MC14 ramsinks

[rragazan]

Should i apply them on the 8800 gtx with thermal pads or arctic silver ? how do I fill the EK multi-option reservoir 250 ? What are the best fans for the Thermochill 120.3 ?

[Levish]

unless you are thinking of arctic silver epoxy the pads are fine

[CedricFP]

1. Just use the tape that comes already on the sinks. Push hard, though.
2. How do you fill it? You can unscrew the top; take out the plug in the top and stick a funnel in there; put a barb in the top and lead it to a fillport in your case.
3. The general consensus around here are Yate Loons for maximum silence/CFM efficiency.

[rragazan]

Artic Silver 5, isn't it more effective ? I don't want the ramsinks to fall. Which Yate Loons, particulary ? The 120x38 are supported ? What about fan shrouds ? On Thermochill there are only for the HE series. How about the PA ?

[welshtom]

Arctic silver isnt adhesive, so the ramsinks will just fall off. Thats why you need Arctic silver epoxy instead.

The HE shrouds will fit the PA rads :). Yateloon D12-SL is a nice fan

[rragazan]

ah, i get it, you're right ! So, is artic silver adhesive better than the standard pads ?? There are 2 types : silver adhesive and alumina adhesive. which is best ? What are the best CFM for the PA120.3 rad ?

[welshtom]

I dont really know which is best, I've always used thermal pads though :)

[IanY]

Let me present a contrarian view.

Arctic Silver has been reported by some (and not me) to me harmful and damaging to video card memory. There have been reports (not me) of AS5 killing video cards. Some people use Arctic Ceramique for video cards because its not electrically conductive (I would think that Alumina is conductive). I use Arctic Silver on the gpu itself and nothing on the memory.

If you are going to use adhesive on your memory, think carefully before you do so. I am not saying don't do it! Unless you somehow dilute the bonding strength, the heatsinks will be on there for good. If you somehow need to RMA the card, or sell the card, or if you somehow change your mind and want to use a full covered block, or if you want to give the card to someone else and want to reinstall the stock cooler, you will be pretty much out of luck. Just something to keep in mind. I myself water cool the cards for good and pass them on to others in my family, so I don't care. If they are on there for good, then all the better. But you need to re-evaluate carefully.

I have found the adhesive pads on the MC14 a royal PITA to stick. I use a different padded product that I can link you to if you want. I don't like superglue on my ram lol :)

Yate Loons are very effective for their price. They are dirt cheap and work well. They *are not* the best quality fans in the market. You can spend up to five times more for a fan and get a 5% improvement. So, buy Yate Loons if you are on a budget and want the best for a given budget. Do not buy Yate Loons if you want the very best fans in the market and have money to spend/waste. I myself do not use Yate Loons because I find them of dubious manufacturing quality, but that's just me. Call me a snob, call me stupid, whatever. I have Yate Loons at home and I don't use them.

EK reservoir. I don't have one, so I cannot comment. Logic says that filling with a funnel up top or a fillport would be better than unscrewing the top and worrying about spillage.

[rragazan]

Then which are the best fans ? What about silentx fans ? I will then use the pads that come with them.

[IanY]

The best fans? I wouldn't want to make this into a flamewar thread. Everyone has different perceptions of what is the best fan. Let me put it this way. A Yate Loon costs $5 or $6. You can easily pay $25 or even $30 for a fan. Would you pay $6 or $25 ? If your answer is "I don't care how much it costs. I have the money and I will buy the very best" then I will start talking. :)

[welshtom]

another thing is do you want silence or not. The "best" fan wont be the most silent, in fact, it will be incredibly noisy.

The yates do provide a good silence/performance ratio

[big poppa pump]

The pads that come with the MCW14 ramsinks have got to be the :banana::banana::banana::banana:tiest in the market. I tried sticking them onto my 800GTX and even when I applied pressure and held them for almost 5 minutes, a small nudge would knock them off.

I scraped off the pads that came with MCW14, sanded down the heatsinks with 400/600/1000/2000 and applied Artic Ceramique to the heatsinks and the memory. Used a sewing needle to apply super glue to the ramsinks, which just gives enough strength to hold the ramsinks to the memory, but still prevents the raminks from bieng knocked off with a small nudge or bump. At the same time this method also allows me to remove the raminks without causing any damage to the memory by just slightly twisting the ramsink and popping them off.

