:: Re: [DNG] [Desktop-Environment] Cin…
トップ ページ
このメッセージを削除
このメッセージに返信
著者: Rick Moen
日付:  
To: dng
題目: Re: [DNG] [Desktop-Environment] Cinnamon and MATE
Quoting Alessandro Selli (alessandroselli@???):

> lscpu

[...]
> Model name:          Intel(R) Pentium(R) 4 CPU 3.40GHz

[...]
> dmidecode --string bios-release-date
> 12/21/2007


Thank you for that, Allesandro. Wikipedia's article on the Pentium 4
says this stepping was the _final_ Pentium 4 (if I'm reading that part
of the entry correctly) and was released in _2004_, using the 130 nm
Northwood core. (Reading further, I see that the Pentium That line's
immediate successor was the Pentium D (2005), Pentium D Extreme Edition
(2005), and Pentium 4 HT (2006) -- all x86_64-capable.)

So, I of course take as valuable information that your BIOS date was
from the end of 2007, but the CPU was being phased out in favour of
x86_64 two-plus years before that.

IMO, my points about 14-year-old gear being fragile specialty items
also applies to 12- and 10-year old computers. And, if you bought your
unit around 2007, you _could_ instead have bought x86_64 as
future-proofing.

On the matter that Adam mentioned about power draw (what the Yanks call
AC power, and the Brits call mains power): When my firm VA Linux
Systems was getting out of the hardware business because of the Dot-Com
market crash, many of us employees stocked up on the flagship VA Linux
Systems model 2230, a 2U rackmount unit w/Intel L440GX "Lancewood"
motherboard and PIII 'Coppermine' 800-1GHz CPU, because they were very
good, cheap, and with bountiful parts. A few people who imagined
themselves lucky acquired VA Linux's last product, model 1124, a 1U with
Tyan Thunder K7 motherboard, dual Athlon 760MP CPUs, custom PSU, and
carefully engineered (& patented) case cooling.

It needed the very carefully engineered cooling because the pair of
Athlon 760MPs put out tremendous amounts of heat, which of course the
operator also pays for in the shape of electric bills. Intel followed
the Athlon's example (meaning, releasing CPUs that tremendously
increased power draw). For many years, you could not have 48 x 1U units
in a standard server rack, with either AMD or Intel flagship CPUs,
because the racks' PDUs could not deliver that much power. It took both
firms many years to apply power-saving techniques from their mobile
lines to the flagship ones.

And, getting back to my point, your Northwood-core Pentium 4 with 3.40
GHz clock speed has a TDP of 89 Watts -- because the entire P4 line and
several of its successors sucked power at an amazing rate relative to
prior Intel (and AMD) CPUs. My Coppermine PIII has a TDP of 20.8 Watts.

Now, sure, TDP = thermal design power is only ambiguously a measure of
real power draw, as denotes the largest amount of heat output the CPU's
related cooling systems will be called upon to dissipate when running a
mix of real applications. But let's say it's a reasonable approximation
of real power draw, and the only thing the industry so far consistently
publishes.

Echoing Adam's point, the cost of each CPU sucking, on a 24x7 basis for
us server people, 4x the draw of a PIII really adds up, over time, and
costs significant money. And that, in turn, is actually why I delayed
retiring my spare Pentium III boxes and am still using one in 2017:
Because the entire P4-class architecture sucked too much power, that I'd
have to pay for in my electric bill. (Luckily, 2GB RAM has been enough
for that application, and it can easily saturate the aDSL link its
static IP lives on.)

My intended replacement, still under construction using Devuan, will
reduce power cost to a pittance: CompuLab Intense PC w/16GB RAM,
Celeron 847E 1.1 GHz dual-core, pair of mirrored SSDs on eSATA in
external enclosures. I'm not sure of the total draw yet, but think it
will be almost nothing -- thus even more cheap to run (not to mention
silent and ultra-cool).