Should RAM timings have to be set manually?

[Jon Danniken]

Hello,

I bought a pair of Kingston HyperX RAM sticks, and tried to use them today;
these ones:
http://www.newegg.com/Product/Product.asp?Item=N82E16820144114

Unfortunately, instead of the 2-3-2-6 timings they are sold at, they come up
(auto) on the board (Epox 9NDA3I) as 2.5-3-3-8 timings. Attempting to set
them at the stated speed results in the board slowing them down to 166MHz
(PC 2700, DDR 333), and setting the memory back up to 200MHz results in
errors in memtest86.

I was rather surprised when the tech support fellow at Kingston told me that
I would have to set the memory timings manually in order to get them to run
at the specified timings.

So, my question is, is it generally required to have to set memory timings
manually in order to get the timings that are specified by the manufacturer?

Also, is it generally required to have to manually increase the voltage of
memory in order to get them to run at the timings specified by the
manufacturer?

Thanks,

Jon

 

[Guest]

Hello,

I bought a pair of Kingston HyperX RAM sticks, and tried to use them today;
these ones:
http://www.newegg.com/Product/Product.asp?Item=N82E16820144114

Unfortunately, instead of the 2-3-2-6 timings they are sold at, they come up
(auto) on the board (Epox 9NDA3I) as 2.5-3-3-8 timings. Attempting to set
them at the stated speed results in the board slowing them down to 166MHz
(PC 2700, DDR 333), and setting the memory back up to 200MHz results in
errors in memtest86.

I was rather surprised when the tech support fellow at Kingston told me that
I would have to set the memory timings manually in order to get them to run
at the specified timings.

So, my question is, is it generally required to have to set memory timings
manually in order to get the timings that are specified by the manufacturer?

Also, is it generally required to have to manually increase the voltage of
memory in order to get them to run at the timings specified by the
manufacturer?

The SPD chip on most memory modules is set to JEDEC
standards, regardless of the module's actual speed and
timing capabilities. Manufacturers do this so they can
maintain compliance with JEDEC standards. A system bios
(in auto mode) simply reads whatever timings are stored on
the SPDs, and it's up to the end-user to override these
default timings.

Also, whether a module needs extra voltage to achieve its
rated speed specs depends on the memory chips used in
the module. Some chips need a voltage boost, others don't.
For this reason manufacturers have begun specifying a
voltage range for a given module (e.g. 3.1-3.3V) instead of
a single number.

 

[frodo]

use something like CPU-Z or WCPU-ID to find out what the ram's reported
SPD settings are - these SHOULD be what the mobo wants to set them at by
default. The SPD values may not match the given spec, but the ram should
work at the given spec if you manually set it there. Bump voltage 0.1
volt up if a memtest86+ fails. If still no good at stated spec then send
'em back and get another brand.

are there any other sticks in there? the mobo will need to set a speed
that is compat w/ all sticks.

ALWAYS ALWAYS ALWAYS do a memtest to validate new ram BEFORE booting into
any OS on the HD - boot from floppy and make sure it passes! Booting to an
OS has the potential to corrupt the HD (HD data buffers are stored in RAM
prior to writting, so if ram is bad data written to disk can be bad too.)
NEVER install an OS to HD if ram is not 100% perfect!

 

[Jon Danniken]

use something like CPU-Z or WCPU-ID to find out what the ram's reported
SPD settings are - these SHOULD be what the mobo wants to set them at by
default. The SPD values may not match the given spec, but the ram should
work at the given spec if you manually set it there. Bump voltage 0.1
volt up if a memtest86+ fails. If still no good at stated spec then send
'em back and get another brand.

are there any other sticks in there? the mobo will need to set a speed
that is compat w/ all sticks.

ALWAYS ALWAYS ALWAYS do a memtest to validate new ram BEFORE booting into
any OS on the HD - boot from floppy and make sure it passes! Booting to an
OS has the potential to corrupt the HD (HD data buffers are stored in RAM
prior to writting, so if ram is bad data written to disk can be bad too.)
NEVER install an OS to HD if ram is not 100% perfect!

Hi Frodo,

Yeah, this is a new build, on the kitchen table ATM, haven't even hooked up
a HDD yet. That is the only RAM in it right now; is there a way to get
CPU-Z or WCPU-ID to run from from DOS?

