water block for ram

Vector86 · 02-12-2004, 11:39 PM

has this been done before?
who makes it?
can somene make me one custom?

Etacovda · 02-13-2004, 05:09 AM

Ive heard stories that ram cooling is next to pointless

|kbn| · 02-13-2004, 06:27 AM

why is it next to pointless? i wouldnt expect my gig of corsair pc3500 to last very long with 3.2v Vdimm like some people have used without proper cooling. currently using the timings and fsb and voltage im using (2-2-2-12, 220fsb, 2.9) i *need* to have a fan resting on the heatspreaders or they will *fail* in memtest86 with quite a few errors. the fan keeps them cold enough to work without errors.

heatspreaders though are different - they are pointless. firstly becuase they dont even make contact with the whole of the ram chips, just the middle. second because there so thin - they dont disapeate heat quickly becuase they lack in surface area, they dont really help ram *when* the ram needs it (when its over-volted), therfore the stock heatspreaders are for two resons, to make it look better and also to hide the chip's details unless some vioding of warrentys is done

cheap easy ram cooling would be done by 15mm copper pipe running along top of your rams, with flattened copper pipe soldered to it making full contact with the chips. there isnt realy anyway to make a mmore compact cooler with water, whilst also using 1/2" pipe flowrates (so it can be ran inline with the rest, no need to run it parrellel).
it would be difficult to cool two or more sticks of ram like this, and doublesided ram sticks would also be far from easy

rundymc · 02-13-2004, 07:53 AM

look here http://overclockers.com/tips1003/index02.asp
apparently, Ji Kuang/ Shamino, didn't get anything out of his chilled ram, but he blamed it on his mobo, and I would agree
chilling the ram does add stability, so it would aid in ocing, especially in terms of latencies
in your case, the ram would benifit from water cooling after reading your experience with it

sorry I can't help you in your search- if it comes to the worse, making your own should be easy, a pal of mine in the States used coolermaster ram heatsinks (copper), filed the fins for a thinner base (since the sinks are the same size as the heat source), made space for a copper top, and epoxied the thing up
he keeps all the blocks in parallel, so flow rate with the tiny aquarium tubing he uses is not up to it, but it works well in his case
I would have put four blocks each in series, and run them parallel

msv · 02-13-2004, 09:17 AM

Hm, can´t access #Rotors homepage. If anyone has done it, I bet #Rotor has.
I´ll check it out.
regards
Mikael S.

bigben2k · 02-13-2004, 01:27 PM

It is pointless.

Of all the waterblocks one can put into a PC, this one is the one that is a complete waste of resources, IMO.

The power output of a stick of RAM is almost negligeable (<10 Watts). Water cooling it will not drop its temperature significantly enough to make any difference, either in the ability to overclock, or to add to the longevity of the electronic component.

That aside...

There's a web outfit out there that sells these, even in silver. More senior members than me will know which one. A search through these Forums might bring it up.

SlaterSpeed · 02-13-2004, 02:42 PM

I can see the point to chilling it using peltier or simlar but theres just no point with normal w/c loop.

My ram runs at more or less ambiant temp anyway.

Maybe ddr2 will be differant?

|kbn| · 02-20-2004, 09:17 PM

its not pointless. i have ram sinks on my 9700. they are alu, and have lots of fins. used passivly they get hot, and definatly limit the rams max o/c as the rams get too hot at faster speeds. i have tested them activly cooled and i get quite a few more mhz from them like that. (infinion 3.3ns on 9700np - gets very hot unlike most other chips.)
also my 2x512mb corsair xms3500c2cp need active cooling (fan blowing on heatspreader) to get 2-2-2-12 at 220fsb @ 2.9v. the warrenty is already vioded (bought second hand) so i will take the heatspreaders off, and w/c (which wont cost anything) meaning i can get rid of the annoying fan....

ive started my 9700 ram w/c. after it will look much like this:
http://server5.uploadit.org/files/kbn2k3-9700np2r1.jpg
ive cut the copper pipe to length, got the elbows and the copepr sheet, its all ready to be soldered together. changes from the above pic are the copper goes all the way along the top - allows for w/c the IC's that would normally be under the heatspreader (which i took off becuase the glue holding it on had got burnt by the heat - this needs some kind of fan even just a small silent one, but id prefer no fans

the sheet that touches the ram is a lot thiner than in the pic (0.7mm - its from flattened 15mm copper pipe...)
hopefully ill get it soldered and the card vmod'ed tomorrow, then i can change the gpu block a bit, and maybe put it back in...

for each 10c lower in temp, a chip should o/c about 10-30mhz more, voltage increases would also help it get further...

