On Wed, May 10, 2017 at 08:54:41PM -0700, Michael Christopher Robinson wrote:
> A google article suggests that data corruption and not drive 
> failure is the greatest issue with SSDs.  If the data corrupts, 
> that is equally as bad as a drive failure.  Intel is working on 
> a new technology using Ferro Something for a semiconductor, it 
> is ultra low power and more reliable potentially than current 
> solid state flash memory technologies.  Perhaps someone else
> can fill in the gaps with what Intel is working on.

Intel "announced" 3D Xpoint solid state memory last
year; much fanfare, no details.  Their first Optane
products are scheduled for 2018.  Those will start
out far more expensive than flash or DRAM, with
performance somewhere in between.

"3D Xpoint" is probably chalcogenide phase change glass,
which Intel published a lot of journal papers about, and
filed a lot of patents ... then stopped about a year
before the big press conference announcing 3D Xpoint.
This is Intel's usual behavior.

Phase change glass is sometimes called "ferroelectric".

http://spectrum.ieee.org/semiconductors/memory/has-intel-invented-a-universal-memory-tech

The pretty picture is a computer rendering, not an
actual physical device. 

In semiconductors, a scheduled release of a new chip
technology more tham one year out means there are
serious problems, and no clue when or even if those
problems will be solved.   Remember bubble memory?

Nantero has a competing technology involving carbon 
nanotubes (gee whiz gosh wow!).  They've been working
on it for more than a decade, and are trickling out
samples now.  Starved for cash, like most startups.
I've talked with their CEO, who claims that Xpoint
won't work. 

Maybe nothing in the pipeline will work.

This is a struggle for investment dollars.  Those
dollars are migrating towards molecular biology, and
Intel and Nantero and others are fighting over crumbs.

My guess is that all of this is at the bleeding edge
of the possible, and when the fog clears, there will
be one player with enough customers and resources to
go into production with one more big change.  The
semiconductor industry is maturing; dense solid state
memory may be like Boeing's supersonic transport.

Much is made of Moore's law, doubling capability every
two years.  What is neglected is Moore's SECOND Law:
the cost of a chip-making FACTORY doubles every FOUR
years.  When the cost of a higher-capability factory
exceeds a company's slowing annual revenue growth
(Intel is at 5% now), the rate at which new factories
are built slows down, which slows down the rate of
process improvement, whcih slows down new applications.
Intel has already announced that the cadence of new
processes will slow. 

We may be nearing the end of the fast exponential,
as we did with automobiles, aircraft, electricity,
telegraph, telephone, television, rocketry, ...  

Sorry, no singularity, just a cascading series of
S curves.  It was fun while it lasted.

Further discussion belongs on plug-talk, but the overall
problem frames future Linux development, and software
in general.  I hope that, facing resource limits,
new software startups will focus on reliability and
stability and efficiency rather than exponential
accumulation of buggy complexity.  Perhaps we will
have time for regression testing ...

Keith

-- 
Keith Lofstrom          [email protected]
_______________________________________________
PLUG mailing list
[email protected]
http://lists.pdxlinux.org/mailman/listinfo/plug

Reply via email to