A computer components & hardware forum. HardwareBanter

If this is your first visit, be sure to check out the FAQ by clicking the link above. You may have to register before you can post: click the register link above to proceed. To start viewing messages, select the forum that you want to visit from the selection below.

Go Back   Home » HardwareBanter forum » General Hardware & Peripherals » General
Site Map Home Register Authors List Search Today's Posts Mark Forums Read Web Partners

New thread



 
 
Thread Tools Display Modes
  #11  
Old April 3rd 17, 09:39 PM posted to alt.comp.hardware
Norm X[_2_]
external usenet poster
 
Posts: 91
Default New thread

[major snippage]
Thanks Paul, I've narrowed my search to two old tools that only work
on WinXP to Win7, MemSet41 and CPU-Tweaker. I've copied them to my
BartPE dongle. When it is time, I'll test them on my overclocked 4GB
RAM to see if the RAM voltage really is 1.8V as it should be. I have
a quad core Q6600 CPU which is not really old. If I can see what I
need to see, I'll try to fix the RAM voltage on one RAM module. If
that works for a long time, I'll try plugging in the second 4GB RAM
module. If that works, I'll have 8GB RAM, suitable for the numerous
VM's I wish to run. If I cannot boot, I'll go back to one RAM module
and start from scratch.
You should also be looking at the motherboard, find the regulator
chip in the DRAM power section, and get the datasheet for it.
That will tell you some basic things about whether options
are even available to modify the setting.

Some modern chips, have excellent control options.

One I modified here, had a voltage offset pin, specifically
for boosting.

And lesser chips, they have a summing junction, where you can
inject current to modify the output. Some of those designs,
the overclock settings are way off (not regulated very well).

Paul
Thanks Paul,
For the Intel Q6600, I have downloaded a 30 day trial for Intel's Intel
Vtune Amplifier XE. I have not yet rebooted but the program runs and
shows some info. If I find anything useful I'll post it here.

This URL:
http://ark.intel.com/products/29765/...z-1066-MHz-FSB

says the Q6600 runs at a max 1.5V, so it cannot give 1.8V for RAM. But
maybe an alternative is to provide less aggressive RAM timings.

I still need to boot into BartPE to see if MemSet41 or CPU-Tweaker
address the MOBO.

If you expand this screenshot:
https://images.betanews.com/screensh...52682468-1.jpg

You'll see that CPU-tweaker does not show or modify RAM voltages, only
timings. Maybe underclocking would do the trick.


You want the product image, on the same web page.

http://ark.intel.com/inc/images/diagrams/diagram-5.gif

On LGA775, the MCH connects to memory, rather than the CPU.
It's the MCH specs you need.

This is an example of an MCH/Northbridge. Because back in the
day, we were still calling them Northbridges.

http://ark.intel.com/products/31919/...ory-Controller

We can then pull the datasheet for that sample MCH.

http://www.intel.com/Assets/PDF/datasheet/317610.pdf

VCC_DDR
1.8 V DDR2 / 1.5 V DDR3 System Memory
Supply Voltage with respect to VSS
-0.3 min 4.0 V max

The pad drivers (apparently) have a huge range.

Modern CPUs, where there is an IMC (integrated memory controller),
the voltage range is a lot lower. (1.65V max on 1.5V I/O devices.)

https://en.wikipedia.org/wiki/Stub_S...rminated_Logic

http://www.embedded.com/print/4007526

This is a JEDEC block diagram for SSTL_1.8. The VCCQ would
be 1.8V, VTT would be 0.9V, and the requirement for VTT regulator
is that it is push-pull, with current flow as large as a couple
amps (for the entire bus). The average current flow is a lot lower,
so all it needs is a decent sized cap to hold the voltage on the
source of VTT. If the address/control or data bus were to be floated,
the resting voltage is half way between the rails.

https://s17.postimg.org/3miofq03j/sstl_1_8.gif

Paul


Thanks Paul,

Now, only tweaking RAM voltage is expeditious. Given that the chips in
question are tiny, I need a circuit diagram for the MCP73VE MOBO and
details on whatever IMC is used. There appears to be two pads for testing
voltage in that area. From the BIOS Setup program, I get information I
have not found at the applications level.

