3 Jun
2013
3 Jun
'13
7:29 p.m.
On 4 June 2013 01:03, David Riley <fraveydank at gmail.com> wrote:
On Jun 3, 2013, at 4:30 PM, Liam Proven <lproven at
gmail.com> wrote:
I can only agree with most of that.
And I would /love/ to see someone try to make a go of it with a
reverse-engineered Alpha-compatible chip. Can't see it happening but
it would give me immense pleasure.
Some of the newest Intel Atoms are down around ARM power consumption
now. You can buy mass-market Atom-powered phones here now.
http://www.intel.com/content/www/us/en/smartphones/smartphones.html
I wouldn't rule Intel out just yet.
--
Liam Proven ? Profile: http://lproven.livejournal.com/profile
Email: lproven at cix.co.uk ? GMail/G+/Twitter/Flickr/Facebook: lproven
MSN: lproven at hotmail.com ? Skype/AIM/Yahoo/LinkedIn: liamproven
Tel: +44 20-8685-0498 ? Cell: +44 7939-087884
On 3 June 2013 20:39, Cory Smelosky <b4 at
gewt.net> wrote:
I'm going to go with the latter. That doesn't make it automatically
"bad", but it is a reason. That's the reason Intel has been able to
pour money into making the x86 family perform as well as it does.
That's what bothers me about the x86
architecture...it gets such a kludgy
mess due to tacking it all on. Intel EM64T: it's amd64 so 64-bit with
32-bit x86 which is tacked on to a 16-bit architecture which is tacked on to
an 8-bit architecture and so on...
Yes it is. However, it has thrived when simpler, cleaner CPU
architectures have failed. Possibly an exemplar of the "worse is
better" school of thought, possibly because of its installed base of
software, possibly because of the tools, chipsets etc. - or possibly
because of the strength of Microsoft, who knows. POWER is bigger and runs hotter these days.
These days, it's also designed for different workloads. PowerPC is
vastly more efficient than x86 in the embedded world, both in terms
of computing done per clock and per watt. POWER is mainly used for
heavy datacenter workloads, and when you have applications designed
for it, it really lays the smack down on x86. But it's certainly
not as simple as "buy a bunch of 1U servers, throw them in a rack
and slap Linux on 'em".
SPARC and MIPS were unable to compete on CPU
performance.
Same with PowerPC, at least in the desktop world. I'm still
convinced that's mostly because Intel and AMD, having the lion's
share of the lucrative desktop market, were able to pour enough
money into R&D that they were able to make x86 CPUs faster by
sheer force of will (and they're RISC-ish processors on the
inside anyway, after a fashion).
Sparc and MIPS coulda been contenders, but interest (and thus
money) died out after x86 became "good enough" for the data
center. Newer Sparcs are actually quite compelling from a
performance standpoint, actually, especially when you consider
their virtualization performance (which is becoming increasingly
relevant).
ARM never competed on CPU performance since the
end of the 1980s, but
did on performance/Watt and price.
True. I think the rise of ARM parallels the sunset of PCs as a
whole; Intel has never been able to really compete in the low-power
arena, even with the Atom (which isn't a half-bad attempt). I like
ARM, but I'm not sure how much I like the direction they're going
with the really beefy stuff these days. Their tiny embedded CPUs,
though (like the Cortex-M series) are spectacular.
Alpha coulda been a contender -- maybe -- but I
learned recently that
its inter-CPU comms, synchronisation and so on were, I am told,
terrible, being very basic and rather unreliable.
That's possible. Probably nothing that couldn't be fixed with an
architecture revision. But it got axed, so we'll never know. The
Alpha definitely was originally designed as a pretty bare-bones
instruction set to keep the clock rate high, so it wouldn't
surprise me if some built-in primitives suffered as a result.