For the VRM's, I ended up getting some 3M heat transfer adhesive tape (comes with a lot of the passive northbridge chipset coolers) and sticking the ramsinks to a single 3M strip and then using the same superglue technique above to make them stick to the VRM's without them bieng knocked off.

For the EK reservoir, I use a fillport technique. Used to have a T-line before I installed the reservoir. Removed the T-line and use the fillport to connect to the top of the reservoir with the bottom ones being connected as inlet/outlet. I love this setup as it took me less than 30 minutes to fill and bleed.

BTW......The Yates kick A$$. Best fans you can get as far as price/silence/performance is concerned. There are more expensive solutions available in the market, but none of them justify spending more $$ to get a 3%-5% increase in performance.

[stormshadow]

1. Just use the tape that comes already on the sinks. Push hard, though.
2. How do you fill it? You can unscrew the top; take out the plug in the top and stick a funnel in there; put a barb in the top and lead it to a fillport in your case.
3. The general consensus around here are Yate Loons for maximum silence/CFM efficiency.
qft

but those stickers on the ramsinks are kinda :banana::banana::banana::banana:ty... so yea - as mentioned before... press hard and for a while.

i got fed up after they started flicking off one at a time so i don't use them anymore...

[ramenchef]

I think philly_boy said that he uses a blow dryer to heat the pads up some so they are more sticky, and then put the sinks on and it should stick much better.

[Budwise]

ive said over and over in the same threads just like this. AS5 works great to hold on ramsinks. But a dab in the middle of the ramsink, push it on firm yet flat, and let it kinda suck onto the ram chip. Once a few runs are done and the ram chips have heated up a few times and cooled down the AS5 will cure and form a nice bond although very easy to pop em off when you want.

[sanhacker]

AS5 works great to hold on ramsinks. But a dab in the middle of the ramsink, push it on firm yet flat, and let it kinda suck onto the ram chip. Once a few runs are done and the ram chips have heated up a few times and cooled down the AS5 will cure and form a nice bond although very easy to pop em off when you want.

Can I assume you remove the existing TIM on the ramsink before putting the dab in the middle? Or, do you just leave it on? Thanks.

[ramenchef]

I don't think that works as AS5 is not an adhesive. Anyone else want to confirm?

[Allsorts]

Can I assume you remove the existing TIM on the ramsink before putting the dab in the middle? Or, do you just leave it on? Thanks.

Yes, remove the existing TIM since this could give you remounting + heat transfer problems. I use ArctiClean but I guess you can use alcohol.


I don't think that works as AS5 is not an adhesive. Anyone else want to confirm?

Correct, AS5 is not an adhesive but I think he was refering to the suction holding the heatsinks on. I prefer to use the bare minimum TIM so there's no chance that suction will hold my heatsinks on!

[Budwise]

yea, with a little dab it kinda sucks onto the ram chip. It doesnt take much AS5 to create enough suction to hold it on and once it cures they wont fall off at all. I dont remove the heatpads, i just clean em really well and use the AS5. Ive used this method for over a year now and it works great. If you want to go the epoxy/AS Adhesive route go ahead, but give this method a try if you want as it is alot easier imo.

[Mr. Peanut]

I don't think that works as AS5 is not an adhesive. Anyone else want to confirm?AS5 is an adhesive if you buy their adhesive product. We are talking about the AS5 epoxy, not the thermal paste.

[Mr. Peanut]

By the way, remember that thermal tape works fine (although I still use epoxy) as long as you allow it time to form a good bond. Most tapes take at least 30 minutes to form a tight hold.

[Allsorts]

OK just to clarify things:

AS5 is 'Arctic Silver 5 Thermal Compound' (http://www.arcticsilver.com/as5.htm) and is just a TIM which is non-adhesive (although Budwise manages it :lol:).