CPU Vcore 1.248V
Chipset Voltage 1.344V

The voltages appear to be lower than specified. The Q6600 works fine, so
there is no need to change its voltage. If the Chipset Voltage of 1.344V
is used by RAM, it is far lower than 1.8V given in the RAM SPD block. One
and only one 4GB RAM module works at the current RAM voltage. With two
RAM sticks installed, the S/N ratio does not make memtest happy. I think
it would be best to increase RAM voltage by only one increment. We don't
want overvoltage on the entire Chipset. It would be self defeating to
burn out the MOBO.

One other problem is that the RAM voltage increment needs to be applied
at the hardware level. It won't work at the application level. First, I
should see if the two pads are voltage test points and test the voltage
if the pads see DC. If so, they could also be used as DC injection
points, after a suitable trace is cut.


This is as close as I can get right now, to a block diagram for your
board.
The search engine isn't giving me much to work with.

http://www.theregister.co.uk/Print/2..._nvidia_mcp73/

Those designs use a single chip for Northbridge and Southbridge. The chip
should have a decent-sized cooler on it, because it's a pretty busy beast.
Underneath the heatsink, is a bare silicon die. If putting the heatsink
back on it, you have to be careful not to tilt the heatsink and
crack the die. Those don't use a shim for safety (like an AthlonXP might).

If the GPU in the Northbridge is being used, that makes it a few watts
warmer when gaming. You might even make it to 15W for a chip like that.

In any case, the memory is not on the processor. No IMC involved.
Just a single chipset chip. The memory runs off that big single chip. The
chip could have 1500 balls on it.

The core (logic gates) of chipsets, run at the lowest voltage on
the board. Around 1V or less. On a chip like that, the core logic
could be drawing 5 amps. They use a low voltage, to try to keep
the power dissipation in bounds. On a 15W chip, you might want a
1 inch high heatsink, with a 40mm fan. Some of the cheesy designs,
don't bother with the fan, and the life of the board can be
extended in cases like that, by using a "blow-down" fan on the CPU.
It helps cool the chipset.

The I/O voltages are going to be higher than that. Chipset chips
have at least five rails connected to them. Since this chip
is a dual one, that only increases the potential number of
supplies on it.

You need to find the regulator next to the DIMMs, get the part number,
and see what you can learn. Just because there is a monitor point,
that's not a good place to be injecting anything.

On some of the chips, there is an actual VID interface. And you load
some ones and zeros on that bus, to set the voltage. That's the
"cadillac" interface. The board would use some GPIOs, and a BIOS
setting would drive out the appropriate number to the VID interface.

If a chip is considered a "fixed" design, you have a couple options.
There is a node called "FB", and you inject a tiny current to offset
the output voltage. If you apply too much, the OVP gets tripped,
and the regulator shuts down.

The other option, is the regulator has a "Boost" terminal, but the
same rules will apply. Too much exuberance on the part of
the user, and it'll shut down.

When I modified the Boost on my Asrock board, I only applied
about 0.1V of offset. There is an equation, relating the resistor
value to ground from Boost, to the amount of output modification.
If I'd doubled the offset, it would probably have tripped the
overvoltage protection.

The node coming from the output of the switcher, has zero output
impedance, and is under closed loop control. The output is "stiff"
and cannot be (easily) steered by injection.

If you manage somehow, to increase the voltage on that node, and
it passes the OVP protection point, they turn on the bottom MOSFET
and clamp the rail to ground. With no injector in place, such a design is
perfectly safe. The upper MOSFET is turned off, the lower one turned
on, the output voltage is zero - this feature exists to prevent
"burning" the customer inserted items, such as RAM or CPU.
The idea is, the motherboard burns (worst case), but the customer
components (RAM and CPU) are saved. But if you inject the node
in question, there will then be a fight between the lower MOSFET
clamping the node to ground, and your fancy injector trying to
raise the node.

