[rescue] Sun E6500 question

[Skeezics Boondoggle]

On Sun, 12 Jun 2005, Patrick Finnegan wrote:

> Patrick Finnegan declared on Sunday 12 June 2005 12:41 pm:
> > Ok, so I've got my E6500 now containing only 100MHz capable boards,
> > and 400MHz CPUs.  Still, the firmware sets the centerplane clock to be
> > 80MHz with a 5:1 CPU multiplier, instead of 100MHz with a 4:1
> > multiplier. Does anyone have a clue why this might be happening?  I'm
> > pretty sure that the clock board is 501-5365, so it should support
> > 100MHz operation I think.
> 
> Nevermind, I just realized that the 6500's centerplane only does 
> 83/90MHz.

[Mini-rant(tm) ahead!]

This has always bugged the hell out of me.  WHY not offer a 450Mhz US-II
module for the E6500 so it could run at 5:1?  They ship 480Mhz modules for
the E450... and 450Mhz modules for the E220/420R... and what the hell is
that 464/466Mhz module all about?  Rather than just ship a 450, we waited
a year(?!?) to get 14Mhz and require a 5:2 clock board?  WTF?  Sun's web
site used to tout that an early US-II fitted with a water cooler (or fan
or something) had reached 500Mhz - I think this was before they had
started shipping the US-II in volume in order the demonstrate how much
"headroom" there was in the chip for future clock rate enhancements...
When the 5:1 clock boards were announced, I figured a 500Mhz upgrade was
right behind... woulda been _sweet_ to bump up my 8-way E4500 to 500Mhz!  
I figured wrong.

So, Sun's "strategy" (which I used to think was quite sensible) is - or
was - to introduce pipeline changes in odd-numbered revs, process changes
in even numbered revs.  Right?  But in the US-III, they broke their own
rules... the II languished while we waited, and waited, and waited for the
III to come out - new pipeline AND the first to use copper, smaller
geometries, etc.  Grrrr.  *While* we were waiting, why not drop the II
from .18um to .13um and ramp it up a couple of notches speedwise?  Or,
even if they couldn't squeeze that many more Mhz out of the US-II
pipeline, AT LEAST take advantage of lowered heat dissipation, or maybe
even make some relatively small changes like bringing the L2 on-chip?  
Given the heat death and L2 E$ problems Sun had with the early 400MHz/8MB
modules, you'd think that a .13um "II+" could have solved those problems
quite nicely, spanning the gap until the US-III/SunFire rollout.

And the US-II had life left in it; apparently quite a bit, given that the
new Niagara multi-core chips are still using the US-II pipeline, not the
US-III!?

But, then, I guess the marketing department decided that trying to get
people to trade up their old Enterprise servers for SunFires was more
important than selling them upgrades.  Seeing the big jump up to 750Mhz
was probably supposed to help delineate what a Great Leap Forward the
SunFires were; faster US-IIs would have undermined that?  Meantime, in the
looooonnng gap before the SunFires finally appeared, though, I was one of
those loyal Sun customers finding fewer and fewer reasons to put off
purchasing Intel-based stuff while we waited for faster SPARCs to
arrive... (Fortunately, my E420Rs were kicking serious ass, so we bought
more used 420Rs while I waited for 280R pricing to come down out of the
clouds...)

And the crazy IIi/IIIi (someone please SHOOT the nutball who invented THAT
numbering scheme) came out in between, but I guess that whole series of
chips is on their own roadmap.  Oy...

While the US-II machines are "old news" and the SunFires occupy Sun's
focus (despite that sizeable numbers of E3k-E10k's are still out there in
production) we're seeing the same sort of problems and delays again, with
the III/IV "stuck" at 1.2Ghz for some time now... I'm hoping that the IV+
gets released soon, as promised - "middle of 2005" is _here_, guys - and
with performance numbers that Sun can use to hold off the critics long
enough to see these new multi-core chips see the light of day; they look
neat!

Of course, if the IV+ features a faster memory interface, then are we
waiting on a new lineup of *machines,* or will the IV+ drop in to the
existing US-III or US-IV sockets?  Who knows?  Maybe I'm just crazy, but
I'd rather hear SOMETHING about their progress rather than endure the
tight-lipped, "wait and see" approach Sun still takes with SPARC
development... small steps are fine with me, rather than lurching forward
every other year.  Ah, well.

Annnnnnyway, all I started out to say was, Patrick, I feel your pain about
the E6500.  :-)

While I just can't understand why Sun didn't put a 450Mhz module out with
the right form factor for that machine, but it's a damned nice box to play
with!

> Oh well.

Yup. :-/

-- Chris