BTW, when it failed Memtest, it actually crashed it; I had to hard boot the
machine with the power button. .

I'll try bumping up the voltage a tenth of a volt. Otherwise I'm thinking
of getting the Crucial Ballistix (2 x 512) instead; any thoughts on those?

Thanks,

Jon

 

[Jon Danniken]

"(e-mail address removed)"

The SPD chip on most memory modules is set to JEDEC
standards, regardless of the module's actual speed and
timing capabilities. Manufacturers do this so they can
maintain compliance with JEDEC standards. A system bios
(in auto mode) simply reads whatever timings are stored on
the SPDs, and it's up to the end-user to override these
default timings.

Also, whether a module needs extra voltage to achieve its
rated speed specs depends on the memory chips used in
the module. Some chips need a voltage boost, others don't.
For this reason manufacturers have begun specifying a
voltage range for a given module (e.g. 3.1-3.3V) instead of
a single number.

Interesting, thanks for that. Now if only the actual SPD timings would be
listed by the manufacturers.

Jon

 

[Jon Danniken]

...is there a way to get
CPU-Z or WCPU-ID to run from from DOS?

Following up, there isn't - CPU-Z needs a dos box (not straight DOS), and
WCPUID won't either.

Any other way to read SPD from DOS?

Jon

 

[larry moe 'n curly]

Any other way to read SPD from DOS?

MemTest86 will display the main timings, at least when run on a mobo
equipped with an nVidia nForce chipset.

I've never bought Kingston HyperX memory, only ValueRAM, but I've found
that it can't be overclocked at all, and often it needs to be run at
slower than the SPD settings, especially the 1T/2T command time
parameter (but many BIOSes seem to set this to 2T, even when the SPD
says 1T). Oddly, I've never had a failure with PNY brand DDR memory.

 

[Jon Danniken]

I've never bought Kingston HyperX memory, only ValueRAM, but I've found
that it can't be overclocked at all, and often it needs to be run at
slower than the SPD settings, especially the 1T/2T command time
parameter (but many BIOSes seem to set this to 2T, even when the SPD
says 1T). Oddly, I've never had a failure with PNY brand DDR memory.

Thanks, Larry. Indeed only changing the command rate to 1T caused a failure
in these sticks. I don't know if I got a bad batch, or if it is indicative
of all Kingston HyperX, but I'm not planning to find out. These are going
back to where they came from, and a different brand showing up in their
place.

Jon

 

[frodo]

No, but memtest86+ does show the SPD values (actually, I think it shows
the values BEING USED, not necc. the SPD reported values.)

You could try booting a cd rom (http://www.ubcd4win.com/), I think it has
cpu-z on it, if not you can add whatever you want to it when creating the
UBCD4Win. This is a useful tool to have, take the time to make it now,
it'll save ya someday!

 

[Jon Danniken]

No, but memtest86+ does show the SPD values (actually, I think it shows
the values BEING USED, not necc. the SPD reported values.)

Indeed; it shows the values being used. I read that CPU-Z works in DOS, but
it turns out it only runs in a dos window, not pure DOS.

You could try booting a cd rom (http://www.ubcd4win.com/), I think it has
cpu-z on it, if not you can add whatever you want to it when creating the
UBCD4Win. This is a useful tool to have, take the time to make it now,
it'll save ya someday!

Aye, thanks. I did d/l Bart's PE the other night, gonna have to put it
together and do some poking around with it.

In the meantime, I ordered a pair of the Crucial Ballistix (512 x 2) sticks,
so I'll see how those do when they arrive.

Thanks,

Jon

 

[larry moe 'n curly]

I don't know if I got a bad batch, or if it is indicative
of all Kingston HyperX, but I'm not planning to find out. These are going
back to where they came from, and a different brand showing up in their
place.

My batting average for Kingston ValueRAM is 8 bad PC3200 modules out of
11 or 12 and maybe 30% bad PC2100 modules. The number of errors seemed
to correlate with the markings on the chips

 

[Rod Speed]

Jon Danniken wrote

My batting average for Kingston ValueRAM is 8 bad PC3200 modules
out of 11 or 12 and maybe 30% bad PC2100 modules. The number
of errors seemed to correlate with the markings on the chips

They aint 'bad', the bios just isnt using the appropriate timing etc for them.