I would say a good deal of caution is required. If you're going
to boost, you at least need the datasheet for the regulator.
Any decent "volt modder" knows this. If you don't have a schematic,
at least have a datasheet for each regulator you intend to mod.

In addition, other voltages like Vtt are involved. But I'm
going to pretend for the sake of argument, that the amount
of boost applicable, is too small to bother with tweaking
Vtt or any other voltage for that matter.

Another point is, don't assume the voltage used is 1.8V anyway.
A good designer cheats, and is already using 1.9V
(1.8 plus 5%), bringing the design right up to the JEDEC
limit. Only on server motherboards, do you get strict
religion and exactly 1.8V. You'd be surprised how many
desktop designs are already slightly bumped.

Years ago, on my first motherboard, the motherboard makers
back then were making their own "local 3.3V" supply. Even though
the power supply made 3.3V, these guys were making 3.5V onboard.
I think you can guess why they might do this. Even back then,
they were boosting stuff. It's not clear why they thought this
was necessary - however, the game was up, when you inserted something
like a TNT2, it wanted 6 amps of current from that regulator, and
the regulator was designed for a lot less. You got a black screen
(and the TNT2 was returned for money-back at GameStop :-) ).

Paul


Careful study of the M1640 manual images says I want to know more about the
ITE 8718FX, near the DIMM slots. ITE is a fabless company in Taiwan, 8718FX
is the IC ID. When I do Google search I get a lot of hits for intelligence
agencies. Either they do their own service work, or this IC is hackable.


  #12  
Old April 3rd 17, 10:56 PM posted to alt.comp.hardware
Paul[_28_]
external usenet poster
 
Posts: 1,467
Default New thread

Norm X wrote:


Careful study of the M1640 manual images says I want to know more about the
ITE 8718FX, near the DIMM slots. ITE is a fabless company in Taiwan, 8718FX
is the IC ID. When I do Google search I get a lot of hits for intelligence
agencies. Either they do their own service work, or this IC is hackable.


That should be about a roughly 128 pin quad flat pack, and is your
SuperI/O. That doesn't have regulation functions. It has
things like floppy interface, two serial ports, PS/2 keyboard
and mouse ports, IRDA interface, and so on. It also typically
includes the hardware monitor, three voltages, three fan speeds,
three temperatures and so on.

While I didn't see a datasheet pop up for that, this has a
passing reference to its function.

http://manualmachine.com/acer/verito...manual/page:2/

Super I/O controller: ITE 8718FX

*******

Power regulators don't generally get too big. Maybe 48 or 64 pins max
or so. If a regulator has the MOSFETs inside the package, it may have
a heat slug in the bottom center, but there's no way to see
that from the outside. While there can be a thermal via field
visible on the backside of the motherboard, you'd have to take the
thing apart to examine it.

The gate capacitance of a power MOSFET can be on the order of 3000pF.
That's a large load, being driven at a high frequency. So instead,
the arch is like this.

Regulator ------ Phase 1 --- Buffer chip - Up MOSFET, Down MOSFET
------ Phase 2 --- Buffer chip - Up MOSFET, Down MOSFET
------ Phase N --- Buffer chip - Up MOSFET, Down MOSFET

By doing it that way, the main chip doesn't drive the MOSFETs directly.
Doing it this way would save money (and was done in the past)...

Regulator ------ Phase 1 ----------------- Up MOSFET, Down MOSFET
------ Phase 2 ----------------- Up MOSFET, Down MOSFET
------ Phase N ----------------- Up MOSFET, Down MOSFET

but the FCV^2 of the Phase driver outputs would cause
the Regulator chip to overheat, even with a heat slug
on the bottom. Using the buffer chips, means the heat is
spread out a bit.

