super_pi problem on xp

[sali]

having win/xp on new intel i7/920 cpu
on task manager, it show me 8 cpu, which is ok.
wanted to test multiprocessing, to se on which way xp assignes proc & tasks
and to utilize all 8 procs simultaneously, as much as possible
so, i started super_pi, utility to test number cunching power of cpu
one instance with 32m decimals is finished in appx 15 minutes, running on
cpu_0

then started few instances of super_pi, each of them [according to task
manager] being automaticly [by xp] assigned to separate proc

here comes the surprise:
all successive super_pi tasks [except the first one started in the
sequence], reported error "not konvergent in sqrt", and stopped working,
only that first in the sequence continued and finished [as it was started
alone]

my question is:
is this expected behaviour? as i expected win/xp will seamlessly run few
concurrent tasks of super_pi, having done multiple job almost in the same
time as one is done.
is maybe there some wrong adjustment in my mobo or win subsystem, having
additional instances of super_pi throwing float_point error?

any suggestion?

thnx!

 

[Martha Stewart]

Martha Stewart's Mile High Lemon Pie Recipe

Ingredients:

All-purpose flour, for dusting
1/2 recipe Pate Brisee (follows)
1 large egg
2 tablespoons heavy cream

***Pate Brisee***

2 1/2 cups all-purpose flour
1 teaspoon salt
1 teaspoon sugar
1 cup unsalted butter, chilled and cut into small pieces
1/2 cup ice water

***Filling***

1/3 cup cornstarch
1/3 cup sifted cake flour
1/4 teaspoon salt
1/4 cup sugar
5 large egg yolks, lightly beaten
1/2 cup fresh lemon juice
2 tablespoons lemon rind
4 tablespoons unsalted butter, cut into small pieces
***Swiss Meringue***
7 large egg whites
3/4 cup sugar
1/4 teaspoon salt


Directions:

Heat oven to 400. On a lightly floured surface, roll out dough to a
thickness of 1/8 inch, and use to line a 9-inch pie tin. Crimp the edges
decoratively. Chill until firm.

In a small bowl, whisk together egg yolk and cream to make a glaze.
Prick the bottom of the shell with a fork, brush the top edges with the
glaze, and line with parchment paper. Weight the shell with pie weights
or dried beans, and bake until the edges begin to brown, 10 to 15
minutes. Remove paper and weights; continue baking until golden brown, 7
to 10 minutes more. Let cool.

To make filling, combine cornstarch, cake flour, salt, and sugar in a
medium nonreactive saucepan over medium heat. Gradually add 2 cups cold
water. Bring the mixture to a boil, whisking constantly, about 4 minutes.

Remove the pan from heat. Temper egg yolks by beating a small amount of
hot mixture into the yolks before adding them to pan. Cook over low heat
for 5 minutes.

Remove the pan from heat, and whisk in the lemon juice and rind. Add the
butter one piece at a time. Pour the mixture into a large bowl, and let
cool. Pour the filling into the cooled shell and refrigerate, covered
with aluminum foil, until firm, about 1 hour.

To make the meringue, combine egg whites, sugar, and salt in a
heat-proof bowl. Set over a pan of simmering water; beat until warm and
sugar is dissolved. Remove bowl from heat; whip into stiff peaks.

Spread the meringue over pie so that it touches crust all around. Broil
until brown, about 2 minutes, watching constantly. Serve at room
temperature.

NOTE: Raw eggs should not be used in food prepared for pregnant women,
babies, young children, the elderly, or anyone whose health is compromised.

To make Pate Brisee (PIE DOUGH): (Makes 1 double-crust or 2 single-crust
9- to 10-inch pies)

Pate brisee is the French version of classic pie or tart pastry.
Pressing the dough into a disk rather than shaping it into a ball allows
it to chill faster. This will also make the dough easier to roll out,
and if you freeze it, it will thaw more quickly.

In the bowl of a food processor, combine flour, salt, and sugar. Add
butter, and process until the mixture resembles coarse meal, 8 to 10
seconds. With machine running, add ice water in a slow, steady stream
through feed tube. Pulse until dough holds together without being wet or
sticky; be careful not to process more than 30 seconds. To test, squeeze
a small amount together: If it is crumbly, add more ice water, 1
tablespoon at a time.

Divide dough into two equal balls. Flatten each ball into a disk, and
wrap in plastic. Transfer to the refrigerator, and chill at least 1
hour. Dough may be stored, frozen, up to 1 month.

having win/xp on new intel i7/920 cpu
on task manager, it show me 8 cpu, which is ok.
wanted to test multiprocessing, to se on which way xp assignes proc & tasks
and to utilize all 8 procs simultaneously, as much as possible
so, i started super_pi, utility to test number cunching power of cpu
one instance with 32m decimals is finished in appx 15 minutes, running on
cpu_0

then started few instances of super_pi, each of them [according to task
manager] being automaticly [by xp] assigned to separate proc

here comes the surprise:
all successive super_pi tasks [except the first one started in the
sequence], reported error "not konvergent in sqrt", and stopped working,
only that first in the sequence continued and finished [as it was started
alone]

my question is:
is this expected behaviour? as i expected win/xp will seamlessly run few
concurrent tasks of super_pi, having done multiple job almost in the same
time as one is done.
is maybe there some wrong adjustment in my mobo or win subsystem, having
additional instances of super_pi throwing float_point error?

any suggestion?

thnx!

