Frequency, Bus speed, rated FSB ...cpu-ram memory ?

[Majki Majk]

information
Type CPU / Microprocessor
Family Intel Pentium M
Processor number ? 755
Part number RJ80536GC0412M
Processor markings RJ80536 2000/2M
Frequency (MHz) 2000
Processor speed in LFM mode (MHz) 600
Bus speed (MHz) ? 400
Clock multiplier ? 20
Package type 479-ball Micro-FCBGA

Architecture / Microarchitecture / Other
CPUID 06D6h
Core stepping B1
Manufacturing technology (micron) 0.09
L2 cache size (KB) ? 2048
Features Execute disable bit ?
SpeedStep technology ?
Core voltage (V) ? 1.276 - 1.34
Core voltage in LFM mode (V) 0.988
Case temperature (°C) ? 100

What is the case with this two in relation ?
Bus speed of my processor is working only at 100mhz, like the DRAM
frequency is also at 100mhz. I have 2x 1024Gb 533mhz DDR2 memory modules
installed and this mobile processor installed with multiplier of 20x. In
CPUZ Bus speed is 100mhz and the FSB is 400mhz. My other computer works
fine with core of 2Ghz at Bus speed of 200mhz and HT system of 1Ghz with
multiplier 10x.
Is there a way to adjust this procesor to work fine with multiplier
10x...and why would that be better then multiplier of 20x ?

I had had this cpuz working at multiplier 6x at 600mhz with LFM, and it
woked great... ?

Thanks for any opiniones and advices !

mm

 

[Paul]

information
Type CPU / Microprocessor
Family Intel Pentium M
Processor number ? 755
Part number RJ80536GC0412M
Processor markings RJ80536 2000/2M
Frequency (MHz) 2000
Processor speed in LFM mode (MHz) 600
Bus speed (MHz) ? 400
Clock multiplier ? 20
Package type 479-ball Micro-FCBGA

Architecture / Microarchitecture / Other
CPUID 06D6h
Core stepping B1
Manufacturing technology (micron) 0.09
L2 cache size (KB) ? 2048
Features Execute disable bit ?
SpeedStep technology ?
Core voltage (V) ? 1.276 - 1.34
Core voltage in LFM mode (V) 0.988
Case temperature (°C) ? 100

What is the case with this two in relation ?
Bus speed of my processor is working only at 100mhz, like the DRAM
frequency is also at 100mhz. I have 2x 1024Gb 533mhz DDR2 memory modules
installed and this mobile processor installed with multiplier of 20x. In
CPUZ Bus speed is 100mhz and the FSB is 400mhz. My other computer works
fine with core of 2Ghz at Bus speed of 200mhz and HT system of 1Ghz with
multiplier 10x.
Is there a way to adjust this procesor to work fine with multiplier
10x...and why would that be better then multiplier of 20x ?

I had had this cpuz working at multiplier 6x at 600mhz with LFM, and it
woked great... ?

Thanks for any opiniones and advices !

mm

You are comparing architectures which are not similar, and attempting
to adjust things that are not equivalent. It is like asking whether
overclocking is allowed or supported. Yes, the BIOS may have settings
to adjust things, but you have to understand what the consequences are,
before using such settings.

*******
In the first (Intel) architecture -

CPU_input times Core_Multiplier gives Core_frequency, 100x20 = 2000MHz core

The Intel FSB is 8 bytes wide, and transfers data four times per clock
cycle. 100x4 implicit, gives FSB400. FSB400 * 8bytes/transfer = 3200MB/sec

The Intel Northbridge chip has the memory controller. The memory controller
accepts 100MHz as an input reference, but may operate the memory at
a speed which is the ratio of simple integers. Sometimes, more than
one option is available in the BIOS. The memory purchased, should be
fast enough to handle whatever setting the user chooses in the BIOS.
Perhaps the options are DDR2-400 and DDR2-533 for example. Those rates
are synthesized based on the 100MHz input clock.

*******
In the Athlon64 (your desktop) architecture -

CPU_input times Core_Multiplier gives Core_frequency, 200x10 = 2000MHz core

The Athlon64 FSB is 2 bytes wide, and transfers data two times per clock.
The ref clock for it is CPU_input times multiplier, or 200x5=1000MHz.
(The HT multiplier is a separate BIOS setting, adjustable from 1 through 5).
Transferring two times per bus cycle, of two bytes per transfer, times
1000MHz, gives 4000MB/sec, or slightly faster than the Intel. And
Hypertransport has two separate bidirectional busses, whereas Intel
uses a single bus which changes direction when required.

Generally, there is no observable advantage to overclocking the HT bus.
System performance is relatively insensitive to it.

