Motherboard memory voltage at boot, before SPD is read?

[larrymoencurly]

When a motherboard first turns on, what voltage is applied to the DIMMs,
just before the SPD profiles are read?

Does the motherboard default to a high voltage first, like 1.65V for
DDR3, and then lower it if the SPD specifies lower voltage, or does
the motherboard start out at standard 1.50V and then adjust to the
SPD voltage spec?

 

[Flasherly]

When a motherboard first turns on, what voltage is applied to the DIMMs,
just before the SPD profiles are read?

Does the motherboard default to a high voltage first, like 1.65V for
DDR3, and then lower it if the SPD specifies lower voltage, or does
the motherboard start out at standard 1.50V and then adjust to the
SPD voltage spec?

The BIOS first reads the SPD EEPROM to configure the memory chip
controller (MMC). Unless the MB has support provisions for auto-
overclocking, implemented on some SPD tables, it's going to be a
manual adjustment affair in either case to override and save them
apart from BIOS defaults.

 

[Paul]

When a motherboard first turns on, what voltage is applied to the DIMMs,
just before the SPD profiles are read?

Does the motherboard default to a high voltage first, like 1.65V for
DDR3, and then lower it if the SPD specifies lower voltage, or does
the motherboard start out at standard 1.50V and then adjust to the
SPD voltage spec?

There's nothing to store the default.

A regulator, whether it's VCore or VDIMM, relies on hardware
strapping for an initial value. On VDIMM, where there are no
VID signals, the GPIOs are tri-stated, and straps on the
GPIOs would cause nominal voltage to be applied to the DIMMs
at T=0.

After the BIOS code has been running for a few milliseconds,
then it's possible to set up the voltages you want to use.
The GPIOs send a value for voltage to the VDIMM regulator,
such as if a boost is needed.

On an XMP DIMM, the boost value would be applied after the
SPD is read out. Since the SPD is independent of the memory
chips themselves, the memory chips don't have to be sane,
for the SPD to be parsed by the BIOS.

System memory is not needed immediately. The BIOS code can
be written for register-only operation at first, until
the memory is set up.

Motherboards have the option of using a double-start, if
the designers feel changing clocks or voltage will cause
hardware insanity. If all hardware can tolerate ramped
shifts in conditions, then a single start might be all
that is needed.

On my old Nforce2 motherboard, there was an
"overclocker chip", which stored VID signal values between
restarts. The motherboard could do a double-start, and
store the "boost value" of voltage, for the second start.
And to the user, this is transparent. The reset pulse
can be pretty narrow, and the effects not observable to
the naked eye.

On double-start designs, occasionally there are stability
issues at nominal conditions. Such as the processor
crashing, before the application of boost is complete.
But, hardware is supposed to be tested to work at
nominal So this is not necessarily a fault of the
design or anything. It should have worked. Maybe if
a processor has electromigration damage, it can no
longer survive the "meager diet" delivered at T=0.
It can actually be the application of an overclock
for a few years, that damaged the CPU. This has happened
to certain AMD CPUs. They can no longer even run at
nominal clock and voltage.

Paul

 

[Flasherly]

You did say before a BIOS SPL read...eXcuSe me -- correction: The MB
manufacturer meets compliance to JEDEC memory industry specifications,
standards of memory class determination. Known as Stub Series
Termination Logic voltage, its ratings as given under said standards
will default either to 1.8V for DDR2 and 1.5V for DDR3. Since the pin-
out slots are not interchangeable (by in large to negate a few extant,
oddballed cross-dresser designs), accordingly, there's no apparent
reason for a regulator stage to be pumping in 1.65V over JEDEC
publications. As applicable, however, the "Tigers" known to inhabit
the Pacific Rim well may deign during a course of implementation to
supersede a stricter sense of compliance, to include naming
conventions of a related complexity involving SSTL. Or so I suspect.
How else could there be such pieces of crap manufactured under the
namesake of a motherboard, such as BioStar? (....Duly bearing in
mind, innovation, of course, is invariably the two-sided sword when
cutting through lazy standards involving scribblers and dabblers,
loose women, or the thinest slice of bread imaginable, nobody really
is likely to notice amiss.)