 

[Jon Danniken]

My batting average for Kingston ValueRAM is 8 bad PC3200 modules out of
11 or 12 and maybe 30% bad PC2100 modules. The number of errors seemed
to correlate with the markings on the chips

Damn, that's some bad sticks there!

Further experimentation on mine (still the HyperX PC3200) reveals that
either single stick will run just fine at 1T command rate, but putting them
both in (dual channel) gives errors with 2T command rate (errors in Memtest
Test #8, to be precise). .

Hmmmmmmmm.

Jon

 

[kony]

They aint 'bad', the bios just isnt using the appropriate timing etc for them.

Depends on your definition of appropriate. If the bios uses
the SPD programmed timings and can't run them stable in
situations where other brands of memory can run at their
same, SPD programmed timings, it does indicate a problem.

The problem might be that the Kingston specs aren't as
conservative as other modules. Trying enough boards you
might find one that can run them but they're not sold as
board specific modules so that can't rule out boards
demonstrating stability with other modules.

 

[larry moe 'n curly]

They aint 'bad', the bios just isnt using the appropriate timing etc for them.

One BIOS was in an Asrock/Asus KT400 mobo, the other in an ECS nForce3
mobo. Thaiphoon reported the same SPD timings for all of these PC3200
modules, but all the modules with the completely unmarked chips all
failed, most of those with Dxxxxxx chips failed, while almost all the
modules with Fxxxxx chips tested fine, and even the failing ones showed
only 3-5 errors.

 

[Rod Speed]

Depends on your definition of appropriate.
Nope.

If the bios uses the SPD programmed timings and can't run them
stable in situations where other brands of memory can run at
their same, SPD programmed timings, it does indicate a problem.

'a problem' isnt the same thing as BAD ram.

The problem might be that the Kingston specs
aren't as conservative as other modules. Trying
enough boards you might find one that can run them

Makes more sense to try different timing detail etc.

but they're not sold as board specific modules so that can't
rule out boards demonstrating stability with other modules.

Separate matter entirely to whether the module is BAD.

 

[Rod Speed]

Rod Speed wrote

One BIOS was in an Asrock/Asus KT400
mobo, the other in an ECS nForce3 mobo.

The technical term for that is 'pathetically inadequate sample'

Thaiphoon reported the same SPD timings for all of these PC3200
modules, but all the modules with the completely unmarked chips all
failed, most of those with Dxxxxxx chips failed, while almost all the
modules with Fxxxxx chips tested fine, and even the failing ones
showed only 3-5 errors.

Very unlikely indeed that all the ones that were
unmarked were deliberately shipped by Kingston BAD.

Much more likely that the data in the spd is incorrect with SOME chipsets etc.

And we are currently seeing with DDR2 ram, some motherboard bios
like one Gigabyte particularly going thru plenty of revisions of the bios
to even allow the system to boot with known good high quality ram, so
its not just a simple matter of reading the spd and using the values in it.

 

[larry moe 'n curly]

The technical term for that is 'pathetically inadequate sample'

That's all I can afford. OTOH why did all the Corsair and PNY modules
work fine with those mobos? The only modules that performed as badly
as the Kingstons were Muskin Enhanced (budget) -- all three showed lots
of errors, even when the command timing was set to 2T. Oddly,
Thaiphoon said that those Mushkins were made by Kingston.

Very unlikely indeed that all the ones that were
unmarked were deliberately shipped by Kingston BAD.

Why would they ship such chips, especially considering the costs of
warranty replacements?

 

[Rod Speed]

Rod Speed wrote

That's all I can afford.

Irrelevant to your original claim.

OTOH why did all the Corsair and PNY modules work fine with those mobos?

It liked those better. Doesnt mean that the others are BAD.

The only modules that performed as badly as the Kingstons
were Muskin Enhanced (budget) -- all three showed lots of
errors, even when the command timing was set to 2T. Oddly,
Thaiphoon said that those Mushkins were made by Kingston.

Nothing odd about it, they would all have worked fine with appropriate timing etc.

Why would they ship such chips, especially
considering the costs of warranty replacements?

Basically because on the whole they do end up ahead.

Not all motherboards have a problem with them.

 

[kony]

Irrelevant to your original claim.


It liked those better. Doesnt mean that the others are BAD.

If the board runs stable with the other memory at certain
timings, but can't run stabiliy with that memory at same
timings, and those timings are what the Kingston was spec'd
to use, it implicates the memory.