| leading the way to the
new storage frontier
by Zsolt Kerekes,
|Editor:- September 2, 2010 -
this week a reader (Andrew Hancox) sent me a list of 10 key questions
about the future of the SSD market. |
As I'd already answered most of
them to some degree in articles my reply was mostly a list of links.
1st question was - "How long do you think it will be before the pricing
of SSDs comes down to a level where they are a viable option to be used as
primary drives in portable devices for
I've been answering that question in articles for 5 years - starting with my
penetration model, numerous comments on the
page, quotes from other
SSD analysts and
comments in past
news pages - but I've never been able to give a deterministic number because
there are parameters involved which depend on vendors in the market changing
Here's what I actually said in my email reply - "A
big obstacle is not media pricing - but how well the SSD design is integrated in
the notebook motherboard design. Nearly all current notebook designs are
adapted from HDD designs. Adding SSDs into them wastes most of the potential
benefits of the SSD. That's why this is taking years longer than it should
have done... "
Thinking back on my replies to readers I'm never
really sure how well they have been understood. That's because
SSD education is a
Now when I talk about "motherboard design"
that's the electronics and computer architect in me talking in a code language
which translates as - the design of most notebooks is a mismatch for getting
the best out of SSDs. Most notebook vendors are too lazy to design new SSD
notebook products - so instead they integrate me-too SSDs into me-too
HDD-centric notebook designs - to get results which fail to inspire anyone. Then
they complain that the market projections for SSD adoption in notebooks didn't
This morning I thought of a good analogy for what's been
happening in the notebook SSD market.
Imagine that Henry Ford - had
looked at the horse drawn carriage market and the internal combustion engine -
and had decided to design mechanical horses (powered by the new engines) which
were then coupled to a coach in the traditional way.
There you have
today's notebook SSD market...
From the viewpoint of the horseless
carriage customer it's an expensive novelty. The engine is whirring as fast as
it can - but those clippety cloppety legs in the mechanical horse can't run
any faster. The coach rolls along and it's an impressive sight but also a
waste of internal combustion engine horsepower. This will never become a mass
|| Most notebooks today
completely waste the potential horsepower of SSDs. How can you predict when
notebook marketers and designers will stop being stupid? That's where my
crystal ball fails.|
|...Later:- I got this
follow up question from another reader Robert Young|
missing from your article is the completion of the analogy: in what specific
way does an SSD/notebook combination happen which is the analog to Ford's
automobile (_not_ putting the engine in the horses)?
How does this
differ from what's happened with desktop/server machines? Near as I can see,
all three types of machines implement SSD as HDD, and in the same way: SATA.
The only SSD specific support I know of is in linux. The btrfs
filesystem, not yet fully implemented,
Editor's reply:- the motherboard chipsets
translate the disk I/O requests from memory into SATA then inside the SSD
back again from SATA into memory I/O in the flash. So a lot of hard disk world
simulation occurs which adds to latency, makes reliability worse and slows down
throughput. True SSD notebooks would have the flash arrays directly accessible
as in Fusion-io's iodrives.
...a few days later:-
Editor's follow up reply:- I
forgot to mention that in an earlier article I pointed out the differences
between notebook and PC/server motherboard designs when it comes to the ability
to use Legacy SSDs (like SATA). it's at the bottom of this article -
chips are chosen to have just enough speed to do the intended job.
This minimizes power consumption and reduces the price of the chips. The result
is that the performance headroom in these chipsets is much more constrained
than with servers and desktop PCs.
So when you add a fast SATA SSD to
most notebooks - the intermediate notebook chips waste the SSD's potential
performance as seen by the host CPU. You pay good money for low latency in the
SSD - but the notebook chips add latency back and slug the SSD performance -
because they were designed to be good enough for (slower) hard drives where
these factors were not so critical.
If someone wants to add an SSD to
a notebook despite those reservations - it's best to choose a notebook which
was designed to work with a 7,200 RPM HDD - if you're using SATA SSDs - or
better still - one of the many variants of mini PCIe.
is published by