lkml.org 
[lkml]   [1998]   [Jul]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: Merced makes things interesting, eh?
On 7 Jul 1998, Michael Poole wrote:

> Shawn Leas <sleas@ixion.honeywell.com> writes:
>
> > A little excerpt from a writeup by Rebecca Sykes, InfoWorld Electric seems
> > would like to see something like linux instead of it.
>
> I'm not sure why you would think this -- it's not going to be
> particularly better or worse at direct competiton than before; it's just
> that the "high-end" OSes will be running on hardware that is more similar
> to "consumer" OSes than before. The "high-end" will lose some
> competitiveness due to not having so much specially engineered hardware,
> but I'd imagine they'll still have some (so that the vendors can
> differentiate themselves, if nothing else).

Another platform on which Linux can directly compete with more commercial
OSes than ever? I might be missing something, but I thought it sounded
neat. However, as to why I dislike PH-UX... I think everything in PH-UX
is kludged into the OS, like their cdrom fs... PFS (pretty FuckedUp
fileSystem) is a hack if I ever saw one.

> [As a side note, HP will of course migrate to Merced -- they have helped
> Intel in the development of the Merced all along.]

They've also helped port Linux to the PA-RISC.

> The reason that the high-end hardware vendors are moving towards
> Merced (or the Alpha) -- as I see it, at least -- is to be more
> competitive. This comes through amortizing the cost of chip development

Agreed, but if the OS is the ONLY thing they're selling, there's a Linux
threat as I see it. This could lead to all the vendors thinking more
about improving their core OS in order to not be shown up. There is at
least less architectural worries in supporting Linux from a software
vendor's point of view, IMO.

> over many units. Look at the original Crays -- their individual
> processors were blindingly fast. As time progressed, supercomputers have
> gone from a few incredibly fast processors in one box to many smaller
> processors. It's easier to design slower processors than it is to

Agreed, except many tasks requiring supercomputer performance do not scale
to massively parallel systems.

> design faster ones, and to design really fast ones (the kind of chips
> we will be seeing as cutting-edge in 5 or 10 years), it takes a lot of
> work to develop and debug the design before you can even sample the chip.
> SGI (among others) seems to have realized this, and decided to move to
> Merced in the future -- unless you sell a lot of processors or very
> expensive processors[1], the chip division goes in the red, which no
> company likes.

All agreed here, that's what guberment contracts are for.

> It's a win for the hardware vendors to have this development
> done in one place, and Intel is logically positioned to do that since
> they have lots of money to throw at the problem (the Alpha is another
> logical choice since it's already faster than almost any other single
> processor out there, which is companies like Samsung are picking up on
> the 21264).

Does Unigraphics or Alias Wavefront run on Alpha? If not, I wish...

> [1] - and very fast and expensive processors tend to lose; the price
> per performance may be the same as for cheap processors, but the market
> isn't going to be large enough for the chip vendor to profit once a
> competitor enters the market. SGI workstations have long provided the
> best 3D graphics performance, but once cheapo PCs were remotely
> competitive, they stole market share and even expanded the graphics
> workstation market. And now SGI is moving towards the PC market.

Well, I saw it coming a long time ago. Things just didn't have the
acceleration that they did in the PC world. It's been a *somewhat* common
platform, complying to *some* standards at least *loosely*. And, only now
is it viable for HUGE tasks, with Merced coming out, breaking the 32 bit
barrier.. Sheesh, just a few years ago people were talking about 32 bit
as if it was the holy grail of Personal Computing (tm). (Not me, though)

-Shawn
> -- Michael


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu

\
 
 \ /
  Last update: 2005-03-22 13:43    [W:0.298 / U:1.084 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site