'Arctic Silver Thermal Adhesive' (http://www.arcticsilver.com/arctic_silver_thermal_adhesive.htm) or Arctic Silver epoxy (note it doesn't have '5' in the name) is a thermally conductive permanent adhesive.

If you want to permanently attach the ramsinks and then forget about it, then the Arctic Silver Thermal Adhesive will suit you fine. If you want to keep your options open in the future then using AS5 or Arctic Silver Ceramique (http://www.arcticsilver.com/ceramique.htm) with tiny dabs of glue at the corners seems to be less stressful than trying to get thermal pads to stick.

[serialk11r]

I stuck aluminum chunks (from my case) onto the chips of my hard drive for fun with as5. It had pretty good "suction" as you could call it. The only problem is, if you move them around a little, they start getting AS5 all over the place which isn't good. I got as5 all over the bottom of my hard drive and I have no short circuits or problems for some reason...

[Snyxxx]

I have found by testing and experience that a 75%/25% mix of AS Epoxy/AS5 holds just right and yet allows you to remove them in the future without ripping the chip out of the circuit board. 50%/50% dilutes the epoxy too much and a bump will knock the sink off.

I also use the Alumina Epoxy/ Alumina thermal paste (same mix ratio) for RAM chips just in case of the contacts and such.

I tried unsuccessfully to use thermal pads. They start out fine, but reaching your hand in to unplug something is when I normally bump them and they fall off. I am glad others have done well with them , just not for me though.

Key for me is the dilution of the epoxy with the similar product.

[Sparky]

I keep seeing it said that arctic silver is conductive so I want to set something straight here ;) It is NOT conductive, says so on the website and has been proven before. What it is though is it IS slightly capacitive, so it is suggested to not get it on chip pins and such. However, this is just as a precaution, odds of it doing any damage are very small but since they are there arctic silver is just covering their butts :)

I have verified this in my own system, I got arctic silver across the top of my GPU next to the core, making contact with those little chips there. I cleaned them off but there is still a residue, no problems however. Also have gotten it on the top of my naked opty and on a few other CPUs and never a problem. (Yes I can be messy sometimes :p:) I do clean it off but the residue stays so I just leave it sit there.

Don't want to sound harsh or offend anyone, I'm just pointing something out that has become one of my pet peeves :p:

:toast:

[ramenchef]

I think the issue lies when the silver bridges two pins and creates a short.

[sanhacker]

I think the issue lies when the silver bridges two pins and creates a short.

I think the point was that it is non-conductive. Therefore, no short.
However, it is capacitive. i.e. Capacitor in effect.

This could lead to intermittent issues if one gets a bit sloppy on application.

[Visioneer]

So what is an alternative to the stock pads now, I experienced the same problem, they fall right off and epoxy is unfortunately permanent and you cant sell the card anymore.

[ramenchef]

Hmm, cuz my friend fried his gfx card when he didn't properly apply the AS5 a while back.

[Snyxxx]

Visioneer,

Did you read my post on Page 1? Dilute the epoxy to make a weaker bond. It will come off later.

[SPL15]

It's been proven over and over and over in the electronics industry that thermal adhesive pads are only slightly better than no thermal interface at all. The pads dont fill the little gaps very well no matter hwo sticky they are. The company I work for has done lots of tests with thermal adhesive tape's and they are only marginally better to lots worse than running dry with no interface. Tape would be nice since it would reduce our manufacturing costs by several hundred thousand, but it doesnt perform near as well as epoxy based adhesives.

Thermal tape only works mediocre if there is still a clamping force. If your relying on the thermal tape for adhesive purposes, it aint doin much good for thermal transfer.

Do a simple google search on this subject and you'll find the info.

The only tape that performs halfway decent is phase change material (PCM). but it's adhesive qualities are non existant when operated at the temperature it is supposed to work at (around 60C). Below the phase change temp, it conducts heat about as good as plastic. If PCM interfaces never reach the phase change temp, it cannot spread out and fill the voids and works like crap.

Applying pressure to thermal epoxy while curing leaves a very thin interface that fills voids very well and transfers heat almost as good as mediocre quality thermal grease applied correctly (which is pretty good considering there is no contact pressure.)

