RAM heatsinks

[Shrek]

Does RAM really need heatsinks?

All the RAM I have never seems to get very warm.

 

[theFOoL]

To me no. Even OC I don't feel one needs them. Now for Storage MVME I'd say yes *Kinda

 

[27MaD]

RAMs running @high frequencies & low timings / heavily overclocked might slightly benefit from heatsinks, they're mostly bling bling that's it.

 

[tabascosauz]

Most lower end heatspreaders are absolute crap and have no surface area (see Viper Steels), probably would do about the same thermally without them. Honestly I'm not even aware of any DDR4 heatspreader that actually looks like it has impressive surface area. Trident Z could be okay if it didn't have that plastic piece/stupid RGB diffuser on top - I removed the plastic piece on my E-die Tridents but for the RGB/Neo/Royals you can't just take off the diffuser. I took the heatspreaders off my 4133CL19 Viper Steels for clearance, they've been running naked fine at 4200 17-18-18 1.45V for a while now.

It's really only a specific subset of DDR4 that might benefit. Within DDR4 there are only a few ICs that can safely handle 1.5V or more > within those there are only a few that can do tight timings (B-die) or high freq (Rev.B, DJR) > within those there are a few that are temp sensitive and destabilize at x temp so actually need the cooling. DJR comes with crazy factory XMP up to 1.65V.

Mostly it's airflow that matters much more than what heatspreader is on there. Heatspreaders with poor airflow are useless, but not vice versa. I'm not so sure about dual rank sticks, a heatsink might knock off a few degrees (again, not just any old dual rank 3200CL16, but dual rank B-die/Rev.E/(Rev.B?)/(DJR?) @ 1.5V+).

But then again, people love parroting the idea that watercooling RAM is useless, but a properly put-together waterblock setup can actually be beneficial for a B-die daily system pushing 1.55-1.6V (say, 4000CL14 on Zen3, 4600CL16 on CML etc). Especially with a 200W+ GPU in the system dumping heat onto the DIMMs.

 

[caroline!]

Yes, depending on what kind of RAM is it, and what kind of heatsink. My old DDR2 memory was OC'd from 800 to 1066 MT/s, that thing used to get HOT so I did what any ghetto modder would: glue some heatsinks on it.
Those modules are really sturdy, ran at 2.3V for years and are still alive, case cooling was simply a box fan standing next to the barebones case blowing as much air as possible, it was ugly and noisy but managed to play a ton of games with it, even Half-Life 2 which was quite the feat, I even added MODS to it when I got my hands on an used 40GB drive which was literally as twice as big as the original 20GB it came with (prebuilt). Those were the years.

 

[Braegnok]

+1,.. properly put-together waterblock setup can actually be beneficial for a B-die system pushing 1.55-1.6V.

I started water cooling my memory a few years ago, as B-die modules quickly become unstable when the RAM IC's go above 52c,..

I'm running Hynix IC's now on my daily driver @ 5866MHz 1.725V,.. with custom made copper heatspreaders & watercooling.

 

[ir_cow]

@Braegnok!!!!!!!!!! I was going to ask you how you manage that clock but I see you are rocking that Apex MB

My answer is yes for most things that is 1.35v or above. I can tell you right now that if you don't have a heatspreader on for 1.45v+ it will create errors. Dual rank is the worse. Requires active cooling (aka fan directly on it). Even though ram is rated for 85c (JEDEC Specs). If you pass 45c on Samsung-B die either the system crashes or you will get crazy errors.

 

[RyzenMaster.sys]

Most ram won't need heatsinks. Usually it's just for looks. However, higher voltage ram or overclocked ram may begin to want at least some airflow or a heatsink.

 

[tabascosauz]

@Braegnok!!!!!!!!!!

My answer is yes for most things that is 1.35v or above. I can tell you right now that if you don't have a heatspreader on for 1.45v+ it will create errors. Dual rank is the worse. Requires active cooling (aka fan directly on it). Even though ram is rated for 85c (JEDEC Specs). If you pass 45c on Samsung-B die either the system crashes or you will get crazy errors.

Can never get a clear answer on Bdie temp sensitivity. Is it 45 or 50? Or is 45 just a sub-140ns tRFC or 1.55v+ thing? I have my single rank and dual rank B-die down to 133ns and 140ns without destabilizing at 1.45-1.53v during long repeated TM5/LinX/HCI sessions, as long it stays below 50 - hence why I treated 50 as the magic number.

+1,.. properly put-together waterblock setup can actually be beneficial for a B-die system pushing 1.55-1.6V.

I started water cooling my memory a few years ago, as B-die modules quickly become unstable when the RAM IC's go above 52c,..

I'm running Hynix IC's now on my daily driver @ 5866MHz 1.725V,.. with custom made copper heatspreaders & watercooling.

View attachment 222168

Sweet DJR setup. Have you noticed any temp sensitivity on Hynix? Heard somewhere that DJR is, but could never confirm and CJR doesn't seem to be

 

[ir_cow]

@tabascosauz I think it is Kit dependent. I usually have problems between 43-48c for Samsung-B Die. But if it is stable for you with no errors during memory stress testing, that is a win!
I have no idea for Hynix DJR temp. Just that you can abuse the crap out of the voltage. 1.6v is not a problem. Considering I will do hours of stress testing at 1.6v, I think DJR is the bomb. I do put a direct fan on them though, which is why I haven't ever looked at the temps.

