Why? Why put thermal insulators on SDR?

All that is going to do is make the RAM run hotter.

If you really need SDR to run cooler than just point a case fan at it.

cyclone3d wrote on 2022-07-30, 06:46:

Why? Why put thermal insulators on SDR?

All that is going to do is make the RAM run hotter.

If you really need SDR to run cooler than just point a case fan at it.

Just a casemodding thing, no practical need for heatsinks. (although i use CL3 certified RAM at CL2, the dimms are barely warm, maybe 30 degrees C)

useless
https://media-www.micron.com/-/media/client/g … aa129b2689e1c8f
128MB SDR dimm uses ~0.5A at 3.3V. Below 2 watts when transferring data, jumps to 3W during refresh.

I always imagined the SDRAM and early DDR ram module heatsinks were just there for the look more than a real need. Considering sometimes the "quality" of thermal pads often used I always wondered if any heat would even transfer to the heatsink. I imagine higher clocks decreased stability before any temperature problem at least in those times.
I tried lately on a SBC ARM board which the temperature went up to even 80/90°C easily and using a cheap heatsink with its own factory thermal pad, the temperature in idle even increased and maybe beginning to be useful only when temperatures were really high. So I imagine the heatsink is not always the solution if depending on variables to consider. That SBC board sure need a heatsink but also a case air flow imho and the best possible thermal paste considering how fast temperature increase or decrease.

The only gen of ram that got warm was DDR3. All of the ramsinks we had were manufacturers trying to differentiate their brand. These GSkill DDR4 sticks are the most hilarious. I took the sinks off a stick last year and it was just taped on with 2-sided tape.

DDR5 could be different though. I don't know much about that.

I always imagined the SDRAM and early DDR ram module heatsinks were just there for the look more than a real need.

DDR Samsung TCCD 500+ Mhz (modules which were sold with these specs btw) - requires heatsink.
DDR Winbond (BH/CH) on 3.3v - requires heatsink and sometimes active cooling.

The only gen of ram that got warm was DDR3. All of the ramsinks we had were manufacturers trying to differentiate their brand.

Before DDR5 arrival, enthusiast DDR2 and FB-DIMMs were the hottest RAM modules, surpassing even infamous RDRAM.

psychofox wrote on 2022-07-30, 06:36:

Does DDR heatsinks fit to SDRAM dimms? They probably do fit fine but i just in case ask since i have no DDR dimms lying around here at the moment for comparison.

i am looking at these:
https://www.aliexpress.com/item/4000309965844 … 286289312%21sea

So these are Kingston HyperX Fury ramsinks? Are these salvaged off of dead Kingston RAM or something or are these brand new? (no pun intended 🤣)
I didn't know people were actually reselling these.

I have a 512mb stick that has a blue heat spreader.

Came like that.

Try it? They will probably fit

Can you find rgb ones? That would be cool.

RGB won't work.

smtkr wrote on 2022-07-30, 21:58:

The only gen of ram that got warm was DDR3. All of the ramsinks we had were manufacturers trying to differentiate their brand. These GSkill DDR4 sticks are the most hilarious. I took the sinks off a stick last year and it was just taped on with 2-sided tape.

DDR5 could be different though. I don't know much about that.

You forget RDRAM. The heatsinks on that are there for a reason.

The Serpent Rider wrote on 2022-07-30, 22:18:

I always imagined the SDRAM and early DDR ram module heatsinks were just there for the look more than a real need.

DDR Samsung TCCD 500+ Mhz (modules which were sold with these specs btw) - requires heatsink.
DDR Winbond (BH/CH) on 3.3v - requires heatsink and sometimes active cooling.

I was thinking to slower late SDRAMs and early DDR modules which I imagine did not require heatsinks by factory and even the ones that had some (in the early times) didn't exactly warm up that much, not the modules and not the heatsinks imho. Considering how many of those early DDR ram modules perfectly worked without until later or higher end modules probably with higher requirements, clocks and even voltages (I think some +0,1v by factory compared to the usual voltage) which I imagine in a closed atx high end config case suggested the heatsinks usage, I'm not sure if by the ram datasheet or the manufacturer choice.

smtkr wrote on 2022-07-30, 21:58:

The only gen of ram that got warm was DDR3. All of the ramsinks we had were manufacturers trying to differentiate their brand. These GSkill DDR4 sticks are the most hilarious. I took the sinks off a stick last year and it was just taped on with 2-sided tape.

DDR5 could be different though. I don't know much about that.

