Pulling heatspreaders off DDR2 RAM

[Psi]

I Pulled the heatspreaders off some failed DDR2 ram to see what brand chips they were using, and all of the BGA chips came off with it.
It didn't take much force either.

[PeterG]

Made by Dodgy Brothers Corp by the looks of it....

Regards

[Rerouter]

welcome to the world of 1 way construction, where its cheaper to use a glue that isnt designed to pry apart at a later date, even if that isnt all that bad a design feature (metal object loose in a computer can be very, very bad,)

still thanks for the image, it confirms for me that bugger all has changed about ram since the 1980's only higher speed considerations have started to matter, (normally dont get to see a bare ram pcb)

[Monkeh]

A lot of those solder balls look like they didn't bond properly, if at all.

Try warming the heatspreader next time. Usually it softens the adhesive.

[SeanB]

Some probably were meant for a larger capacity BGA chip, and using a common board reduces the cost, and perhaps they were for a different pinout alternate part bas well.

A good thing is that at least the heatsink had good thermal transfer to the chips..........

[Monkeh]

Some probably were meant for a larger capacity BGA chip, and using a common board reduces the cost, and perhaps they were for a different pinout alternate part bas well.

A good thing is that at least the heatsink had good thermal transfer to the chips..........

Not that they need it, unless they're running over specified voltage.

[T4P]

Some probably were meant for a larger capacity BGA chip, and using a common board reduces the cost, and perhaps they were for a different pinout alternate part bas well.

A good thing is that at least the heatsink had good thermal transfer to the chips..........

Not that they need it, unless they're running over specified voltage.

Or really hot running chips

[SeanB]

You mean those batches that passed test, but failed the quiescent current test in production? Probably were self soldering without the heatsink on.

[Psi]

Yeah, quite a few of the BGA joints came apart with the balls still on the PCB.
So they were crap joints to start with.

The brand is SUPER TALENT if anyone wants to know.

[amyk]

Maybe that was why it failed --- some of the balls were already loose.

[Psi]

Maybe that was why it failed --- some of the balls were already loose.

very likely

[entereev]

Btw, heatspreaders for ram are for decoration purposes

[Psi]

Btw, heatspreaders for ram are for decoration purposes

Yeah, unless you have expensive/fast ram and are running it at significantly higher voltage than stock.
Then they do help to get rid of excess heat.

In this case yeah, the ram was just cheap stuff and the heatspreaders were just for show.
They may in fact have contributed to the failure of this ram because they were only attached by the glue strip to the BGA chips.
So pushing the ram into a motherboard is pushing sideways on the chips and may have cracked some of the joints.

[T4P]

Btw, heatspreaders for ram are for decoration purposes

Yeah, unless you have expensive/fast ram and are running it at significantly higher voltage than stock.
Then they do help to get rid of excess heat.

In this case yeah, the ram was just cheap stuff and the heatspreaders were just for show.
They may in fact have contributed to the failure of this ram because they were only attached by the glue strip to the BGA chips.
So pushing the ram into a motherboard is pushing sideways on the chips and may have cracked some of the joints.

You know man ... the heatspreaders are for "marketing" purposes in this case

[Psi]

That's what "just for show" means

[metalphreak]

Ch. 0 DIMM 0:  +75.0°C
Ch. 0 DIMM 1:  +73.5°C 
Ch. 1 DIMM 0:  +82.5°C 
Ch. 1 DIMM 1:  +69.5°C 
Ch. 2 DIMM 0:  +82.5°C 
Ch. 2 DIMM 1:  +70.5°C 
Ch. 3 DIMM 0:  +83.5°C
Ch. 3 DIMM 1:  +70.0°C

That's fully buffered ECC DDR2 server ram (FBDIMMs). Massive heatsinks on them with forced airflow, and they still get that hot. I measured the power increase when I added four more modules.... ~35W 

[T4P]

Ch. 0 DIMM 0:  +75.0°C
Ch. 0 DIMM 1:  +73.5°C 
Ch. 1 DIMM 0:  +82.5°C 
Ch. 1 DIMM 1:  +69.5°C 
Ch. 2 DIMM 0:  +82.5°C 
Ch. 2 DIMM 1:  +70.5°C 
Ch. 3 DIMM 0:  +83.5°C
Ch. 3 DIMM 1:  +70.0°C

That's fully buffered ECC DDR2 server ram (FBDIMMs). Massive heatsinks on them with forced airflow, and they still get that hot. I measured the power increase when I added four more modules.... ~35W 

Imagine them without heatsinks
But yeah, because DDRIII runs on a lower voltage and smaller architecture it is cooler ... nah they do crazy things with the frequency
But heck, that is a shitload of power wasted ... it's like your CPU's quiescent current!

[Rerouter]

the buffering and eec must be doing something, standar 800Mhz ddr2 ram only sucks down approx 1.2-1.9W per chip max, with ddr3 being 0.5- 0.8W for 1600,

[metalphreak]

Yeah they have that advanced memory buffer (AMB) chip on them that uses heaps of power. Was all the rage with Intel dual and quad cpu (not cores) boards back in the DDR2 days.

Hurts a bit when you accidentally put your hand on them when working on a machine that's just been turned off