Basically, for good performance, use epoxy. If you need to remove the ramsinks in the future, try anything besides thermal adhseive tapes.

There is a reason why adhesive tapes are not widely used without clamping pressure. It doesnt perform well at all.

For the TMPS chip on the 8800 cards, use somthing better than a MC14 ramsink even with epoxy. This chip dissipates around 15 watts which is more than a mc14 is sized to dissipate under optimal thermal interface conditions. I would say that the mc14 is good for about 5 to maybe 10 watts.

EDIT: The TMPS chip is that small processor looking one near the front of the card. It has been mistakenly called the RAMDAC in the past.

[skindogg]

What would you recommend, maybe epoxy a zalman NB47J on that chip? I just happen to be putting on a MCW60 block tomorrow on my 8800gtx and got every thing lined up. I also got one of these lying around, I'll have to see if it'll fit.

[sanhacker]

For the TMPS chip on the 8800 cards, use somthing better than a MC14 ramsink even with epoxy. This chip dissipates around 15 watts which is more than a mc14 is sized to dissipate under optimal thermal interface conditions. I would say that the mc14 is good for about 5 to maybe 10 watts.

EDIT: The TMPS chip is that small processor looking one near the front of the card. It has been mistakenly called the RAMDAC in the past.

Hmmm. Sounds like a good argument for a full cover block.
Or, are there alternatives to the MC14 for the TMPS ???

[ramenchef]

I believe swiftech has a ramsink kit designed specifically for for G80 cards.

http://www.petrastechshop.com/swmcsmccokit.html

Here it is.

[sanhacker]

I believe swiftech has a ramsink kit designed specifically for for G80 cards.

http://www.petrastechshop.com/swmcsmccokit.html

Here it is.

Yes, but doesn't that kit include an MC14 for the TMPS chip in question ???

[ramenchef]

Ahh, you're right. I would think an MC14 would be good enough though if swiftech uses it in their kit designed specifically for the 8800s, but I guess not.

[JargonGR]

Well I finally managed to stick the MC14s on my 2 8800GTxs but yes it SUCKS! so bad and I am worried on how long they will stick there. I ofc soaked them in Alcohol and removed the stock pads. I used the 3M adhesive tape and some pressure for 20 secs each to make them stick.

However if you ask me this is not any great cooling for the ram also the cards will not do anything above 680Mhz core and 2100 memory although the temps are about 38C on load )PA120.3 dedicated waterlopp and Apogess 1U on the GPUs.


I will try to improvise and maybe make a pressure plate or something for them using the holes on the PCB.

[ramenchef]

You also need to realise that 8800gtx will not perform much better on water than on air as they are mostly volt limited in overclocking.

[Philly_Boy]

Ramenchef is right. My clocks on stock air (after replacing the thermal pads with T411 thermal tape) were like 623/1018. Temps were like 51c idle and 65c loaded. In my dedicated water loop I've managed 653/1070 but not consistantly. The temps are like 35C idle and 39-40c loaded (looping 3D Mark 06 all night). When my G80 water block review is done these cards will get volt modded and we'll see what they can really do.

And yes, I do heat my MC14's with a decent hair dryer for 40-45 secs then they attach a bit easier. You don't have to keep insane pressure for more than 30-40 secs. First I clean them twice with Artic 1 and Artic 2 solutions followed my high purity isopropyl alcohol. If I get a particularly hard one to attach, I scrape off the stock pad with a razor blade (carefully) and replace it with a square of T411 tape. Heat the sink and it should attach. I've even had a scant few extremely stubborn chips that I had to lay down a square of T411 on the memory chip then a dab of AS epoxy them the ram sink.

Using a laser thermometer, my memory modules never get hotter than 28-30c even during hard benching. Even the TMPS chip stays a fairly cool 33-34c. I try to get a few fans blowing directly on the ram sinks. That helps a lot. Fans take about 5c off the the temps of the ram sinks alone without fans.

[SPL15]

Ramenchef is right. My clocks on stock air (after replacing the thermal pads with T411 thermal tape) were like 623/1018. Temps were like 51c idle and 65c loaded. In my dedicated water loop I've managed 653/1070 but not consistantly. The temps are like 35C idle and 39-40c loaded (looping 3D Mark 06 all night). When my G80 water block review is done these cards will get volt modded and we'll see what they can really do.

