Hey guys...

I'm looking for a deep technical answer on this question from someone
who knows something about chip design or at least chip quality
control, technical specs, etc.

I'm curious about whether undervolting can actually damage a chip. I
know it could make it unstable, but could it actually damage the chip
or shorten its life span (or the life span of other components)? Is
this something to be concerned about or is it something like "yeah, it
might shorten the theoretical life span from 100 years to 50 years"?

The reason I'm asking is this comment:

"Depending on the voltage your adjusting you could fry the chip by
under-volting. Most chips these days require a few different voltages
to run, and if you lower one, there are sometimes sneak paths where
the logic that is not getting enough power can draw power from the I/O
voltage, or other auxiliary voltages on the chip which can burn out IO
and other parts since they were never designed to carry the amount of
current drawn through the sneak path. There are usually some
protections in place to prevent the burn out, but with millions of
transistors on a chip it's sometimes hard to catch all of the possible
sneak paths."

Here's the context of the comment:

http://forums.macrumors.com/showthread.php?p=5213018

Is this really a potential issue that one would face in the normal
life span of a computer? I would think that undervolting (keep in mind
no overclocking here) would be much safer than overvolting/
overclocking where I have heard fry stories.

-Adam

On Tue, 25 Mar 2008 09:07:54 -0700 (PDT), Adam Ierymenko
<(E-Mail Removed)> put finger to keyboard and composed:

>Hey guys...
>
>I'm looking for a deep technical answer on this question from someone
>who knows something about chip design or at least chip quality
>control, technical specs, etc.
>
>I'm curious about whether undervolting can actually damage a chip. I
>know it could make it unstable, but could it actually damage the chip
>or shorten its life span (or the life span of other components)? Is
>this something to be concerned about or is it something like "yeah, it
>might shorten the theoretical life span from 100 years to 50 years"?
>
>The reason I'm asking is this comment:
>
>"Depending on the voltage your adjusting you could fry the chip by
>under-volting. Most chips these days require a few different voltages
>to run, and if you lower one, there are sometimes sneak paths where
>the logic that is not getting enough power can draw power from the I/O
>voltage, or other auxiliary voltages on the chip which can burn out IO
>and other parts since they were never designed to carry the amount of
>current drawn through the sneak path. There are usually some
>protections in place to prevent the burn out, but with millions of
>transistors on a chip it's sometimes hard to catch all of the possible
>sneak paths."
>
>Here's the context of the comment:
>
>http://forums.macrumors.com/showthread.php?p=5213018
>
>Is this really a potential issue that one would face in the normal
>life span of a computer? I would think that undervolting (keep in mind
>no overclocking here) would be much safer than overvolting/
>overclocking where I have heard fry stories.
>
>-Adam

I know that in the early days of DRAMs and EPROMs, there were devices
(eg 4116 16Kx1 DRAM) which required three supply rails, +5V, -5V, and
+12V. For these chips, the -5V rail had to be applied before and
removed after the other supply voltages.

- Franc Zabkar
--
Please remove one 'i' from my address when replying by email.

"Adam Ierymenko" <(E-Mail Removed)> wrote in message
news:7f53d811-7dfc-404c-b6b6-(E-Mail Removed)...
> Hey guys...
>
> I'm looking for a deep technical answer on this question from someone
> who knows something about chip design or at least chip quality
> control, technical specs, etc.
>
> I'm curious about whether undervolting can actually damage a chip. I
> know it could make it unstable, but could it actually damage the chip
> or shorten its life span (or the life span of other components)? Is
> this something to be concerned about or is it something like "yeah, it
> might shorten the theoretical life span from 100 years to 50 years"?
>
> The reason I'm asking is this comment:
>
> "Depending on the voltage your adjusting you could fry the chip by
> under-volting. Most chips these days require a few different voltages
> to run, and if you lower one, there are sometimes sneak paths where
> the logic that is not getting enough power can draw power from the I/O
> voltage, or other auxiliary voltages on the chip which can burn out IO
> and other parts since they were never designed to carry the amount of
> current drawn through the sneak path. There are usually some
> protections in place to prevent the burn out, but with millions of
> transistors on a chip it's sometimes hard to catch all of the possible
> sneak paths."
>
> Here's the context of the comment:
>
> http://forums.macrumors.com/showthread.php?p=5213018
>
> Is this really a potential issue that one would face in the normal
> life span of a computer? I would think that undervolting (keep in mind
> no overclocking here) would be much safer than overvolting/
> overclocking where I have heard fry stories.

Hi,

There are 2 things that could go wrong:
1) What you're describing: when multiple voltage come into play and start to
leak, you can expect side effects like latch up or a higher voltage going to
a low-voltage region of the chip and frying your transistors which have only
a few atoms of isolation (typically, the IO is made with thicker oxide FETs
to run higher voltages than your core in order to pump the nescessary energy
fast enough to the peripheral chips).
2) If you have too much current flowing from one part of the chip to
another. Even if it comes from the same supply, the current drawn would be
too much to handle (imagine powering you CPU with 1 pair of pins instead of
the 100s they use). In this case, your electrical connection melts away like
a fuse.

Regards,
Alvin.

"Adam Ierymenko" <(E-Mail Removed)> wrote in message
news:7f53d811-7dfc-404c-b6b6-(E-Mail Removed)...
> Hey guys...
>
> I'm looking for a deep technical answer on this question from someone
> who knows something about chip design or at least chip quality
> control, technical specs, etc.
>
> I'm curious about whether undervolting can actually damage a chip. I
> know it could make it unstable, but could it actually damage the chip
> or shorten its life span (or the life span of other components)? Is
> this something to be concerned about or is it something like "yeah, it
> might shorten the theoretical life span from 100 years to 50 years"?
>
> The reason I'm asking is this comment:
>
> "Depending on the voltage your adjusting you could fry the chip by
> under-volting. Most chips these days require a few different voltages
> to run, and if you lower one, there are sometimes sneak paths where
> the logic that is not getting enough power can draw power from the I/O
> voltage, or other auxiliary voltages on the chip which can burn out IO
> and other parts since they were never designed to carry the amount of
> current drawn through the sneak path. There are usually some
> protections in place to prevent the burn out, but with millions of
> transistors on a chip it's sometimes hard to catch all of the possible
> sneak paths."
>
> Here's the context of the comment:
>
> http://forums.macrumors.com/showthread.php?p=5213018
>
> Is this really a potential issue that one would face in the normal
> life span of a computer? I would think that undervolting (keep in mind
> no overclocking here) would be much safer than overvolting/
> overclocking where I have heard fry stories.
>
> -Adam
>

There are a few main exposures that I can think of....

1. The N-wells are typically connected to one of the power supplies.
Which one is dependent on the design and what supplies are being used in
that circuit. If the well bias gets more than a diode drop below the
voltage that the Pfet is connected to, current will flow raising
disipation, and potentially causing latchup.

2. Some high voltage outputs may use multiple supplies and depend on
values of each to avoid overstressing...

3. Some level shift circuits might depend on power supply being in
tolerance..

However chips are usually designed so as not to require any power
sequencing. I can't promise that every chip in use is in fact designed
that way...

But I wouldn't worry about it particularily.

del