And to show how loopy some of the motherboard designers are,
there is one motherboard, where a tiny 8 pin switcher with a
MOSFET inside, runs with a case temperature of 100C. And
that's before the room gets hot. You wouldn't know it was
there, unless you touched it... It was detected via an IR
camera photograph. That's how someone figured it out.

Paul
  #13  
Old April 13th 17, 06:28 AM posted to alt.comp.hardware
Norm X[_2_]
external usenet poster
 
Posts: 91
Default New thread

Careful study of the M1640 manual images says I want to know more about
the ITE 8718FX, near the DIMM slots. ITE is a fabless company in Taiwan,
8718FX is the IC ID. When I do Google search I get a lot of hits for
intelligence agencies. Either they do their own service work, or this IC
is hackable.


That should be about a roughly 128 pin quad flat pack, and is your
SuperI/O. That doesn't have regulation functions. It has
things like floppy interface, two serial ports, PS/2 keyboard
and mouse ports, IRDA interface, and so on. It also typically
includes the hardware monitor, three voltages, three fan speeds,
three temperatures and so on.

While I didn't see a datasheet pop up for that, this has a
passing reference to its function.

http://manualmachine.com/acer/verito...manual/page:2/

Super I/O controller: ITE 8718FX

*******

Power regulators don't generally get too big. Maybe 48 or 64 pins max
or so. If a regulator has the MOSFETs inside the package, it may have
a heat slug in the bottom center, but there's no way to see
that from the outside. While there can be a thermal via field
visible on the backside of the motherboard, you'd have to take the
thing apart to examine it.

The gate capacitance of a power MOSFET can be on the order of 3000pF.
That's a large load, being driven at a high frequency. So instead,
the arch is like this.

Regulator ------ Phase 1 --- Buffer chip - Up MOSFET, Down MOSFET
------ Phase 2 --- Buffer chip - Up MOSFET, Down MOSFET
------ Phase N --- Buffer chip - Up MOSFET, Down MOSFET

By doing it that way, the main chip doesn't drive the MOSFETs directly.
Doing it this way would save money (and was done in the past)...

Regulator ------ Phase 1 ----------------- Up MOSFET, Down MOSFET
------ Phase 2 ----------------- Up MOSFET, Down MOSFET
------ Phase N ----------------- Up MOSFET, Down MOSFET

but the FCV^2 of the Phase driver outputs would cause
the Regulator chip to overheat, even with a heat slug
on the bottom. Using the buffer chips, means the heat is
spread out a bit.

And to show how loopy some of the motherboard designers are,
there is one motherboard, where a tiny 8 pin switcher with a
MOSFET inside, runs with a case temperature of 100C. And
that's before the room gets hot. You wouldn't know it was
there, unless you touched it... It was detected via an IR
camera photograph. That's how someone figured it out.

Paul


Thanks again Paul. I'm also making some progress in setting RAM timings
using Memset. Memset 4.1 ran smoothly on my Acer Aspire One netbook. It
accepts the chip set. I could have changed RAM timings if I wanted to.
Googling suggested that my best bet would be with Memset 3.1, for the M1640.
However using the Way Back Machine, I was only able to obtain Memset 3.4. It
'almost' ran on my Acer M1640. Maybe the authors reply to email.
Alternatively, with time I could disassemble, alter and compile what I want.


 




Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

vB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Cant fine that thread Fievel J Mousekewitz Overclocking AMD Processors 1 August 1st 08 12:24 AM
New thread Robert E. Watts Packard Bell Computers 8 February 7th 08 09:57 PM
What happened to the IMD thread? Jeffrey Alsip Compaq Servers 5 September 15th 05 07:55 AM
Wow i just try x2 the thread and it not like it was. DDC Ati Videocards 0 August 21st 05 09:11 PM
Why doesn't - this thread die! Skeptic General 0 May 2nd 05 12:02 AM


All times are GMT +1. The time now is 08:14 AM.


Powered by vBulletin® Version 3.6.4
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Copyright ©2004-2024 HardwareBanter.
The comments are property of their posters.