#Rotor · 02-20-2004, 09:58 PM

Quote:

Originally Posted by msv

Hm, can´t access #Rotors homepage. If anyone has done it, I bet #Rotor has.
I´ll check it out.
regards
Mikael S.

U Betcha I did it.... but that was long ago.... and indeed, even back then I eventually also came to the same conclusions as the people trying it now...

"it's not really worth the effort...."Don't get me wrong... That ain't no excuse for not doing it...

there you can see, my early efforts was actually very simple, two plates squishing a tube, and clamping the stick of ram, that has chilled liquid going through it.... worked rather well for it's simplistic design....

-J- · 02-21-2004, 01:15 AM

some nice heatsinks should do the trick.

Blackeagle · 02-21-2004, 05:32 PM

I'd suggest this to you.

Install a side blower in your case if you don't have one where it can direct cool room temp air on the RAM.
Next if you have to buy a quality thermometer and attach one or more thermisters to the heat spreader. Get some accurate temps at the highest voltage and O/C you can run at under full load.

Now remove the heat spreaders and repeat the testing @ the same settings, then try for a higher O/C.

At Directron they have some skived fin heat sinks for system RAM. I'd suggest you buy sets of that and attach with thermal epoxy. Again run the same tests.

I think you'll find that just taking the heat spreaders off will drop temp & increase your O/C, may be only a small gain, but I'd bet you'll get a gain.

Adding the heat sinks will also give a boost if you are pushing a pretty high O/C at a good high voltage (3.2+), and the higher the voltage the more it'll help.

There was a thread on this topic a while back at Xtreme Systems. Most, but not all, who ditched the heat spreaders made some gain. And a couple who added heat sinks gained more.

For those after MAX O/Cs a few extra mhz gained on the FSB is well worth the cost of finned heat sinks.

Blackeagle · 02-21-2004, 05:50 PM

Quote:

Originally Posted by bigben2k

It is pointless.

Of all the waterblocks one can put into a PC, this one is the one that is a complete waste of resources, IMO.

The power output of a stick of RAM is almost negligeable (<10 Watts). Water cooling it will not drop its temperature significantly enough to make any difference, either in the ability to overclock, or to add to the longevity of the electronic component.

That aside...

There's a web outfit out there that sells these, even in silver. More senior members than me will know which one. A search through these Forums might bring it up.

For those driving RAM with 3.2 up to 4 volts water cooling wouldn't be a Waste Ben. It would in fact increase the RAM's life (most of all if above 3.5v) at the least, and would most likly aid O/C some.

For those driving really Xtreme O/C's good RAM cooling is needed & a help.

All that said however, most making use of 3.5v or below can do fine with a side blower fan on the RAM & some finned heat sinks added with epoxy thermal glue instead of worthless heat spreaders. But the same can sure be said for water cooling of hard drives & mosfets as well, and I believe you are planning on making use of those aren't ya Ben?

|kbn| · 02-21-2004, 06:29 PM

heatspreaders are useless though... they bearly increase surface area of heat to air, so they wont disipate it much quicker than the naked chips... for normal heatspreaders to be of any use you need airflow on them, otherwise there useless.. they will just store the heat (and add an extra TIM joint...).
the problem is there isnt really room with system ram for heatsinks, usually the ram is too close together. you could use 3 or 4 40mm fans theres not much point, they dont make enough air flow for the niose imo, and they would only cool the inner sides as there not wide enough. i wouldnt use fans + stock heatspreaders for any voltage over 2.9v on my corsair 3500

ive been using a 80mm fan due to lack of 60mm fans and it just gets in the way. ive had to move my cd drive out a lot, and the fan is secured only by 2 peices of 1/0.6 wire tied around cables near it.
all the ram ive had the performance will decrease at 2.8 and 2.9v due to overheating. when it gets to the point where a small fan will aloow it to overclock more, thats the point they need better cooling - more voltage wont help. im not sure how people can run 3.5v through there ram without better cooling! also thats 1v over SSTL-2 stock voltage (JEDEC standard voltage for ddr ram), usuall i dont go much more than 0.4v over stock voltages in fear of electromigration :o
#rotor your design looks interesting, but im confused how the water cools/makes contact with the alu plates?

bigben2k · 02-21-2004, 11:04 PM

Well, you're right aboput that BE! I do plan to go through the marginally beneficial HDD water cooler, and even though the MOSFET cooler seems complicated, and the benefits dubious, I do believe that that's one worth the effort.

But I certainly wouldn't go around telling people about running RAM at more than 2.9v without a fair warning!

