On Sat, 03 Jun 2017 14:58:55 -0400, Paul
wrote:

It's a price war.

https://www.theinquirer.net/inquirer...whitehaven-cpu

But it's not going to magically revive the desktop business.

*******

The 18-core for $2000 isn't good value, because
it's more suited to commercial server applications,
than to a desktop. Your 7ZIP compression run will go
faster, but the decompression step will be as slow
as ever. If you're the kind of person who leaves
four videos rendering in the background, then it'll be
a win for you.

If you're just a gamer, not so much. I'm not aware of
any games designed for that many cores. Maybe Fritz Chess
can scale to use them, but not much else.

What the 18-core processor will be doing, is a lot
of benchmarks. Until the thrill wears off.

Paul

I've seen the benchmarks on AMD's non-Ryzen octal and it's also
disturbing. And that was an E-series, further Bulldozer technology
adjustments, halfway through its life span and ostensibly worthy of
mention as Vishera.

What that means is there's 95-watts is the tip of an iceberg that's
capable of drawing 300-watts with all 8 cores fully engaged. Perhaps
a minor distinction to however AMD's processor-dependency Turbo Mode
is triggered for increasing core speeds by a factor of one-quarter
faster.

The reviews have been around for some time, and no less apt to involve
a gamer setup, tickled-pink about running 4.5GHz on water cooling with
some sort of "extreme" placed in MB labeling, capable of non-linear MB
power distribution over stock operating characteristics. (Most MBs
when broaching this territory are quick to red-line -- to most
definitely not fit an upper 225-watt Vishera model into their
socket.)

What not all benchmark and reviews reveal is an extremity of heat
generated on MBs to satisfy that power draw. 200 degree Fahrenheit,
nor the latest generation of solid-state capacitors, is not of
significant assurance between those for whom new octal builds fail due
to a MB melt-down.

Of course applicability, how one approaches stressing stated limits,
should offer some residual results. Whether, conversely, the
processor's multiplier is reduced and under-clocked would seem easiest
to first approach before setting up program scenarios for engaging
core multiples.

That's my intended plan after a favorably impression from implementing
it on a quadcore. Besides, there's nothing whatsoever "Extreme" about
an expenditure in parts I chose to have, a bare minimum, eight cores
accessible. Both their limitations to instruction threading and, all
of the above, in a drawback of underlying efficiency should not suffer
an impression beyond applicable engagement and benefit to a basic
budget, eight-core build.