The Athlon64 has the memory controller inside the processor. The DIMMs
connect to the processor. The processor operates the DIMMs using
clock synthesis, using the ratio of simple integers. The BIOS will
strive to select a value which does not exceed the memory speed
rating. Users can override this by adjustment in the BIOS.

*******

You should consider each of these computers, without respect to the other.
Can parts of a computer be overclocked ? Yes, but the impact may be a
reduction in stability, hotter operation, increased power consumption
and so on. There must be a good reason to do it. Proper testing
will establish whether the new settings are correct.

Overclocking a laptop is dangerous, because the laptop already has
a rather high internal operating temperature. Laptop cooling is
barely adequate, even when the heatsinks are kept clean. With a
desktop computer, you can modify the cooling systems, to prepare
the computer for overclocking.

HTH,
Paul

 

[Majki Majk]

Basically You are saying there is no possibility to force my laptop to
Bus speed 200mhz from 100mhz at the moment ?

COuld you explain in short what is better for computer performance
....when everything is on higher frequencies let say 800mhz FSB or FSB
400mhz, one guy asked me that, in me didn't know what to answer to him !?

Is computer performance really better with higher FSB and memory frequency ?
mm

 

[Paul]

Basically You are saying there is no possibility to force my laptop to
Bus speed 200mhz from 100mhz at the moment ?

COuld you explain in short what is better for computer performance
...when everything is on higher frequencies let say 800mhz FSB or FSB
400mhz, one guy asked me that, in me didn't know what to answer to him !?

Is computer performance really better with higher FSB and memory
frequency ?
mm

Here are the test results from my latest build. The processor is
an E4700 dual core 2.6GHz with normal FSB of FSB800. The first pair
of lines in the table, test the processor at its normal conditions.
I used SuperPI to test performance. A lower test time is better.
I tested RAM in single channel mode and then in dual channel
mode. (View the table in a fixed font like Courier.)

SuperPI Memtest 1.65
1M (sec) Bandwidth
lower is (MB/sec)
better

200 x 13 = 2.60GHz, FSB800, DDR2-533, Single channel 24.05 2203
Dual channel 22.87 2668

266 x 10 = 2.66GHz, FSB1066, DDR2-533, Single channel 23.47 ----
Dual channel 22.52 ----

266 x 13 = 3.46GHz, FSB1066, DDR2-533, Single channel 19.37 2419
Dual channel 18.42 3305

In the second test, I ran the FSB at a higher setting, then reduced
the multiplier setting while in Windows. The resulting core speed
is almost the same as the nominal value (to make it easier to
compare changing just the FSB.) My BIOS doesn't allow
the multiplier to be set in the BIOS. I couldn't measure the
bandwidth in Memtest in the second set of test conditions,
because Memtest is not Windows, and won't allow a program
to run except for the memory test.

You can see from my result, that there is a slight improvement with
the higher FSB.

The third result, is with the core frequency overclocked a fair
bit. That is making a larger improvement, than changing the FSB alone.

So changing the FSB makes a difference, but not much of a
difference.

I'm not able to run a memory frequency series of tests on
my motherboard, because the memory controller doesn't work
properly at higher speeds (made by VIA). That is why there
aren't tests done at higher memory speeds than DDR2-533.

Changing the memory speed makes a small difference, but you
cannot guess at the impact based on frequency alone. For
example, if you had DDR2-533 memory and DDR2-800 memory,
there is not a 800/533 improvement. The improvement per
step is one or two percentage points. But since I cannot
change my memory speed, I'm not able to demonstrate that
using my $70 motherboard.

HTH,
Paul

 

[Majki Majk]

Here are the test results from my latest build. The processor is
an E4700 dual core 2.6GHz with normal FSB of FSB800. The first pair
of lines in the table, test the processor at its normal conditions.
I used SuperPI to test performance. A lower test time is better.
I tested RAM in single channel mode and then in dual channel
mode. (View the table in a fixed font like Courier.)

SuperPI
Memtest 1.65
1M (sec)
Bandwidth
lower is (MB/sec)
better

200 x 13 = 2.60GHz, FSB800, DDR2-533, Single channel 24.05 2203
Dual channel 22.87 2668

266 x 10 = 2.66GHz, FSB1066, DDR2-533, Single channel 23.47 ----
Dual channel 22.52 ----

266 x 13 = 3.46GHz, FSB1066, DDR2-533, Single channel 19.37 2419
Dual channel 18.42 3305

In the second test, I ran the FSB at a higher setting, then reduced
the multiplier setting while in Windows. The resulting core speed
is almost the same as the nominal value (to make it easier to
compare changing just the FSB.) My BIOS doesn't allow
the multiplier to be set in the BIOS. I couldn't measure the
bandwidth in Memtest in the second set of test conditions,
because Memtest is not Windows, and won't allow a program
to run except for the memory test.

