Which optimizes multicore CPU, Windows or chipset?

[John Doe]

Which more causes the efficient use of a multiple core CPU, the
operating system or the motherboard? In other words... Does
Windows 8 put a multiple core CPU to work much better than Windows
XP? I would ask whether upgrading the motherboard helps, but Intel
has prevented that from being a possibility, by changing its
sockets.

 

[John Doe]

I should've also asked...

Does the upgrade version worked like prior upgrade versions? You
can just use the CD key from Windows XP, and do a clean install?

I see that Microsoft has the download for $40. I'm probably going
for that.

 

[John Doe]

Sorry, wrong thread...

 

[Paul]

Which more causes the efficient use of a multiple core CPU, the
operating system or the motherboard? In other words... Does
Windows 8 put a multiple core CPU to work much better than Windows
XP? I would ask whether upgrading the motherboard helps, but Intel
has prevented that from being a possibility, by changing its
sockets.

Both WinXP and Windows 8, use a scheduler to decide when a
task gets a time slice. WinXP is already sophisticated enough
to get good performance from quad core processors.

WinXP probably doesn't care that much about cache state. Your
Q9550 consists of two silicon die, with a shared FSB joining
them together and to the chipset.

dual core dual core
| |
+------+-----+ <--- FSB inside LGA775 MCM
|
|
Video --- Northbridge --- Memory
|
Southbridge

Say we number the core from left to right 1,2 and 3,4.

If the scheduler is currently running a task on "1", WinXP
has no qualms about bouncing the task to "4". This causes
a cache flush on the left-most silicon die, and so there
is a tiny cost associated with migration.

Later OSes, the idea is, they assign a "weight" to migration.
If "2" isn't busy, an OS after WinXP knows it can
"bounce for free" from 1 to 2. Now, if "2" is loaded to 100%
right now, then bouncing to "2" would be stupid. Maybe "3"
has less of a load. So at some point, there is a tradeoff
between the tiny cost of a cache flush, versus a non-optimal
task placement.

I suspect that's all that would be relevant from an OS
later than WinXP.

And while it's fun to pretend this stuff matters,
I don't think it does for the most part. If one OS
was 5% faster, would you notice ? Maybe if you
were doing video rendering all day long. But if
doing email, playing Solitaire, surfing the web,
it probably is an undetectable difference. You'd
need to use a stopwatch, to see a difference.

Paul

 

[Yousuf Khan]

Which more causes the efficient use of a multiple core CPU, the
operating system or the motherboard? In other words... Does
Windows 8 put a multiple core CPU to work much better than Windows
XP? I would ask whether upgrading the motherboard helps, but Intel
has prevented that from being a possibility, by changing its
sockets.

It's the OS that's more responsible for interacting with the processor
cores, than anything else.

All the motherboard does (through its BIOS) is initialize the cores
prior to starting the OS. The OS takes care of the rest.

Yousuf Khan

 

[Flasherly]

It's the OS that's more responsible for interacting with the processor
cores, than anything else.

All the motherboard does (through its BIOS) is initialize the cores
prior to starting the OS. The OS takes care of the rest.

Yousuf Khan

Make shift and without a very minimal overall import for writing code
to utilize stages of progression planned across cores;- a few
companies have, although the actual significance might be as
compelling as contacting a travel agent to obtain a private vacation
on an orbiting space station. ...We're certain, fairly so to like to
be, that some day these things will come around.

On the other hand, say my audio processor program -- which end is
basically to render all musical passages with equal clarity, whether
originally produced to be perceived for loud or correspondingly soft
-- is multi-core programmed, as well utilized with the best PCI-
slotted, preamp sound board processing equipment. For all intent it's
a singular event on an island surrounded by a vast sea without a game
to play on the horizon. But, of course, such times well might temper
what there is to love about listening. The program costs nothing,
unless perhaps for a $1000US commercial radio or studio license, where
it's not below popular regard, similarly to state by recent demands
for soundboard pricing -- at $19US, is a culmination of audiophile-
regarded superiority placed for a time against all MBs equipped with
"better" sound chips and OPAmps -- although possibly fairer yet to
state in terms of commercial sound racks, if comparatively equipped
with such as a Behringer UltraCurve Pro (a slight differences to my
UCP, depending, as it'll also handle microphones, commercial and
studio impedances, and far-distance stage timing sequences between
offset speakers and undesired echo effects, some feedback circuitry
and whatnot).