And yes, I do heat my MC14's with a decent hair dryer for 40-45 secs then they attach a bit easier. You don't have to keep insane pressure for more than 30-40 secs. First I clean them twice with Artic 1 and Artic 2 solutions followed my high purity isopropyl alcohol. If I get a particularly hard one to attach, I scrape off the stock pad with a razor blade (carefully) and replace it with a square of T411 tape. Heat the sink and it should attach. I've even had a scant few extremely stubborn chips that I had to lay down a square of T411 on the memory chip then a dab of AS epoxy them the ram sink.

Using a laser thermometer, my memory modules never get hotter than 28-30c even during hard benching. Even the TMPS chip stays a fairly cool 33-34c. I try to get a few fans blowing directly on the ram sinks. That helps a lot. Fans take about 5c off the the temps of the ram sinks alone without fans.


Infrared thermometers (aka laser thermometers) are not that accurate position wise or temperature wise for that matter. I've got an expensive well respected Fluke infrared temp meter and the reading spot is about 2 inches below what the spot light says it is at when about 1 foot away and it is about 15C above what it is actually measuring, proven by an expensive calibrated temp probe I use at work for mission critical work.

Also, reading the temperature of a heatsink is hardly an indication of die temperatures which is what really matters.

I'll give you this exagerated non real world example to clearly state my point:

Say you place a piece of dry regular cotton between a CPU and heatsink for the thermal interface. Cotton as you already know has poor thermal conductance properties and would be a very poor thermal interface. With cotton as a thermal interface, the heatisnk temperature will measure very low while the CPU temp will be very high because the cotton thermal interface is not transfering heat to the heatsink very efficiently. If you read the temperature with an accurate infrared sensor, you would be reading the relatively cool heatsink temperature and NOT the burning hot CPU temp. Since the thermal transfer of cotton is VERY low, the temperature difference between the CPU and heatsink is VERY high. the CPU is hotter than heck while the heatsink is running cool due to poor thermal transfer of the cotton. If you base temperature on heatsink temperature with a poor thermal interface your going to get an overly optimistic result.

Chomerics, a respected thermal interface manufacturer, clearly admits in its own private testing that the thermal adhesive tape they manufacture is only marginally better than a dry air interface. Lots of specialized professional electronics journals have stated the fact that thermal adhesive tape is only slightly better than no inerface at all.

[Philly_Boy]

Infrared thermometers (aka laser thermometers) are not that accurate position wise or temperature wise for that matter. I've got an expensive well respected Fluke infrared temp meter and the reading spot is about 2 inches below what the spot light says it is at when about 1 foot away and it is about 15C above what it is actually measuring, proven by an expensive calibrated temp probe I use at work for mission critical work.

Also, reading the temperature of a heatsink is hardly an indication of die temperatures which is what really matters.

I'll give you this exagerated non real world example to clearly state my point:

Say you place a piece of dry regular cotton between a CPU and heatsink for the thermal interface. Cotton as you already know has poor thermal conductance properties and would be a very poor thermal interface. With cotton as a thermal interface, the heatisnk temperature will measure very low while the CPU temp will be very high because the cotton thermal interface is not transfering heat to the heatsink very efficiently. If you read the temperature with an accurate infrared sensor, you would be reading the relatively cool heatsink temperature and NOT the burning hot CPU temp. Since the thermal transfer of cotton is VERY low, the temperature difference between the CPU and heatsink is VERY high. the CPU is hotter than heck while the heatsink is running cool due to poor thermal transfer of the cotton. If you base temperature on heatsink temperature with a poor thermal interface your going to get an overly optimistic result.

Chomerics, a respected thermal interface manufacturer, clearly admits in its own private testing that the thermal adhesive tape they manufacture is only marginally better than a dry air interface. Lots of specialized professional electronics journals have stated the fact that thermal adhesive tape is only slightly better than no inerface at all.

What do you suggest we use as a thermal interface between the mem modules and the ram sinks? You seem knowledgeable and able to eloquently point out the shortcomings in our methodology, yet you don't offer any solutions.