 

[eidairaman1]

Does RAM really need heatsinks?

All the RAM I have never seems to get very warm.

It does help, ive felt ram get plenty hot.

+1,.. properly put-together waterblock setup can actually be beneficial for a B-die system pushing 1.55-1.6V.

I started water cooling my memory a few years ago, as B-die modules quickly become unstable when the RAM IC's go above 52c,..

I'm running Hynix IC's now on my daily driver @ 5866MHz 1.725V,.. with custom made copper heatspreaders & watercooling.

View attachment 222168

View attachment 222169

Alphacool wb there.

Been wanting to WC the ram lol.

 

[Mussels]

Does RAM really need heatsinks?

All the RAM I have never seems to get very warm.

At stock voltages, no.
Even overclocked, often no.

 

[Toothless]

Some server ram sure does need it. Been almost burned by ECC when it was in my SR-2 rig.

 

[delshay]

CAUTION: Strong Recommendation By Me.

When inserting any DDR Memory module I recommend inserting the memory module by pressing utilizing the edge of the PCB "if visible". I don't have enough evidence to support why, but it's been look at here why so many are going faulty. It's to do with stress on the solder balls & solder pads why they are lifting so easy when worked on. Below is a example where I recommend to apply pressure when inserting the memory. Personal I have stop inserting memory module via the heatsink. I have been doing this for some time now but have never posted this.

QUESTION: Does anyone know the best way to take memory modules apart that have been glued? Heating it up does not work very well.


Strong Recommendation: Insert memory module where marked with a red circle (press here). Not on the heatsink like most of us do.

 

[ir_cow]

QUESTION: Does anyone know the best way to take memory modules apart that have been glued? Heating it up does not work very well.

Hair dryer. 30 seconds - go back and forth. Try to peal one side. If it fights you, keep heating it up. Try the other side as well.

Most kits only need one round to come apart. It is also easy to bend the heatsink, so be aware that is likely to happen.

I've only do 30 or so kits so far. Im sure someome who does it all day for custom WC builds might have a better and easier way.

 

[delshay]

Hair dryer. 30 seconds - go back and forth. Try to peal one side. If it fights you, keep heating it up. Try the other side as well.

Most kits only need one round to come apart. It is also easy to bend the heatsink, so be aware that is likely to happen.

I've only do 30 or so kits so far. Im sure someome who does it all day for custom WC builds might have a better and easier way.

Sorry, should have pointed out the heatsink is directly glued to the chips. Even with a very strong heat source it very difficult to get it apart.

 

[eidairaman1]

CAUTION: Strong Recommendation By Me.

When inserting any DDR Memory module I recommend inserting the memory module by pressing utilizing the edge of the PCB "if visible". I don't have enough evidence to support why, but it's been look at here why so many are going faulty. It's to do with stress on the solder balls & solder pads why they are lifting so easy when worked on. Below is a example where I recommend to apply pressure when inserting the memory. Personal I have stop inserting memory module via the heatsink. I have been doing this for some time now but have never posted this.

QUESTION: Does anyone know the best way to take memory modules apart that have been glued? Heating it up does not work very well.


Strong Recommendation: Insert memory module where marked with a red circle (press here). Not on the heatsink like most of us do.

If not enough contact area press on the ends of the HS.

Never had a ram failure due to pressing the IHS, only ESD, DOA- defect

 

[ir_cow]

Sorry, should have pointed out the heatsink is directly glued to the chips. Even with a very strong heat source it very difficult to get it apart.

Are you sure its really glue and not super sticky tape? I would still heat it up, but I guess use a knife to cut/peal as you go. It will ruin the heatspeaders if its really glued, but that is the only option really.

 

[delshay]

If not enough contact area press on the ends of the HS.

Never had a ram failure due to pressing the IHS, only ESD, DOA- defect

No. Press down/insert the ram stick by pushing down on the edge of the PCB (red marking) if exposed. Some heatsink covers this so you may have no choice but use the heatsink to insert in socket.

Are you sure its really glue and not super sticky tape? I would still heat it up, but I guess use a knife to cut/peal as you go. It will ruin the heatspeaders if its really glued, but that is the only option really.

It's real glue. Large collection of DDR2 to repair/restore..

 

[Shrek]

I seem to recall a video where cooling in liquid nitrogen helped with removing the heat sinks; not very practical, I know.

(1) Removing DDR Memory Heatspreaders with GamersNexus - Bing video

At 7:40 looks like tape to me

 

[delshay]

I seem to recall a video where cooling in liquid nitrogen helped with removing the heat sinks; not very practical, I know.

(1) Removing DDR Memory Heatspreaders with GamersNexus - Bing video

Will try overnight freezing it.

 

[Mussels]

Will try overnight freezing it.

airtight bags, no condensation!

 

[Shrek]

I wonder if soaking overnight in isopropyl alcohol would help.

 

[maxfly]

Will try overnight freezing it.

You don't need to keep it overnight. An hour or so will freeze it up. Be quick but gentle when prying the spreaders off or you'll have to refreeze.
GL!

 

[delshay]

You don't need to keep it overnight. An hour or so will freeze it up. Be quick but gentle when prying the spreaders off or you'll have to refreeze.
GL!

Don't have time for monitoring. It's in a plastic bag in the freezer, will check on it in the morning.