I don't know about modern DDR4 specifications (but I think I've seen many modules without any heatsink like in the past maybe low end but still running at the usual clocks) but in the DDR PC3200@400Mhz times some high end modules had those thin heatsinks with thermal pads I still wonder if they even worked as supposed to. I understand many mainboards having close ram sockets would warm up more but the heatsink also didn't permit the air flow to pass between the DIMM modules which is the other side of such solution. Maybe in modern times they might be really needed like the old RDRAMs or some highest end DDR2 but I imagine only as an highest end product for extreme clocks and higher voltages.

cyclone3d wrote on 2022-07-31, 20:50:

smtkr wrote on 2022-07-30, 21:58:

The only gen of ram that got warm was DDR3. All of the ramsinks we had were manufacturers trying to differentiate their brand. These GSkill DDR4 sticks are the most hilarious. I took the sinks off a stick last year and it was just taped on with 2-sided tape.

DDR5 could be different though. I don't know much about that.

You forget RDRAM. The heatsinks on that are there for a reason.

I was under the impression that the heatspreaders were needed because of the way RDRAM works it could pound individual chips differently.

For OP:
I'm running DDR2 OCZ heatpipe ram "covers" on various DDR and SDR kits.
Purely for the lols =)

386SX wrote on 2022-07-31, 21:39:

in the DDR PC3200@400Mhz times some high end modules had those thin heatsinks with thermal pads I still wonder if they even worked as supposed to. I understand many mainboards having close ram sockets would warm up more but the heatsink also didn't permit the air flow to pass between the DIMM modules

There is an unreasonable and unintuitive amount of effect even a small heatsink has on the ability to radiate out heat. Plus the goal is not having temperature comfortable to human touch, but one below Tjmax of targeted silicon.
https://www.youtube.com/watch?v=8ruFVmxf0zs EEVblog - Electronics Thermal Heatsink Design Tutorial

Apple is really "good" at this. In order to lower noise they cook all electronics close to 80-100C depending on manufacturer provided specification instead of actually efficiently cooling. Thats how you get multiple recalls for fried macbook GPUs or recent M2 heavy throttling.

rasz_pl wrote on 2022-08-01, 08:01:

There is an unreasonable and unintuitive amount of effect even a small heatsink has on the ability to radiate out heat. Plus the […]
Show full quote

386SX wrote on 2022-07-31, 21:39:

in the DDR PC3200@400Mhz times some high end modules had those thin heatsinks with thermal pads I still wonder if they even worked as supposed to. I understand many mainboards having close ram sockets would warm up more but the heatsink also didn't permit the air flow to pass between the DIMM modules

There is an unreasonable and unintuitive amount of effect even a small heatsink has on the ability to radiate out heat. Plus the goal is not having temperature comfortable to human touch, but one below Tjmax of targeted silicon.
https://www.youtube.com/watch?v=8ruFVmxf0zs EEVblog - Electronics Thermal Heatsink Design Tutorial

Apple is really "good" at this. In order to lower noise they cook all electronics close to 80-100C depending on manufacturer provided specification instead of actually efficiently cooling. Thats how you get multiple recalls for fried macbook GPUs or recent M2 heavy throttling.

I also don't think too that the heatsink temperature finger test is much useful beside to understand if it's working when sensors read high values. But I'd wonder how much heat such thick thermal pads can even transfer to those thin heatsinks compared to the same modules without, in a good air flow case scenario (just wondering with those old early DDR DIMM I suppose in some cases it might have not been always a priority).
Because as said looking at those old high end DDR modules running close to each other basically with no space for any air cooling, I wonder they might even warm up more. In the past there were similar choices (like also nowdays RGB leds on components which have not much sense). Of course I'd never remove a factory installed heatsink beside to change the thermal paste/pad but I'd not easily add one if there isn't a real reason.

Like the example of a SBC ARM board. I added an heatsink with a weak thermal factory pad, it didn't solve much the maximun values but once I decided to let the SoC run as supposed without, idle temperatures decreased like -5/10°C. Which make the heatsink a factory choice that should be found in a IC datasheet suggestion imho. 😀

cyclone3d wrote on 2022-07-31, 20:50:

smtkr wrote on 2022-07-30, 21:58:

The only gen of ram that got warm was DDR3. All of the ramsinks we had were manufacturers trying to differentiate their brand. These GSkill DDR4 sticks are the most hilarious. I took the sinks off a stick last year and it was just taped on with 2-sided tape.

DDR5 could be different though. I don't know much about that.

You forget RDRAM. The heatsinks on that are there for a reason.

RDRAM doesn't exist in my metaverse 😁

I was getting 140Mhz out of 8ns PC-100 with just aluminum plate spreaders on back in the day. That was before they were cherrypicking and rebinning for 7.5ns "barely" PC-133 though and PC-133 had 7ns. (i.e. the pool of 8ns parts had parts that might otherwise have been binned 7.5ns.) But yeah, depended a lot on airflow in your case, particularly bad if you had a PIII or tually overhanging the RAM off a slotket and the spreaders might have been collecting heat rather than getting rid of it.

Source of generic, module sized (as opposed to high rise with frills and flounces... oh, ineffective notchy fins) spreaders might be ECC server RAM, which goes by the bucket load for cheap.