That aside, I can see water cooling ram at above 3.5 helping to save the RAM, for sure, but there'll be other support chips that'll be running way outside spec. If I was running RAM above 3.5 (which I'd never do), I'd keep a close eye on the northbridge temp, and the power supply chip(s) to the ram, along with anything else that's related, and running hot.

Could be fun, but too risky for my taste.

02-12-2004, 11:39 PM	#1
Vector86 Cooling Savant Join Date: Jun 2002 Location: KANSAS "it's not just cows"! Posts: 211	water block for ram has this been done before? who makes it? can somene make me one custom?

02-13-2004, 09:17 AM	#5
msv Cooling Savant Join Date: Dec 2002 Location: Sweden Posts: 336	Hm, can´t access #Rotors homepage. If anyone has done it, I bet #Rotor has. I´ll check it out. regards Mikael S. __________________ The only constant factor in all Your failures is You. Catapultam habeo. Nisi pecuniam omnen mihi habis, ad caput tuum saxum immane mittam.

02-13-2004, 01:27 PM	#6
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	It is pointless. Of all the waterblocks one can put into a PC, this one is the one that is a complete waste of resources, IMO. The power output of a stick of RAM is almost negligeable (<10 Watts). Water cooling it will not drop its temperature significantly enough to make any difference, either in the ability to overclock, or to add to the longevity of the electronic component. That aside... There's a web outfit out there that sells these, even in silver. More senior members than me will know which one. A search through these Forums might bring it up. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

02-21-2004, 05:32 PM	#11
Blackeagle Thermophile Join Date: Oct 2002 Location: U.S.A = Michigan Posts: 1,243	I'd suggest this to you. Install a side blower in your case if you don't have one where it can direct cool room temp air on the RAM. Next if you have to buy a quality thermometer and attach one or more thermisters to the heat spreader. Get some accurate temps at the highest voltage and O/C you can run at under full load. Now remove the heat spreaders and repeat the testing @ the same settings, then try for a higher O/C. At Directron they have some skived fin heat sinks for system RAM. I'd suggest you buy sets of that and attach with thermal epoxy. Again run the same tests. I think you'll find that just taking the heat spreaders off will drop temp & increase your O/C, may be only a small gain, but I'd bet you'll get a gain. Adding the heat sinks will also give a boost if you are pushing a pretty high O/C at a good high voltage (3.2+), and the higher the voltage the more it'll help. There was a thread on this topic a while back at Xtreme Systems. Most, but not all, who ditched the heat spreaders made some gain. And a couple who added heat sinks gained more. For those after MAX O/Cs a few extra mhz gained on the FSB is well worth the cost of finned heat sinks. Last edited by Blackeagle; 02-21-2004 at 06:01 PM.

02-21-2004, 11:04 PM	#14
bigben2k Responsible for 2% of all the posts here. Join Date: May 2002 Location: Texas, U.S.A. Posts: 8,302	Well, you're right aboput that BE! I do plan to go through the marginally beneficial HDD water cooler, and even though the MOSFET cooler seems complicated, and the benefits dubious, I do believe that that's one worth the effort. But I certainly wouldn't go around telling people about running RAM at more than 2.9v without a fair warning! That aside, I can see water cooling ram at above 3.5 helping to save the RAM, for sure, but there'll be other support chips that'll be running way outside spec. If I was running RAM above 3.5 (which I'd never do), I'd keep a close eye on the northbridge temp, and the power supply chip(s) to the ram, along with anything else that's related, and running hot. Could be fun, but too risky for my taste. __________________ WBTA (Water Block Testing Alliance) Nordic Hardware WC 101

02-13-2004, 05:09 AM	#2
Etacovda Cooling Savant Join Date: Feb 2004 Location: Dunedin NZ Posts: 735	Ive heard stories that ram cooling is next to pointless

02-13-2004, 06:27 AM	#3
\|kbn\| Cooling Savant Join Date: Feb 2004 Location: W. Sussex, UK Posts: 329	why is it next to pointless? i wouldnt expect my gig of corsair pc3500 to last very long with 3.2v Vdimm like some people have used without proper cooling. currently using the timings and fsb and voltage im using (2-2-2-12, 220fsb, 2.9) i need to have a fan resting on the heatspreaders or they will fail in memtest86 with quite a few errors. the fan keeps them cold enough to work without errors. heatspreaders though are different - they are pointless. firstly becuase they dont even make contact with the whole of the ram chips, just the middle. second because there so thin - they dont disapeate heat quickly becuase they lack in surface area, they dont really help ram when the ram needs it (when its over-volted), therfore the stock heatspreaders are for two resons, to make it look better and also to hide the chip's details unless some vioding of warrentys is done cheap easy ram cooling would be done by 15mm copper pipe running along top of your rams, with flattened copper pipe soldered to it making full contact with the chips. there isnt realy anyway to make a mmore compact cooler with water, whilst also using 1/2" pipe flowrates (so it can be ran inline with the rest, no need to run it parrellel). it would be difficult to cool two or more sticks of ram like this, and doublesided ram sticks would also be far from easy

