Memory woes

On Mon, 20 Mar 2006 12:55:14 -0800, beerspill wrote:


The little lost angel wrote:

Definitely the cheaper ValueRAM :P I started on them simply because at
that time, they were the only ones cheap yet with lifetime warranty
and a local policy of no hassle one to one exchange. Since I've very
little problems with them, I've continued using/recommending them. But
bear in mind I come into contact with, including those for friends,
only a small number of Kingston modules over the years.

In comparison, my friend who tried the PNY does something in the order
of hundreds of computers in a year. I asked him after my first post on
this issue and he claims that he has used close to 2000 DIMMs in the
past year including PNY, Patriot (their cheap PDP modules), Twinmos and
Kingston. Kingston has given him the least grief, PDP and Twinmos
while not perfect has been generally ok while nothing quite bit him
like PNY did. Personally I must add that it could just be his luck
with a bad batch.

Take a look at Thaiphoon, an SPD viewer/editor:
http://cbid.amdclub.ru/html/download.html

I've gotten many DIMMs to work reliably by simply using it to reprogram
the tC setting from 1T to 2T, a setting not alterable with every BIOS..

....and you think this is acceptable? If it doesn't work as advertized,
send it back!

--
Keith

On Mon, 20 Mar 2006 19:51:33 +0000, wrote:

On Mon, 20 Mar 2006 09:21:37 -0600, Alan Walpool
wrote:


After reading this thread it really makes you wonder about memory. I
have purchased many different brands of memory over the years and I
have so called brand name memory to fail. Anyway the promise was that
memory manufacturing is good now you did not need ECC memory or even
parity memory because the memory failure rate was so low.

Anyway I try to purchase motherboards that support ECC but this is
hard to find these days. Almost impossible with AMD64 consumer chips.
It is interesting the last motherboard I had that was supposed to
support ECC did not support ECC.

I have read similar threads like this one and it make the case even
more for ECC memory and ECC memory support. It sure would save running
memtest86+.

Whatever I guess you call this progress. ;-\

Alan

Go with socket 940 - they all support ECC. In fact, almost all
registered memory (req'd for 940) comes with ECC.

....and that's a *good* thing! I wouldn't go socket-940 these days though.
The price penalty is far too high for a desktop.

--
Keith

Trent wrote:
On Sun, 19 Mar 2006 16:13:25 GMT

(The little lost angel) wrote in Message id:
:

YMMV but for me, Kingston's been a pretty good brand and I do
recommend it for a lot of my friends.

Frankly, I think they suck. You never know whose third tier, gray market
chips they're going to put on them.

The worst Kingston among the ones I recently tried probably had chips
that Kingston themselves had cut out of wafers and put into plastic
packages.

On Tue, 21 Mar 2006 16:59:09 +0000, wrote:

On Mon, 20 Mar 2006 21:41:09 -0500, Keith wrote:

On Mon, 20 Mar 2006 19:51:33 +0000, wrote:

On Mon, 20 Mar 2006 09:21:37 -0600, Alan Walpool
wrote:


After reading this thread it really makes you wonder about memory. I
have purchased many different brands of memory over the years and I
have so called brand name memory to fail. Anyway the promise was that
memory manufacturing is good now you did not need ECC memory or even
parity memory because the memory failure rate was so low.

Anyway I try to purchase motherboards that support ECC but this is
hard to find these days. Almost impossible with AMD64 consumer chips.
It is interesting the last motherboard I had that was supposed to
support ECC did not support ECC.

I have read similar threads like this one and it make the case even
more for ECC memory and ECC memory support. It sure would save running
memtest86+.

Whatever I guess you call this progress. ;-\

Alan

Go with socket 940 - they all support ECC. In fact, almost all
registered memory (req'd for 940) comes with ECC.

...and that's a *good* thing! I wouldn't go socket-940 these days though.
The price penalty is far too high for a desktop.

But this is the only way to get 2+ sockets, if you need or, what the
heck, just _want_ it.

;-)

Well, these days it just might make sense to
wait for the next gen DDR2 Opterons...