 

[Paul]

having win/xp on new intel i7/920 cpu
on task manager, it show me 8 cpu, which is ok.
wanted to test multiprocessing, to se on which way xp assignes proc & tasks
and to utilize all 8 procs simultaneously, as much as possible
so, i started super_pi, utility to test number cunching power of cpu
one instance with 32m decimals is finished in appx 15 minutes, running on
cpu_0

then started few instances of super_pi, each of them [according to task
manager] being automaticly [by xp] assigned to separate proc

here comes the surprise:
all successive super_pi tasks [except the first one started in the
sequence], reported error "not konvergent in sqrt", and stopped working,
only that first in the sequence continued and finished [as it was started
alone]

my question is:
is this expected behaviour? as i expected win/xp will seamlessly run few
concurrent tasks of super_pi, having done multiple job almost in the same
time as one is done.
is maybe there some wrong adjustment in my mobo or win subsystem, having
additional instances of super_pi throwing float_point error?

any suggestion?

thnx!

Test with Prime95. It is multithreaded, and also does a numerical
calculation to prove the processor is working correctly. I let this
run for four hours, and no errors are acceptable.

http://majorgeeks.com/Prime95_d4363.html

Also, you can control which core is used in your Super_Pi testing,
by using the affinity settings in Task Manager. So if you want to
move one copy of Super_Pi around on your multicore processor, try
making changes to the affinity tick boxes for the Super_Pi process.

Paul

 

[Ian D]

having win/xp on new intel i7/920 cpu
on task manager, it show me 8 cpu, which is ok.
wanted to test multiprocessing, to se on which way xp assignes proc &
tasks and to utilize all 8 procs simultaneously, as much as possible
so, i started super_pi, utility to test number cunching power of cpu
one instance with 32m decimals is finished in appx 15 minutes, running on
cpu_0

then started few instances of super_pi, each of them [according to task
manager] being automaticly [by xp] assigned to separate proc

here comes the surprise:
all successive super_pi tasks [except the first one started in the
sequence], reported error "not konvergent in sqrt", and stopped working,
only that first in the sequence continued and finished [as it was started
alone]

my question is:
is this expected behaviour? as i expected win/xp will seamlessly run few
concurrent tasks of super_pi, having done multiple job almost in the same
time as one is done.
is maybe there some wrong adjustment in my mobo or win subsystem, having
additional instances of super_pi throwing float_point error?

any suggestion?

thnx!

The Core i7 does not have 8 cores. It has 4 cores and 8 HT units.
The HT allows for more efficient thread scheduling, but does not
increase processing power. There is still only one processor per
core. The 8 CPUs Task Manager shows are the 8 HT units. In
XP, Task Manager shows the 4 HT0 units then the 4 HT1 units.
Vista correctly shows them as HT0/HT1 pairs.

 

[sali]

The Core i7 does not have 8 cores. It has 4 cores and 8 HT units.
The HT allows for more efficient thread scheduling, but does not
increase processing power.

thnx, my post was over-simplified.
i know the difference between real proc and just the hyper-thread over it.
ht allows us to use some of cpu's parallel units which will be otherwise
left unused, so you may have better proc utilisation with ht than without
it. it is simple idea: if i may get more power with ht, why don't use it

so, here goes also the question: with automatic task assignment to core/ht,
the operating system should first try to use all real procs, and onla when
all used, trying to use ht, isn't it?

 

[Ian D]

thnx, my post was over-simplified.
i know the difference between real proc and just the hyper-thread over it.
ht allows us to use some of cpu's parallel units which will be otherwise
left unused, so you may have better proc utilisation with ht than without
it. it is simple idea: if i may get more power with ht, why don't use it

so, here goes also the question: with automatic task assignment to
core/ht, the operating system should first try to use all real procs, and
onla when all used, trying to use ht, isn't it?

It works that way with MS Flight Simulator FSX.

The MS Flight Simulator FSX program with the SP1 and SP2
updates can use multiple cores. I think the design max is 16
cores, but it may be more. With HT on in BIOS it uses only
the HT0 unit of each i7 core, and the total CPU utilization
only shows 50% in Task Manager. There's the odd little blip
in one or two of the HT1 units, but I think it's something the
OS is doing. With HT off in BIOS, FSX uses all 4 cores at 100%,
when required with the total CPU utilization at 100%. With FSX,
core 0 runs the main thread and is always at 100%, and the other
cores vary as required, with the load spread equally across the
3 cores.

 

[sali]

It works that way with MS Flight Simulator FSX.

The MS Flight Simulator FSX program with the SP1 and SP2
updates can use multiple cores. I think the design max is 16
cores, but it may be more. With HT on in BIOS it uses only
the HT0 unit of each i7 core, and the total CPU utilization
only shows 50% in Task Manager. There's the odd little blip
in one or two of the HT1 units, but I think it's something the
OS is doing.

in my case [with multiple super_pi instances] there are 8 cpu history,
according to the doc, left four are real cores 0..3, while right ones are
ht-cores, 4..7, and sometimes, super_pi is started in one of the right
ht-cores, according to cpu history of that core/ht, rising to 100%.
that are the results from few days ago, i need time to perform further tests