 

[larrymoencurly]

The MB manufacturer meets compliance to JEDEC memory
industry specifications, standards of memory class
determination. Known as Stub Series Termination Logic
voltage, its ratings as given under said standards
will default either to 1.8V for DDR2 and 1.5V for DDR3.
Since the pin-out slots are not interchangeable (by in
large to negate a few extant, oddballed cross-dresser
designs), accordingly, there's no apparent reason for
a regulator stage to be pumping in 1.65V over JEDEC
publications. As applicable, however, the "Tigers"
known to inhabit the Pacific Rim well may deign during
a course of implementation to supersede a stricter
sense of compliance, to include naming conventions of
a related complexity involving SSTL. Or so I suspect.
How else could there be such pieces of crap manufactured
under the namesake of a motherboard, such as BioStar?

I thought manufacturers would start off their motherboards
at higher than normal voltage to make it more likely they'd
boot successfully with low quality DRAM. Most retail DRAM
is made from chips that are either factory rejects or
overclocked 30-100%. Here's some 2666 MHz Corsair made
from 1333 MHz chips:

http://www.xbitlabs.com/articles/memory/display/corsair-dominator-platinum_5.html#sect1

I have a BioStar G41-M7 motherboard with a minimum 1.95V
voltage for its DDR2 memory. The BIOS also allows setting
the voltage as high as 2.65V, or 0.3V above the absolute
maximum for DDR2 chips, and I think an older BIOS allowed
up to 2.75V. This is among the reasons BioStar is my 3rd
least favorite brand of motherboard.

 

[Yousuf Khan]

When a motherboard first turns on, what voltage is applied to the DIMMs,
just before the SPD profiles are read?

The DIMMs aren't activated until the SPD profiles are read. The
motherboard first boots from its non-volatile EEPROM modules, not from
its RAM.

Yousuf Khan

 

[Flasherly]

I thought manufacturers would start off their motherboards
at higher than normal voltage to make it more likely they'd
boot successfully with low quality DRAM. Most retail DRAM
is made from chips that are either factory rejects or
overclocked 30-100%. Here's some 2666 MHz Corsair made
from 1333 MHz chips:

http://www.xbitlabs.com/articles/memory/display/corsair-dominator-pla...

I have a BioStar G41-M7 motherboard with a minimum 1.95V
voltage for its DDR2 memory. The BIOS also allows setting
the voltage as high as 2.65V, or 0.3V above the absolute
maximum for DDR2 chips, and I think an older BIOS allowed
up to 2.75V. This is among the reasons BioStar is my 3rd
least favorite brand of motherboard.

The thought occurred and I agree about giving it the extra juice. I
never had a chance to get into them, one died on me within a year and
the other wouldn't hold the BIOS settings. Although both were dirt
cheap, so I guess I can afford to spit it out there once in awhile.
God, 3rd-least...that's some system. Remind me not to go there
anytime soon. Used to buy exclusively ABIT, if not ASUS, MSI is
nothing spectacular but has treated me well over the years for a
budget board (they're still viable imo, improved from a bad spell,
with their newer offerings), whereas more of late I've also been
looking over GigaByte without anything major to report -- used to know
a pretty good code-breaker who would dimly swear by them over his coke-
bottle glasses. My last ASUS failure was spectacularly horrific,
along with some people I've met on realtime hardware chatlines working
for them, I'm going along with a shift upwards on it's prices while QC
is somewhat abated.

JEDEC...who worries about that anymore when eying Ebay for surfeit
parts. Mis-bought my last batch for DDR2 when I required 3. Hardly
phased me with $12 and such sticks shipped free. I just get by on 1M
with a crappy FireFox periodically gobbling it all up through a
programming leak of cache into main memory (have a nifty restore
crash&burn restore feature, tho, from back when FireFox was a real
browser with real, not abandoned, extensions).

