Hi,

I've recently been given a load of parts and some of the RAM sticks included have heatsinks on them.
I'm talking about flat, aluminium shields, in to parts which encase virtually the whole stick (not the 'spiky' individual chip cooling type).
So anyway, I have several spare of these heatsinks and wondered if it would be worth fitting them to the RAM I have installed that don't have these currently fitted.

So...

1) Do these type of heatsinks work?
2) Will I need a special adhesive to attach them to the RAM. It looks like they have been fitted using some type of double sided tape.
3) Could I simply use something like superglue around the edges of the aluminium plate and attach them that way.

Thanks as always.

OwenBear.

Oh - These heatsinks might be copper.

Usually these pretty much just clip on, or use a friction fit. The stuff you found on the sink and chips that they were attached to was probably a thermal transfer paste or tape that helps draw the heat from the chips to the sink where it can be removed by normal heat convection and radiation. You can get this stuff (the thermal paste and tape) from most online electronics stores such as buy.com, newegg, et al. Anyway, don't superglue the sinks to the chips or memory stick! It won't help the thermal transfer characteristics at all, and will make it pretty much impossible to remove!

Anyway, here is a link to one such product on buy.com: http://www.buy.com/prod/startech-com-metal-oxide-thermal-cpu-paste-compound-tube-0-0060-c-w/q/loc/101/10311463.html

Heatsinks are only of real value when using performance RAM. Unless you are going to overvolt and overclock the RAM, there is no real need for heatsinks.
Double sided thermal tape is what is needed if you do decide to install them though.

I've had regular RAM sticks overheat on me where use of a heat sink would probably help reduce that. I fixed the problem by moving the sticks to get better airflow, but a sink would probably have worked as well, or better.

How did you determine they overheated?

I run Scientific Linux 6 (a clone of RHEL 6) and it has full support for system thermal sensor monitoring (lm_sensors), so I can monitor both RAM and CPU temperatures. Under load one RAM stick would exceed limits, resulting in the OS mapping that RAM out of the system. By moving the sticks I was able to reduce the heat by over 25 degrees centigrade under load - such as when I would run all 8 cores to build a new kernel in 15 minutes rather than 2 hours.

How long have RAM had heat sensors? I was under the impression they had none.

Mine are fully buffered ECC RAM (Crucial brand) and the motherboard is an Intel S5000XVN workstation/server board. Custom built with dual 3GHz E5450 processors, 8GB (4x2GB) RAM, and 2.5TB of disc cost me about $5000USD when I built it 3 1/2 years ago.

I think they are an exception rather than the rule at the moment though.

I think they are an exception rather than the rule at the moment though.

Possibly. Myself, I depend upon my systems for my business (IT consulting), so except for my laptop, all my systems utilize ECC memory, and I continuously monitor them for RAM/CPU usage, temperature, drive utilization, etc, with alarms to notify me when things start to get "wonky", such as memory utilization starts to hit the swap space. Since I run a variety of operating systems in virtual machines (Windows, Solaris, QNX, Linux) and an Android emulator (not all at the same time), this can help me sort out any number of strange issues.

Be a part of the DaniWeb community

We're a friendly, industry-focused community of developers, IT pros, digital marketers, and technology enthusiasts meeting, networking, learning, and sharing knowledge.