| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1895.1 | It's a low end machine :-) | WRKSYS::SCHUMANN | | Sun Mar 16 1997 16:34 | 14 |
| Firstly, the Miata program (until very recently) was intended to deliver
our LOW END Alpha workstations. That's why the Miata performance is where
it is. For a low end machine, the performance is very good!!
Since it appears now that EV6 will not ship this year, there is
interest in extending Miata to be a high end machine. We are working on
a variety of things that will make Miata even faster. Bigger cache is
on the list.
Please forward you concerns to Keith Bellamy, since there is limited
engineering budget money available and it appears that some programs may
be cut back.
--RS
|
| 1895.2 | | BIGUN::nessus.cao.dec.com::Mayne | Churchill's black dog | Sun Mar 16 1997 20:23 | 6 |
| > Since it appears now that EV6 will not ship this year
First I've heard of this. What's the story?
PJDM
|
| 1895.3 | Miata is a low-end machine :-) | WRKSYS::SCHUMANN | | Mon Mar 17 1997 08:57 | 19 |
| [Notes 1895.1 and 1895.2 have been set hidden by the moderator because they
discussed sensitive subject matter. This is reply .1, with the sensitive
sentence deleted... (.2 is merely a request for more information about the
sensitive material.)]
Firstly, the Miata program (until very recently) was intended to deliver
our LOW END Alpha workstations. That's why the Miata performance is where
it is. For a low end machine, the performance is very good!!
To tide us over until EV6 arrives, there is interest in extending Miata to be a
high end machine. We are working on a variety of things that will make Miata
even faster. Bigger cache is on the list.
Please forward you concerns to Keith Bellamy, since there is limited
engineering budget money available and it appears that some programs may
be cut back.
--RS
|
| 1895.4 | Look beyond the single geometric mean! | PERFOM::LICEA_KANE | when it's comin' from the left | Mon Mar 17 1997 13:57 | 49 |
| | Have we totally lost it or what?
Or what.
| This is the same as the AlphaStations 500/500 Specs we have been
| shipping already.
No, it is not the same.
There are *10* benchmarks in the SPEC CFP95 suite.
They are run with two different set of run rules, peak and base.
So there are a total of 20 results to be considered. (Not just
the *one* SPECfp95.)
At SPEC CPU95, looking at ALL 20 RESULTS (peak and base of the 10
benchmarks) a "Miata" is as much as (approximately) 30% faster than a
"Bret"/500. It is also as much as (approximately) 30% slower
than a "Bret"/500.
More often than not, at the 20 results, a "Miata" can meet or beat
a "Bret"/500.
Why? Memory bandwidth, memory bandwidth, memory bandwidth.
(Note that HP's C160 and C180, and SGI's Octane can also beat a "Bret"/500
at some of the individual CPU95 results - but also lose on them as
well. SGI's O2/R10000 is a turkey that loses to just about everything.
Then we have a couple of HALstations. That's followed by a Sun Ultra
at 200MHz which can beat us on only *ONE* CPU95 result. Same
with an IBM RS6000/591.)
Only in Digital could there be the simultaneous arguments that:
It is not possible to sell a "Bret" because of the price of the
"Miata."
*AND*
It is not possible to sell a "Miata" because of the performance
of the "Bret."
(Or worry that the "Bret" is stuck at *ONLY* 500MHz and the "Miata"
is stuck at *ONLY* 4MB cache.)
-mr. bill
|
| 1895.5 | s | SCASS1::MARIA | John Maria | Mon Mar 17 1997 21:44 | 37 |
| I don't know about the rest of you workstation sales people out there,
but I am not losing deals because of performance. I have won a number
of opportunities because I had the right mix of applications, the right
support (ISV or Digital), and the right price.
In almost every case where I have lost we kicked the competitions butt,
sometimes running their own benchmarks. I lost because I didn't run the
applications they needed (Catia, Softdesk, SDRC, PDGS etc.), I didn't
have the right partner, or I did not know the competitions price.
At major wheel manufacturer we blew away an HP C180, UltraSPARC II, and
an SGI O2 R10000. We lost 70 units to SUN because they ran Catia, PDGS,
and the latest versions of SDRC.
At a heavy equipment manufacturer we lost 80 seats bidding a DPW180i to M
icron. The VAR knew the industry better, and Micron had a better price.
Yes, a few customers need the hottest box on we can offer, but in my
experience even those customers are buying less than 2% of boxes in
that performance range.
I can't even get a major helicopter manufacturer to evaluate an Alpha
because it does not run Catia.
I am not made aware of workstation opportunities in our installed base
because we are not incenting our account managers for workstation
business (oops that is for another conference)
.0 you and I know SUN and SGI kicked butt at Los Alamos and Sandia in
an era where Digital and SGI had the hottest boxes. You kicked butt in
those accounts with DECstations! !!!running ULTRIX!!!
My take is... don't let up on the performance, but let's put the effort
where it can really boost sales!
John Maria
WAM
|
| 1895.6 | Will we win for twelve months? | SWAM1::POIANI_MI | | Tue Mar 18 1997 00:07 | 19 |
| I couldn't agree with you more on software driving all the sales,
however working with what we have (a limited number of Apps) Itseems
the only deals we can win are those opportunities when all the apps
line up (i think it's more than 2%maybe 5%). In those oppotrtunities
we win because our performance.
I also agree we have been kicking but on the week competitors today. I
was upset that down the road I couldn't see any performance increases
until Monet 12 Months from now. My opinion is that the only thing Ihad
(performance) to win was stagnent for 12 Months. If our performance
goes away and we don't have applications.....??? There goes the 5%.
I can only vividly remember the surprise of the Ultra Sparc, and how we
scrambled to redo the line (and we knew it was coming). Sun is telling
my customers there are more surprises coming from SUN.
Look I did get the info I needed from .3, I will give my input to
Kieth. I would also like to see the Spec breakdown for the Miata's, but
I could not find them anywhere, are they available?
|
| 1895.7 | SPEC CPU95 on Miata and on NT | PERFOM::HENNING | | Tue Mar 18 1997 06:40 | 15 |
| Miata has been extensively measured with SPEC95 running under an
operating system that has NOT been announced on Miata yet. Therefore
we haven't submitted or published the numbers yet.
Meanwhile, as to NT, the current state of the NT SPEC95 kit compared
to Unix on a Bret/500 is:
- SPECint_base95 better than Unix by about 9%
- SPECint95 not yet measured with a valid, submittable sw set
- SPECfp_base95 about 10% lower than Unix
- SPECfp95 about 15% lower than Unix
Both fp_base95 and fp95 are expected to improve within days, and int95
should improve within weeks.
/john
|