 

[Franc Zabkar]

The DIMMs aren't activated until the SPD profiles are read. The
motherboard first boots from its non-volatile EEPROM modules, not from
its RAM.

Yousuf Khan

I notice that DIMMs provide a separate supply pin (VDDSPD) for the SPD
serial EEPROM.

- Franc Zabkar

 

[Yousuf Khan]

I notice that DIMMs provide a separate supply pin (VDDSPD) for the SPD
serial EEPROM.

Exactly, and chances are likely that this is a very standard power input
voltage and current that probably never changes over generations.

Yousuf Khan

 

[Paul]

Exactly, and chances are likely that this is a very standard power input
voltage and current that probably never changes over generations.

Yousuf Khan

The only incentive for it to change, is if the motherboard chipset
can no longer tolerate such voltages (on SMBUS). As long as USB
continues to be supported, things might not change (as USB uses
a relatively high potential, on the legacy pins).

Paul

 

[larrymoencurly]

Since the SPD is independent of the memory chips themselves,
the memory chips don't have to be sane, for the SPD to be
parsed by the BIOS.

That explains it. Thanks, Paul!

I think I'll be editing the SPDs of some modules to see if I can
get them to work.

 

[Yousuf Khan]

The only incentive for it to change, is if the motherboard chipset
can no longer tolerate such voltages (on SMBUS). As long as USB
continues to be supported, things might not change (as USB uses
a relatively high potential, on the legacy pins).

Paul

And most PSU's have a variety of voltages that they output, plus there
are voltage converters onboard the motherboards these days, which take
the standard voltages supplied by modern PSU's and downstep them as needed.

Yousuf Khan

 

[Paul]

And most PSU's have a variety of voltages that they output, plus there
are voltage converters onboard the motherboards these days, which take
the standard voltages supplied by modern PSU's and downstep them as needed.

Yousuf Khan

It's voltage tolerance, that drives changes to features like this.
If the SPD runs off 5V, and the chipset only has 3.3V or 1.8V tolerant
I/O pads, then you can't connect 5V swing signals to it. That's what
I was referring to.

*******

It's not a supply side issue as such. The PSU has "high voltages".
Local regulators on the motherboard make any "low voltages" needed.
For example, on my old NForce2 board, there was an 8 pin DIP that
made 1.200V for the chipset core logic. Being a local regulator,
it gives a precise voltage as needed.

The power supply itself, isn't really that well regulated. There
are the three main rails (3.3,5,12V), and a heavy load on one,
affects the voltage on the others. (That's called cross-loading.)
When you look at how well the local regulators on the motherboard
can do the job, they're as steady as can be. By comparison.
And for lots of critical job, that's why there is a local
regulator.

Even the sound chip has local regulation, but that's to clean up
analog noise, so it doesn't leak into the sound chip.

Paul

 

[Flasherly]

And most PSU's have a variety of voltages that they output, plus there
are voltage converters onboard the motherboards these days, which take
the standard voltages supplied by modern PSU's and downstep them as needed.

Yousuf Khan

Seems a reasonable expectation there ought be lots and lots of variety
among sexy power supplies...

http://www.tomshardware.com/picturestory/612-17-gift-guide-pc-hardware-recommendation.html

Not a sole determinate of course to plainer varieties, even if
straightly appealing to sole-source 12V rails...

http://www.tomshardware.com/news/Triathlor-Enermax-PSU-Power-Supply,21213.html

 

[Flasherly]

Excuse the above links if you exercise discretionary filters. Tom's
of late evidently is infiltrated with 20 external links out to those
pages, according to my main browser's filter/blocker;- I found them
just after while using a second browser, as I couldn't duplicate the
site, apart from a standalone install with routines for wiping out
prior residuals, freezing the registry, and denying any attempts to
modify its plug-in support.