Re: DDR memory

Rookie wrote:

> I am looking for a reliable brand, and I can buy dimms from those
> manufacturers:
>
> Supertalent (comes with heatsinks)
> TWIMOS
> KINGSTON
>
> The price is more or less the same. Which one would you suggest?

You may want to check the memory forums at places like www.overclockers.com,
www.hardOCP.com, www.tomshardware.com, etc.

I've heard good things about TwinMOS but have no experience with hit.

My track record with budget Kingston modules hasn't been good. One of
their 256MB PC2100 modules failed testing, and a whopping 8 out of 11
or 12 of their 512MB PC3200 modules failed. Also bad was a 512MB
PC3200 Mushkin whose SPD said was a Kingston. OTOH the 1GB PC5300
DDR2 modules I bought a few months ago were fine, even when
overclocked.

I've never heard of Supertalent, but I bought my first heatsinked
memory a few days ago, a matched pair of 1GB OCZ Gold PC6400 modules.
Their heatsinks are nothing but a nuisance, and one of the modules
showed a few bad spots, always in the 26th bit, whether run alone or
with the other module. The next pair I tried had a much worse module
that made the computer beep-beep-beep on boot, every time when run
with the other module, half the time when run alone. But the third
pair of OCZ modules have so far passed all diagnostics. BTW, each
time I went to the store to exchange the bad OCZ, I saw a customer
trying to return some, too.

Can you buy Crucial memory? Because it's about the only retail memory
that always comes with good chips, either Micron (Micron owns Crucial)
or Samsung. Almost every other brand is made with untested chips (UTT
-- a search will return horror stories).

The diagnostics I use are from www.MemTest86.com (but MemTest+, from
www.MemTest.org, is very, very similar) and Gold Memory ver. 5.07 from
www.GoldMemory.cz. Both were highly praised in a review by www.RealWorldTech.com,
but I found that newer versions of Gold Memory weren't nearly as good
at finding errors. I ran each program overnight or until errors were
reported, whichever came first. I run both programs because some
modules have passed one of these diagnostics but failed the other.

Rookie wrote:

> Thanks for the suggestions. I can find Crucial memory, but the price is
> really expensive! Though I do use the memory diagnostics, I find they
> can't be precise and the result is too many 0x1e stop errors.
>
> I think I will go for the Twinmos memory.

Have you been running the diagnostics from Windows? Because the only
time most memory tests work correctly is when they're booted from
their own floppy, CD-ROM, or USB drive and have direct control of the
computer. But it's common for error detection to be imprecise when
the memory is only marginally defective (but still unacceptable for
use).

On Sun, 2 Dec 2007 06:08:39 +0100 (CET), Rookie
<rookie@hates.spam> wrote:

>"larry moe 'n curly" <larrymoencurly@my-deja.com> wrote in news:3125162b-
>b80d-48f3-9557-9b1cc7ac3e10@e23g2000prf.googlegroups.com:
>
>> Rookie wrote:
>>
>>> Thanks for the suggestions. I can find Crucial memory, but the price is
>>> really expensive! Though I do use the memory diagnostics, I find they
>>> can't be precise and the result is too many 0x1e stop errors.
>>>
>>> I think I will go for the Twinmos memory.
>>
>> Have you been running the diagnostics from Windows? Because the only
>> time most memory tests work correctly is when they're booted from
>> their own floppy, CD-ROM, or USB drive and have direct control of the
>> computer. But it's common for error detection to be imprecise when
>> the memory is only marginally defective (but still unacceptable for
>> use).
>>
>
>I use the memtest modules that run from a boot cd, both the ms and the open
>source ones. However, I find that they may run fine for days, and still
>have problems with memory when under windows.

When you see these problems what do you do to resolve it?

If Memtest86+ doesn't find a hard error on an address
regularly reocurring, lowering the bus speed may then
temporarily remove the error state (providing timings are
manually kept higher instead of allows to drop via the SPD
table)

On Sun, 2 Dec 2007 18:16:12 +0100 (CET), Rookie
<rookie@hates.spam> wrote:


>>>I use the memtest modules that run from a boot cd, both the ms and the
>>>open source ones. However, I find that they may run fine for days, and
>>>still have problems with memory when under windows.
>>
>> When you see these problems what do you do to resolve it?
>>
>> If Memtest86+ doesn't find a hard error on an address
>> regularly reocurring, lowering the bus speed may then
>> temporarily remove the error state (providing timings are
>> manually kept higher instead of allows to drop via the SPD
>> table)
>>
>
>Indeed this may be the case, as both memory modules worked fine on an
>older motherboard. Since DDR has reached its technology peak, I hope the
>modules I am gonna buy will be fine.
>
>I could try lowering the ddr speed from bios settings, but somehow I
>think it would be better to buy some new modules.


This brings me to my other point, that if you are having
windows problems, it is not necessarily the memory to blame
unless the problems completely go away by getting that
memory stable, or being able to swap in other memory.

We just don't know enough about your situation to know is my
point, that often windows errors referencing memory have
nothing to do with memory errors, it just lists an address
but it was really a driver or something else to blame.