IMO, it never makes sense to wait. There is *always* something better
coming down the pike. I bought my S/940 a few months before S/939 was
available. I'm not sorry at all.

--
Keith

Keith wrote:
On Mon, 20 Mar 2006 12:55:14 -0800, beerspill wrote:

Take a look at Thaiphoon, an SPD viewer/editor:
http://cbid.amdclub.ru/html/download.html

I've gotten many DIMMs to work reliably by simply using it to reprogram
the tC setting from 1T to 2T, a setting not alterable with every BIOS..

...and you think this is acceptable? If it doesn't work as advertized,
send it back!

Of course it's not acceptable, but the quality of retail memory is now
like the quality of automobiles back in the 1970s - horrible, and
sending bad memory back often results in nothing but a waste of time
and money and more bad memory. So the most practical solution may be
to simply correct the overly optimistic SPD settings and reduce the
computer's performance by only an inperceptible amount

I've never accepted subpar memory, or at least I've never paid for any.
I do have 2 modules that eventually cost nothing, thanks to credit
card chargebacks, and they work fine when slowed from their rated
PC3200 to PC2700.

On Tue, 21 Mar 2006 22:12:47 -0500, Keith wrote:

....snip...

Go with socket 940 - they all support ECC. In fact, almost all
registered memory (req'd for 940) comes with ECC.

...and that's a *good* thing! I wouldn't go socket-940 these days though.
The price penalty is far too high for a desktop.

But this is the only way to get 2+ sockets, if you need or, what the
heck, just _want_ it.

;-)

Well, these days it just might make sense to
wait for the next gen DDR2 Opterons...

IMO, it never makes sense to wait. There is *always* something better
coming down the pike. I bought my S/940 a few months before S/939 was
available. I'm not sorry at all.

To the contrary, I'd rather hold until the old system starts giving
too much aggravation so the upgrade becomes inevitable. But as long
as the old clunker still gets the job done I wouldn't spend a penny.

On Wed, 22 Mar 2006 01:53:57 -0800, beerspill wrote:

Keith wrote:
On Mon, 20 Mar 2006 12:55:14 -0800, beerspill wrote:

Take a look at Thaiphoon, an SPD viewer/editor:
http://cbid.amdclub.ru/html/download.html

I've gotten many DIMMs to work reliably by simply using it to reprogram
the tC setting from 1T to 2T, a setting not alterable with every BIOS..

...and you think this is acceptable? If it doesn't work as advertized,
send it back!

Of course it's not acceptable, but the quality of retail memory is now
like the quality of automobiles back in the 1970s - horrible, and
sending bad memory back often results in nothing but a waste of time
and money and more bad memory. So the most practical solution may be
to simply correct the overly optimistic SPD settings and reduce the
computer's performance by only an inperceptible amount

No, *demand* the product you were sold! Practicalities have little to do
with it. If you accept this crap you're aiding the manufactures in
their, umm, optimistic testing.

I've never accepted subpar memory,

You just said you did!

or at least I've never paid for any.
I do have 2 modules that eventually cost nothing, thanks to credit
card chargebacks, and they work fine when slowed from their rated
PC3200 to PC2700.

If you charged back the memory, shouldn't you have returned it?

On Wed, 22 Mar 2006 19:26:49 +0000, wrote:

On Tue, 21 Mar 2006 22:12:47 -0500, Keith wrote:

...snip...

Go with socket 940 - they all support ECC. In fact, almost all
registered memory (req'd for 940) comes with ECC.

...and that's a *good* thing! I wouldn't go socket-940 these days though.
The price penalty is far too high for a desktop.

But this is the only way to get 2+ sockets, if you need or, what the
heck, just _want_ it.

;-)

Well, these days it just might make sense to
wait for the next gen DDR2 Opterons...

IMO, it never makes sense to wait. There is *always* something better
coming down the pike. I bought my S/940 a few months before S/939 was
available. I'm not sorry at all.

To the contrary, I'd rather hold until the old system starts giving
too much aggravation so the upgrade becomes inevitable. But as long
as the old clunker still gets the job done I wouldn't spend a penny.

....and your dissagreement is what, exactly?

--
Keith