that the G4 does not scale well.
... which is why Apple is looking at the 7460 for low power and high
clockspeeds.
> So what the hell makes you think the 7460 will???
... because if the 7460 _didn't_ scale upward well, Apple wouldn't be
interested in it, would they?
> because its at 1.33 Ghz??? that isnt an indication of scalability joey.
True, the _yields_ Moto is getting at 1.33 ghz is the indication of
scalability. Moto's minting 7460s now. I'd be greatly surprised if they
weren't.
>you know aswell as i do i meant Graphics appz not graphics,
1) no, i didn't.
2) graphics apps would be greatly accelerated with altivec and multiple
execution units due to the parallelism of the apps in question.
>what is the relevance of the similarity of the G3 & G4 compared to what
the consumer needs???
Your argument centers around Apple 1) saving a few bucks per unit so they
can forgo beefier float and altivec, both of which they need, as i've
pointed out repeatedly and 2) gaining higher clockspeeds which the 7460
can do anyway.
>It doesnt matter how similar a chip is you can still make conclusions at
what each chip is good at.
... the G4 is a G3 w/altivec and better float. So they're ~equal in
integer apps, the G4 wins in simd and float. (of course there's
caveats, the processor with more execution units might win parallel
computation, cache sizes, pipeline design etc.)
>And as for IBMs vapourware???
Erm, i don't see any reports of IBM minting the 750FX. Remember Moto
announced their high-clockspeed G4s long before Apple started using them.
>tell me the G4 has a 400mhz bus and i will accept your statement.
"An incarnation of the G4 architecture could conceivably run on a FSB
clocked at 400 MHz".
erm, i hate to tell you this, the `bus' is on the logic board, genius.
`400 MHz' is only one of the FSB speeds the processor can accept, and it's
probably the highest (marketeering goes here).
>And after all that tell me joey, who has a better track record over
recent years. IBM or motorola?
You're only as good as your last game so their `track records' are
irrelevant -- IBM hasn't produced any FXs for Apple and the 7460/8500
rumors are too conservative to be false, Apple wants them ASAP.
>Sure apples media appz are nice with a G4,
... and many of them are intended at consumers, so it follows that Altivec
is a desired feature.
> but are the majority of consumer based appz altivec friendly???
Should they be? iTunes, iMovie, iDVD, anything that involves the same
operation repeated ad infinitum. So Apple has an interest in SIMD units.
> no they are not you blind sh*t.
Pfffht, Apple NEEDS altivec in the consumer machines if their consumer
apps (theirs and third-party) _can_ be vector-optimized. As I've already
pointed out repeatedly, YES, there is plenty of scope for altivec
enhancement (re examples above).
>Is it easy to create an altivec enabled app??? no it is not D*ckhead.
Altivec enabled apps can be developed in C. Lots of developers use CW for
example, its compiler works fine w/altivec.
Or you could try a vectorizing compiler. I don't know if one exists
though.
In any case it's easier than SSE/SSE-2.
>that the majority of my statements are right,
And precisely which statements of yours would they be?