Genesi: Pegasos III with PPC 970 announced

[BlackMonk]

Hey amigamad, after the G5, there will be a G6, after the G6, there will be a G7.

Yeah, but the G5 is such an overall improvement to the PPC line that his argument has validity.  The G3/G4 has been crippled for years.  Going from a G3 to a G4 is an incremental improvement, at best.  Going from a G3/G4 to a G5 has alot more impact.

Why get a system based on a CPU that is 6 years old (G3 was introduced in 1997, I believe) and still crippled by a FSB from 1999 (introduction of 100 MHz FSB)?  I must claim some ignorance, but is the G3 available at higher that 100 MHz even now?

The G4 harkens back to 1999.  4 years old.  The G4 finally got a 133 MHz FSB back in the first month of 2001.  So that's basically 2 and a half years old.  But hell, I guess the Pegasos is reaching back to 1999 for its processor bus as it's listed at 100 MHz anyway:

http://www.pegasosppc.com/tech_specs.php

So please tell me again, why is waiting a few more months for such a generational increase in power (PegPower, right?) and overall system performance such a foolish notion?

Now if, say, it was "wait 5 more months for a faster clocked G4", ok, you got a good point there.  But it's like saying, "you're stupid to wait 5 months for a 64-bit CPU with one of the best FSB's on the market and some of the best performance currently shipping."  Especially when you compare it to the under-performing G3/G4 line which has stagnated for years.

But hey, I guess low power consumption is also an issue.  The question then is, will the Pegasos 3 be underclocked to lower the power requirements?  I mean, there's got to be SOME reason that Apple's throwing in a 450 watt power supply even for the low-end 1.6 GHz G5/PPC970/GPUL/whatever you want to call it.

http://developer.apple.com/documentation/Hardware/Developer_Notes/Macintosh_CPUs-G5/PowerMacG5/PowerMacG5.pdf

Page 12 for details.  Maybe it's the AGP Pro slot that's sucking all that extra juice, eh?

Well, as long as you don't go over 1.2 GHz (18 watts) you should be ok.  If you want performance?  1.8 GHz is sucking 42 watts.  That's more in line with the 1 GHz G4e at 30 watts.  

So should people not wait because of such an improvement?  Or should people say, "oh, it won't be as power efficient for performance as the G4s and G3s will, I'll just get a G3 instead?"

Or will that timeframe put the Pegasos 3 up against cheap systems from IBM themselves?

http://arstechnica.com/archive/news/1058761450.html

Perhaps Mr. Buck is trying to schmooze with IBM and convince them to supply 'em with cheap CPUs and chipsets or something?

http://arstechnica.com/archive/news/1058552694.html

Economies of scale, after all.

But Mark, you certainly cannot fault anyone for holding off when the next product is such a monumental improvement over the current generation of PowerPC products?  Can anyone, honestly, justify getting a G3 when there might be a G5-based Pegasos only months away?  And probably selling for around the same price point?

Most of y'all have been using tricked-out Amigas for years.  YEARS.  What's another few months for an actual cutting-edge product?  

I'm not saying the AmigaOne is a good buy, either.  But if I were an Amiga loyalist and AInc announced the AmigaPPC970 due in the next half year or year, I sure as hell wouldn't waste my money on an AmigaOne.

And certainly, Mark, it's not like only a few people here have expressed more interest in a G5 version of the Pegasos.  Mr. Buck linked to an explanation of why they are pursuing this line--very lackluster response to Pegasos 2.  And a larger demand for a Pegasos 3, I'm assuming with something other than a run-down G4 as the CPU.

I must admit, though, I have a hard time grasping this concept:

Just remember we are trying to eliminate CPU speed from the marketing discussion/buying decision. The issue is what WORK (activity exerted through power or faculty applied to perform a task) can be accomplished in the time allotted -- not how fast we ran to accomplish the work!

What work we can accomplish in the time allotted.  Versus how fast we ran to accomplish the work.

Would it not seem, from a requirements point of view, that needing X amount of work done in Y amount of time would require at least Z speed?  If Pegasos or Pegasos 2 do not supply at least Z speed, should you settle for less work done in the allotted time?

Why are you trying to eliminate CPU speed from the buying decision and marketing aspect?  Amount of work done in a set amount of time is intimately tied to the speed of a system.  

I guess you're trying to turn expectations on their head, i.e. getting people to look at it from an end-product perspective (you got something done, aren't you happy?).  But it seems that most people tend to look at things in a "how fast can I get this done so I can move on to other things" perspective (you got something done in half the time it would have taken you if you had used a competing product, aren't you happy?)

Next thing we know, the Pegasos will come with PR ratings like the Athlons.  You know, to take speed out of the equation.

[BlackMonk]

Ah, and to compare with the x86 world--their products haven't had such a leap in development through time.  As in, they didn't get stalled at one FSB or MHz rating for a few YEARS.

Thus, going from a 3 GHz P4 with a 400 MHz FSB to one with an 800 MHz FSB isn't as big of a leap as, say, going from a 100 MHz FSB G3 at 600 MHz to a 1.2 GHz G5 with a 600 MHz FSB.  Same thing with AMD's offerings, even they are at a 400 MHz FSB with their next product offering 800 MHz.

And before someone jumps on my back, yes, I've simplified it.  Instead of saying the Athlon uses a 200 MHz DDR FSB for an effective rate of 400 MHz and the Athlon64 uses a hypertransport link with an effective bandwidth different from what you'd expect with an 800 MHz signal, blah blah plus it integrates the memory controller on-die, blah blah P4 uses double pumped FSB, etc.  Hopefully that will head criticizms that I'm mistaken on the technical details off at the pass.  Remember:  I simplified only to get the point across and for readability.  I could throw in technically exact details into this post and the one above but it would make the long-ass posts even LESS readable.

[BlackMonk]

It'll be interesting to see what chipset BBRV will be using with the PPC 970. Mainly because the frontside bus is designed by Apple! So, if you want to use the high speed bus, you have to get either a chipset from apple, or a licence from apple.

http://www.arstechnica.com/cpu/03q2/ppc970-interview/ppc970-interview-2.html

Hannibal: You guys worked with Apple on the supporting logic--the chipset--for this thing, right? Or is this something that you did that is going into Apple boxes?

Peter Sandon: The chipset is an Apple design, and we're fabbing it.

Hannibal: So what are you guys using in the blades, then--the 970 blades? A chipset of your own design?

Peter Sandon: I haven't kept up with what we've announced about that, so I guess I don't know.

--------------

So maybe IBM will have their own chipset solution, as well.  Then again perhaps a 3rd party will magically make yet another solution, too.  I think that's an issue that's more of a "wait and see" kind of thing.  There's just not enough info at this point.