02-13-2004, 07:53 AM	#4
rundymc Cooling Neophyte Join Date: Feb 2004 Location: Singapore Posts: 67	look here http://overclockers.com/tips1003/index02.asp apparently, Ji Kuang/ Shamino, didn't get anything out of his chilled ram, but he blamed it on his mobo, and I would agree chilling the ram does add stability, so it would aid in ocing, especially in terms of latencies in your case, the ram would benifit from water cooling after reading your experience with it sorry I can't help you in your search- if it comes to the worse, making your own should be easy, a pal of mine in the States used coolermaster ram heatsinks (copper), filed the fins for a thinner base (since the sinks are the same size as the heat source), made space for a copper top, and epoxied the thing up he keeps all the blocks in parallel, so flow rate with the tiny aquarium tubing he uses is not up to it, but it works well in his case I would have put four blocks each in series, and run them parallel

02-13-2004, 02:42 PM	#7
SlaterSpeed Cooling Savant Join Date: Nov 2003 Location: Suffolk, UK Posts: 234	I can see the point to chilling it using peltier or simlar but theres just no point with normal w/c loop. My ram runs at more or less ambiant temp anyway. Maybe ddr2 will be differant?

02-20-2004, 09:17 PM	#8
\|kbn\| Cooling Savant Join Date: Feb 2004 Location: W. Sussex, UK Posts: 329	its not pointless. i have ram sinks on my 9700. they are alu, and have lots of fins. used passivly they get hot, and definatly limit the rams max o/c as the rams get too hot at faster speeds. i have tested them activly cooled and i get quite a few more mhz from them like that. (infinion 3.3ns on 9700np - gets very hot unlike most other chips.) also my 2x512mb corsair xms3500c2cp need active cooling (fan blowing on heatspreader) to get 2-2-2-12 at 220fsb @ 2.9v. the warrenty is already vioded (bought second hand) so i will take the heatspreaders off, and w/c (which wont cost anything) meaning i can get rid of the annoying fan.... ive started my 9700 ram w/c. after it will look much like this: http://server5.uploadit.org/files/kbn2k3-9700np2r1.jpg ive cut the copper pipe to length, got the elbows and the copepr sheet, its all ready to be soldered together. changes from the above pic are the copper goes all the way along the top - allows for w/c the IC's that would normally be under the heatspreader (which i took off becuase the glue holding it on had got burnt by the heat - this needs some kind of fan even just a small silent one, but id prefer no fans the sheet that touches the ram is a lot thiner than in the pic (0.7mm - its from flattened 15mm copper pipe...) hopefully ill get it soldered and the card vmod'ed tomorrow, then i can change the gpu block a bit, and maybe put it back in... for each 10c lower in temp, a chip should o/c about 10-30mhz more, voltage increases would also help it get further...

02-21-2004, 01:15 AM	#10
-J- Registered User Join Date: Dec 2002 Location: Buenos Aires, Argentina Posts: 217	some nice heatsinks should do the trick.

02-21-2004, 06:29 PM	#13
\|kbn\| Cooling Savant Join Date: Feb 2004 Location: W. Sussex, UK Posts: 329	heatspreaders are useless though... they bearly increase surface area of heat to air, so they wont disipate it much quicker than the naked chips... for normal heatspreaders to be of any use you need airflow on them, otherwise there useless.. they will just store the heat (and add an extra TIM joint...). the problem is there isnt really room with system ram for heatsinks, usually the ram is too close together. you could use 3 or 4 40mm fans theres not much point, they dont make enough air flow for the niose imo, and they would only cool the inner sides as there not wide enough. i wouldnt use fans + stock heatspreaders for any voltage over 2.9v on my corsair 3500 ive been using a 80mm fan due to lack of 60mm fans and it just gets in the way. ive had to move my cd drive out a lot, and the fan is secured only by 2 peices of 1/0.6 wire tied around cables near it. all the ram ive had the performance will decrease at 2.8 and 2.9v due to overheating. when it gets to the point where a small fan will aloow it to overclock more, thats the point they need better cooling - more voltage wont help. im not sure how people can run 3.5v through there ram without better cooling! also thats 1v over SSTL-2 stock voltage (JEDEC standard voltage for ddr ram), usuall i dont go much more than 0.4v over stock voltages in fear of electromigration :o #rotor your design looks interesting, but im confused how the water cools/makes contact with the alu plates?

Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)