Memory voltage problem...

Hello

I've been using 2 gigabytes of OCZ Platinum 6400 for a few months now
with no problems. Recently I was experiencing problems when loading
games so I ran memtest which discovered errors when reading and
writing from the RAM. The RAM is running at its rated timings of 4 4
4
15 1T at 2.1V.
I found that when I lowered the voltage to 2.0V the errors vanished.


Is this something I should be worried about as OCZ state that the
memory should support voltages up to 2.2V


I would appreciate any advice


Thanks

"Calum" <calumtomeny@gmail.com> wrote in message
news:ef3942de-605c-43fa-9827-69401af3255e@r60g2000hsc.googlegroups.com...
> Hello
>
> I've been using 2 gigabytes of OCZ Platinum 6400 for a few months now
> with no problems. Recently I was experiencing problems when loading
> games so I ran memtest which discovered errors when reading and
> writing from the RAM. The RAM is running at its rated timings of 4 4
> 4
> 15 1T at 2.1V.
> I found that when I lowered the voltage to 2.0V the errors vanished.
>
>
> Is this something I should be worried about as OCZ state that the
> memory should support voltages up to 2.2V
>
If a new memtest proves OK, you`re good.

Calum wrote:
> Hello
>
> I've been using 2 gigabytes of OCZ Platinum 6400 for a few months now
> with no problems. Recently I was experiencing problems when loading
> games so I ran memtest which discovered errors when reading and
> writing from the RAM. The RAM is running at its rated timings of 4 4
> 4
> 15 1T at 2.1V.
> I found that when I lowered the voltage to 2.0V the errors vanished.
>
>
> Is this something I should be worried about as OCZ state that the
> memory should support voltages up to 2.2V
>
>
> I would appreciate any advice
>
>
> Thanks

There are a couple possible boundary conditions.

If there is too little voltage, the memory won't make its rated
timing. This is because the manufacturers insist on using the
timing they get via "overclocker" conditions.

If the industry was an honest one, all memory would be rated by
its 1.8V abilities. You'd apply the "normal" voltage of 1.8V for
DDR2, and the memory would meet its timing. Then it would be up
to the enthusiast, to apply more voltage and get more performance.

Instead, for enthusiast memory, they test at the higher voltage
condition. Not all motherboards have voltage adjustments, and
not all can be adjusted as high as the module maker's boosted spec.

So, if less voltage is being used, it can have an impact on the
tightness of the timing that can be achieve. Maybe you have to
go to 5-5-5 at 1.8V, instead of 4-4-4 at 2.0V.

Now, as more voltage is applied, the memory can get hot. It really
depends on how the memory is designed. I found one academic paper
from the DDR era, that suggested portions of a memory chip, use
a voltage regulator. Such a regulator throws off excess heat, if a
higher supply voltage is used. The memory chip designer, is designing
for 1.8V, and if 2.2V is applied, that is outside their expectation.

Based on that, I'd try the finger test. Run at 2.2V. Measure the
temperature of the DIMM with your finger. Does it seem hot ?
Does the temp change measurably at 2.1V ?

When adjusting memory, you should only apply as much voltage as is
needed to eliminate errors, Keep reducing the voltage, until memtest
reports errors. Then bump it up a notch for your final setting.

Memory performance will change with temperature (it is based on CMOS
design principles as far as I know). CMOS get slower with elevated
temperature. So turning down the voltage, may reduce the temp enough
to help it meet timing. But too little voltage can also prevent
it from making timing.

Some enthusiasts, who are overvolting memory a lot, place a fan over
the DIMM slots, to enhance cooling. Personally, I'd rather crank
things down a notch, so that only ordinary cooling is required.
I use the odd extra fan in my computers, but I don't know how I'd
suspend one over the DIMMs :-)

Paul

On Nov 26, 7:39 pm, Paul <nos...@needed.com> wrote:
> Calum wrote:
> > Hello
>
> > I've been using 2 gigabytes of OCZ Platinum 6400 for a few months now
> > with no problems. Recently I was experiencing problems when loading
> > games so I ran memtest which discovered errors when reading and
> > writing from the RAM. The RAM is running at its rated timings of 4 4
> > 4
> > 15 1T at 2.1V.
> > I found that when I lowered the voltage to 2.0V the errors vanished.
>
> > Is this something I should be worried about as OCZ state that the
> > memory should support voltages up to 2.2V
>
> > I would appreciate any advice
>
> > Thanks
>
> There are a couple possible boundary conditions.
>
> If there is too little voltage, the memory won't make its rated
> timing. This is because the manufacturers insist on using the
> timing they get via "overclocker" conditions.
>
> If the industry was an honest one, all memory would be rated by
> its 1.8V abilities. You'd apply the "normal" voltage of 1.8V for
> DDR2, and the memory would meet its timing. Then it would be up
> to the enthusiast, to apply more voltage and get more performance.
>
> Instead, for enthusiast memory, they test at the higher voltage
> condition. Not all motherboards have voltage adjustments, and
> not all can be adjusted as high as the module maker's boosted spec.
>
> So, if less voltage is being used, it can have an impact on the
> tightness of the timing that can be achieve. Maybe you have to
> go to 5-5-5 at 1.8V, instead of 4-4-4 at 2.0V.
>
> Now, as more voltage is applied, the memory can get hot. It really
> depends on how the memory is designed. I found one academic paper
> from the DDR era, that suggested portions of a memory chip, use
> a voltage regulator. Such a regulator throws off excess heat, if a
> higher supply voltage is used. The memory chip designer, is designing
> for 1.8V, and if 2.2V is applied, that is outside their expectation.
>
> Based on that, I'd try the finger test. Run at 2.2V. Measure the
> temperature of the DIMM with your finger. Does it seem hot ?
> Does the temp change measurably at 2.1V ?
>
> When adjusting memory, you should only apply as much voltage as is
> needed to eliminate errors, Keep reducing the voltage, until memtest
> reports errors. Then bump it up a notch for your final setting.
>
> Memory performance will change with temperature (it is based on CMOS
> design principles as far as I know). CMOS get slower with elevated
> temperature. So turning down the voltage, may reduce the temp enough
> to help it meet timing. But too little voltage can also prevent
> it from making timing.
>
> Some enthusiasts, who are overvolting memory a lot, place a fan over
> the DIMM slots, to enhance cooling. Personally, I'd rather crank
> things down a notch, so that only ordinary cooling is required.
> I use the odd extra fan in my computers, but I don't know how I'd
> suspend one over the DIMMs :-)
>
> Paul- Hide quoted text -
>
> - Show quoted text -

Many thanks. I found this very useful.

I just wanted to make sure that my RAM was ok; I'll play with the
voltage and see how it runs at 1.9V with the same timing.

At least now I know that the RAM is not faulty.