Linus Torvalds (torvalds@linux-foundation.org) on 2/6/09 wrote:
---------------------------
>savantu (savantu@email.ro) on 2/6/09 wrote:
>>
>>If the Special Olympics is worth $20B per year and Intel
>>manages to get IPF to a 40-50% share in the next 5 years,
>
>What "next 5 years". You already used that argument in 2001.
>
>It's expired. Your argument is sitting in the corner of the
>fridge, and it's smelling bad. It's smelling so bad that
>people now are scared to even open the fridge door, much
>less take it out and throw it away.

In 2001 sales were 0. In 2008 sales were close to $5B a year.
Wanna predict how sales will be in 2013 ?

>
>Besides, I made it pretty clear that "Special Olympics" was
>about comparing with sparc. If that was a $20B/a market, the
>capitalization of Sun would be a tad higher, wouldn't you
>say?
>

The capitalization of SUN is a clear indication of its future. From a HW vendor it turned into a SW vendor incapable of monetizing it's most precious assets.Giving away for free what you develop with an R&D budget of over $1.2B a year isn't really smart.
SPARC sales are in the toilet ( thankfully Niagara helps a little ) and the x86 business leveled off.

>
>And that time starts anew each quarter, right?
>
Think in perspective.
When Intel developed x86 they thought it was useless.They needed 10 years until they saw the light.


>Ahh. That stupid argument.
>
>You know what? If you take IBM's decision to stop selling
>IPF as a great argument for how strong IPF is, I have an
>even better argument for you.
>
>Look at SGI. If just stopping selling IPF is such a great
>argument, what a wonderful argument SGI is! They did
>not just stop selling it, they went all the way to chapter
>11! Now that shows more than any small IBM failure!
>Taking the whole company down rather than dismantling just
>a tiny part.
>

Well, comparing SGI with IBM is misleading at best. IBM's strength isn't the HW, but the ecosystem it created around that HW.
SGI OTOH is forced to compete more or less on the HW head-on.

>Yeah, that argument shuts me right up. I bow down before
>your incredible powers of persuasion.
>
>For your edification: what makes IBM's in-house processor
>effort questionable is what made all the other boutique
>processors questionable: the slow and steady enroachment
>of x86 and the commercial advantages of mass markets.

As hard as it seems, IPF is a mass market product.HP takes 80% of it true, but it is available to anyone.
As for x86 encroaching the higher end , what's better ? Continue to fight x86 with in house CPU&FAB development or buying the said CPUs from the mass market and differentiating at system level?

>
>The IPF apologists talk about how IPF can take advantage
>of the mass markets through the fab, but they never admit
>to the fact that making processors isn't even primarily
>about fabs - it's about the designing and the selling of
>the damn things.
>
>Yes, fabs are expensive, but the only way you can make
>your money back is to sell a metric buttload of the
>end result.
>

That's the beauty of having x86 and IPF on top of that.

>Any POWER strategist isn't worried about IPF. They're
>thinking about x86, you dum-dum!
>
>Why can't you admit it? You almost come out and say it:
>
>>there were rumors of Power7 being socket compatible with
>>Opteron.
>
>.. but you are apparently not able to really think about it.
>
>Notice how they didn't for a second even entertain
>the (obviously idiotic) idea of making Power7 socket-
>compatible with IPF.
>
>Why? Because IPF does not matter.
>

Indeed, doesn't matter in the long run.It is simply allowing Intel to attack the market from above and below.When the 2 sides will meet, IPF is dead.



>
>Your argument is funny, but you are just sad. You're talking
>about how people may "probably porting Solaris to IPF".
>
>And then you apparently totally ignore the fact that Sun
>made a hell of a lot more commitment and ported, and sells,
>Solaris on x86. Yes, it was easier (they had some old and
>languishing support for it already), but that wasn't the
>big deal. It made sense.
>

SUN doesn't sell Solaris on x86 ; they give it away.Secondly their x86 business leveled ; they are nothing more than a niche player.
SUN can't make money on x86 nor can it afford to invest since the ROI simply isn't there.

>
>Intel could drop IPF tomorrow, and nobody would care.
>
>Yes, HP would blink, but then immediately put out some
>press release about how they continue to support OpenVMS,
>and in a day you'll see another one about how Core i7 Xeon
>is a perfect upgrade strategy from PA-RISC and HP-UX, and
>how they will continue to support their IPF strategy with
>newly ported tools to x86-64.
>
>Don't tell me they wouldn't. They've done it before. They
>can just fill in the template with new architecture names
>and change the date. All done.
>
>Linus

In order to make money you need to differentiate.Once you bring everything on the same platform , that becomes increasingly hard. x86 is always in the position of good enough since it is mass market.
Instead of porting HP-UX,OpenVMS,AIX or Solaris to x86 , you'te better off with Windows/Linux.They are good enough in the long run to eliminate everything else.