P4/E S-478 3Ghz ... multicore?

  • Thread starter Thread starter Flasherly
  • Start date Start date
F

Flasherly

Device Manager shows it for a - ACPI under Computer
Further down - Two instances are shown under Processors

Yet Task Manager shows two graphs as well as allowing Affinity for
Core#1 / #2 under the Processes tab

Everest under CPU -- CPU Utilizations shows Multi CPU Units #1/#2 --
however categorically qualified by HTT types (Hyper Threading, I
suspect)

Similarly, SpeedFan shows two CPU graphs

I can take, initiate two instances of a program, MP3Gain, divide up
some MP3s between them and both appear to run concurrently faster than
if all MP3s are given to one process;- heat generated by the CPU at
two instances is 121F and at one instance 116F
 
Flasherly said:
Device Manager shows it for a - ACPI under Computer
Further down - Two instances are shown under Processors

Yet Task Manager shows two graphs as well as allowing Affinity for
Core#1 / #2 under the Processes tab

Everest under CPU -- CPU Utilizations shows Multi CPU Units #1/#2 --
however categorically qualified by HTT types (Hyper Threading, I
suspect)

Similarly, SpeedFan shows two CPU graphs

I can take, initiate two instances of a program, MP3Gain, divide up
some MP3s between them and both appear to run concurrently faster than
if all MP3s are given to one process;- heat generated by the CPU at
two instances is 121F and at one instance 116F
Click to expand...

"Intel Processor Identification Utility - Windows Version"

http://downloadcenter.intel.com/Detail_Desc.aspx?ProductID=1881&DwnldID=7838&lang=eng&iid=dc_rss

File name: pidenu31.msi <--- english version, or select another language

There is a screenshot here, of the utility while it is running.

http://i1-win.softpedia-static.com/screenshots/Intel-Processor-Identification-Utility_2.png

That one shows "Hyper-Threading No", whereas your processor is
probably "Hyper-Threading Yes". Hyperthreading means there is one
core, but with a physical and a virtual presence. The processor
contains enough registers, that if one set of registers blocks
because of an outstanding memory access, the processor uses a
second set of registers. And that counts as a "virtual" core. Such a
scheme (Hyperthreading), yields about 10% more performance from a
single core, under best case conditions. If two threads thrash on
memory access, performance actually drops.

The OS thinks there are two cores, but there really aren't two cores
present. The OS can schedule tasks to each side, but since the
sides fight for resources, one side (and its set of registers)
will block, and the other side will run. It's only when a resource
contention is avoided, that HyperThreading is a win from a
performance perspective. Both the physical and virtual core
can't run at the same time. Making them look like cores, is
to fool the OS into scheduling tasks for each side.

Program Counter #1 Program Counter #2
Other registers... Other registers...
| |
+------------+-----------+
|
rest of the
processor core
|
To memory controller
and/or Northbridge

So only one side or the other, can be running at one time.
And if the processor would normally be stopped waiting for
a memory access, the other set of registers may be able to
function by using information already in L1/L2 (from a previous
fetch perhaps). Since the cache has good rates, it means
fewer computing opportunities are lost. And the scheme squeezes
10% more performance from the processor as a result.

If you want to learn more (an article better than my simple summary),
try this one. It gives details on the real deal.

http://www.xbitlabs.com/articles/cpu/display/replay.html

Paul
 
Yes -- HTT being an effective HyperThread construct, as well as from
another instance of HyperThreading spelled out and support mentioned.
I'm not so sure about where a '10% overall performance yield' would be
most significant, given and since I do have another "official" AMD 2X
4200 dualcore for comparative means. The mentioned instance of audio
normalizers I did use to provide, seems obviously apparent to have
benefited in 2 processes, so arrayed above singularity. A gist for
further optimums, then, I'd agree that ' thrashing isn't apparent if
the registers are so aligned to HyperThread.' Something of added
abstractions as well to consider from a user perspective, as to
whether there's perfunctorily significance to "off the cuff"
assignments to posing applications as suitable for 'pseudo-operational
status' at earlier ingress within a definition most loosely permitted
a multicore construct. Hyperthreading apparently meets that need, as
I haven't personally found another such factor, for a strict single
core to consider by 'virtuality' as well for ascertainable
advantage. ...Then, again, perhaps there simply would not be anything
quite so imposing in that same sense as hyperthreading.

Where the hard does meet the road in tack is indubitable. That same
478 processor came from instances of two discrete program processes
with wider variance both permissible in packing audio-visual
presentment, in the case of the audio only by an application written
for (tm)-awareness, and optional deployment upon the multicore (tm)
platform. Extensive familiarity with the Intel P4/E-S478 in that
environment was best to say a dubious experience of sustainable ends,
in realtime processing of audio signals, similar but more advanced
than the aforementioned example, as roughly designed for simultaneous
expansion, and compression, of frequencies within a normative
broadband stream and accompanying presentment to video.

The AMD X2 4200 simply excels there, all other things being equal in
terms of "affinity" and core identification purposes the OS largely
has to bestow. Apart from an optional multicore "switch" the audio
DSP is endowed, I don't believe I have yet another program with
similar capacity. Perhaps it is faithful to add a note from licensed
broadcasters using the same program in more advanced multicore-
dedicated build configurations, as well rates they're duly charged
commiserate to an author allowing them to use it.

Excepting a supposition the first release of Unreal Tournament might
prove within similar benefit. How very nauseating, the thought.
Chess is all else I've always known, and, like most any math,
something I can always excel with a very least effort upon my part to
be acceptable;- I play at expert levels after perhaps three months of
dedicated study, or did so last I made the attempt. Now seems certain
vast ages ago, alas. A computer is less useful for grand stratagems
than actually a discipliner and bludgeoner of the rote, an avoidance
system, as it were, for being picked apart, finely, into piecemeal.
 
Back
Top