You can see from my result, that there is a slight improvement with
the higher FSB.

The third result, is with the core frequency overclocked a fair
bit. That is making a larger improvement, than changing the FSB alone.

So changing the FSB makes a difference, but not much of a
difference.

I'm not able to run a memory frequency series of tests on
my motherboard, because the memory controller doesn't work
properly at higher speeds (made by VIA). That is why there
aren't tests done at higher memory speeds than DDR2-533.

Changing the memory speed makes a small difference, but you
cannot guess at the impact based on frequency alone. For
example, if you had DDR2-533 memory and DDR2-800 memory,
there is not a 800/533 improvement. The improvement per
step is one or two percentage points. But since I cannot
change my memory speed, I'm not able to demonstrate that
using my $70 motherboard.

HTH,
Paul

Thanks for motherboard and processor review. It's nice to know how it
works alltogether, ok the experience makes a part of its own. So this
Bus speed is like a backbone through which operations are dealt ?
So if as, then my backbone of 100mhz Bus speed is a weak beackbone or...

 

[Paul]

Thanks for motherboard and processor review. It's nice to know how it
works alltogether, ok the experience makes a part of its own. So this
Bus speed is like a backbone through which operations are dealt ?
So if as, then my backbone of 100mhz Bus speed is a weak beackbone or...

If you know the name of your chipset

http://www.intel.com/products/notebook/chipsets/915gms/915gms-overview.htm

you can get a block diagram for it from the Intel site, like this.

http://www.intel.com/Assets/Image/diagram/915gms_diagram_760.gif

In that example, the DDR2 memory at DDR2-400, gives 3.2GB/sec max.
The FSB is FSB400, and since the FSB is 8 bytes wide, it also
transfers at 3.2GB/sec max. So that laptop chipset is balanced.
There is a small amount of performance to be gained by changing
the clock, but it would not represent a major improvement. (Perhaps
a few percentage points.) And if the BIOS will not allow a setting
of DDR2-533, it may not be possible to experiment.

Post the name of the chipset listed in CPUZ, so I can look it up.
In your initial posting, I didn't see the chipset name. I
know you have a Pentium-M 755, but the chipset is probably
something more recent than an 855GME.

Paul

 

[Majki Majk]

If you know the name of your chipset

http://www.intel.com/products/notebook/chipsets/915gms/915gms-overview.htm

you can get a block diagram for it from the Intel site, like this.

http://www.intel.com/Assets/Image/diagram/915gms_diagram_760.gif

In that example, the DDR2 memory at DDR2-400, gives 3.2GB/sec max.
The FSB is FSB400, and since the FSB is 8 bytes wide, it also
transfers at 3.2GB/sec max. So that laptop chipset is balanced.
There is a small amount of performance to be gained by changing
the clock, but it would not represent a major improvement. (Perhaps
a few percentage points.) And if the BIOS will not allow a setting
of DDR2-533, it may not be possible to experiment.

Post the name of the chipset listed in CPUZ, so I can look it up.
In your initial posting, I didn't see the chipset name. I
know you have a Pentium-M 755, but the chipset is probably
something more recent than an 855GME.

Paul

Chipset is i915GMS/i910GML according CPUZ
and as Northbridge says: Intel Alviso-G i915
Southbridge: Intel 82801FBM ICH6-M

I saw numerous bioses, and this one doesn't have many open functions to
play with..

mm

 

[Paul]

Chipset is i915GMS/i910GML according CPUZ
and as Northbridge says: Intel Alviso-G i915
Southbridge: Intel 82801FBM ICH6-M

I saw numerous bioses, and this one doesn't have many open functions to
play with..

mm

So this should be pretty close to your architecture.

http://www.intel.com/Assets/Image/diagram/915gms_diagram_760.gif

Page 25 of the datasheet gives an overview. (FSB400, single channel DDR2-400)
http://www.intel.com/Assets/PDF/datasheet/305264.pdf

On PDF page 292, section 10.14.3.2, it shows only one set of operating
conditions for the FSB and memory on the 915GMS. While there are
likely to be ways to overclock everything at the same time, the
question would be how overclockable the 755 processor is,
especially if you cannot bump up Vcore. (Two examples of
ways to overclock, would be a hardware BSEL style mod to
change what the clock generator delivers for an input clock.
Or to use a program such as SETFSB, which modifies the clock
generator settings while in Windows. Not all clock generators
are supported by that program. It works on my current motherboard.)

http://www13.plala.or.jp/setfsb/

Based on some of the other sections of the Intel specification,
my guess would be that chipset was selected to give as long a
battery life as possible. That is why it only has single channel memory,
as each memory channel increases the load on the battery.
A "desktop replacement" style of laptop, has more room
for burning up power, and the laptop designer could select
dual channel, more powerful separate GPU and so on.

Paul