Installed Ramsinks on my 8500LE

[Joe]

Oddly enough i couldn't find any posts to verify if ramsinks really work.

I have a 6 month old Sapphire 8500LE stock clocked at 250/500 with 4ns
Eton memory. Using Radeonator LE I could push it to 310/570 without
artifacts using 3dmark 2001SE as the indicator. 3dMark seemed to produce
more artifacts than my other games. the nature test which produced the
most artifacts seemed to gradually develop visual flaws during the test
leading me to believe that it was a cooling problem. The ram chips felt
cool to the touch even after a lengthy BF 1942 session. After trying
faster GPU speeds I found that i could push the GPU to about 310 while
the memory was held back at 285 or 570DDR

skeptical ramsinks could improve my setup i chopped up an old aluminum
slot-1 heatsink into 4 pieces and lazily lapped it. This was mainly done
to get it clean and not particularly to get it flat (which is isn't). I
put some white thermal paste on the chips and used a credit card to
spread it and scrape it off. No paste was applied to the sinks. I took
the springs off of 4 plastic clothes pins and used them to clamp the
sinks on the card. I loaded up radeonator LE, pushed the memory 600 and
BF 1942 runs with no artifacts. Nature however will only tolerate 295
artifact free.

The sinks cooling 4 of the chips now feel warm after use.
I wonder how far i could go had i did a better job of lapping and
applying the paste.





K7S5A
duron 1100
no volt mods.

 

[Paul P Mathews]

They don't.

 

[Joe]

They don't.

On what basis do you make that assumption?

My facts:

1)when overclocked the Nature test begins artifact free, then visual
errors appear after a couple of seconds. Evidence implying heat build up

2)Turning off the fan pointed at the sinks produced visual errors. This
is with the case OFF so the rest of the computer is being cooled.

3) I could up my memory overclock by 10mhz without showing artifacts

Perhaps you're reading too much and taking other peoples asumptions as
fact.

BTW, I've reapplyed the thermal paste properly and can get 299/299
without artifacts, 15mhz more than without the ram sinks.

However beyond 300Mhz I get visual errors right when the test starts,
leading me to believe that heat is no longer the cause of the artifacts.

I do find it interesting however how we can interpret this for system
memory. Obviously they cannot tolerate any errors in which case the use
of ram sinks would only be useful in pushing the memory to their max and
in certain circumstances could lead to instability and unpredictability.