When I reassemble my test bed (waiting for my PSU to come back from being modded) I will measure a mem module with temp probe of my UEI DT200 and compare it the temp the laser thermometer gives.

My laser thermometer is a 12:1 unit and I get as close as possible when I measure - like 1-2 inches away from what I am measuring so the spot is no more than 1/6th of and inch wide. The only thing that throws my laser thermometer off is a highly reflective surface as the IR bounces off the shiny material and gives an innaccurate reading. When I measure my water temps with my DT 200 and the IR thermometer they are usually within 0.1c-0.2c of each other.

Most folks rely on some software based solution for measuring temps which is usually very innacurate. We do the best we can with the gear available to us. As long as we are using the same methodology to measure with our readings will be accurate for the system we are testing.

[Serra]

+1 vote for all posters who suggested the use of AS5 (or ceramique) and a dab or two of superglue, applied via sewing needle.

Cheap, easy to do, holds tightly, provides at least *among* the best possible temps (maybe ~.1 difference between this and any "better" method, aside from perhaps using liquid metal... that'd be kind of neat to see...), and *will* come off easily with a razer blade.

[SPL15]

What do you suggest we use as a thermal interface between the mem modules and the ram sinks? You seem knowledgeable and able to eloquently point out the shortcomings in our methodology, yet you don't offer any solutions.

When I reassemble my test bed (waiting for my PSU to come back from being modded) I will measure a mem module with temp probe of my UEI DT200 and compare it the temp the laser thermometer gives.

My laser thermometer is a 12:1 unit and I get as close as possible when I measure - like 1-2 inches away from what I am measuring so the spot is no more than 1/6th of and inch wide. The only thing that throws my laser thermometer off is a highly reflective surface as the IR bounces off the shiny material and gives an innaccurate reading. When I measure my water temps with my DT 200 and the IR thermometer they are usually within 0.1c-0.2c of each other.

Most folks rely on some software based solution for measuring temps which is usually very innacurate. We do the best we can with the gear available to us. As long as we are using the same methodology to measure with our readings will be accurate for the system we are testing.


Thermal epoxies like arctic silver adhesive give pretty good transfer when applied well. BUT the big downside is that it is ususlly permanent. On smooth surfaces like a mosfet or CPU, you can usually just freeze spray the heatsink and break the connection, but with rough surfaces like BGA memory IC, it's stuck pretty good.

There are many other thermal adhesive/epoxies that give similar results as Arctic Silver adhesive, but I personally think (with no measured proof) think AS5 epoxy is the best.

Interface thickness is the main issue with epoxies and I usually clamp down the heatsink with C clamps with moderate pressure while it is curing.

Thermal interface tape and similar products are only good for ease of use, less mess, and to lower manufacturing costs. For performance, it isnt very good unless there is a substantial gap between the heat producing surface and heatsink. But usually in these cases, they are using fiberglass reinforced mat type materials or elastomer gap fillers, but still they dont perform very well... Just good enough to let the thing work safely.

With adequate air flow, the memory on the 8800GTX could run all day long perfectly safe with no heatsik attatched at stock volts, but if we can get the temperature lower then we can get the clocks higher at the stock volts. If we use the absolute best interface we can find, then we maximize our chances for better clocks and protect against overheating when volt modding.

I highly recommend using a larger heatsink than the MC14 on the TMPS IC. My MC14 on that IC runs about 65C with decent airflow above it. If the heatsink runs that warm with airflow then the actual die must be around 70 to 80C which I say is a little too warm for 24/7 use. Yes it is designed to run hot, but anything running that hot all the time wont last too long.

As is common knowledge, the thinner the interface the better, higher clamping pressures give the best results. Anything you can do to maximize both of these will give you the best results, even if it is regular old silcone based thermal grease. cheap Silicone grease performs miles better than any thermal tape.

[HaxR3]

Personally.

Corn Dowing 340 Silicone Based Compund.

Medium size "blob" on ram chip, then push down to create a "suction hold" that keeps it stuck till the compound hardens slightly.

Works great for me :D

[ramenchef]

Could someone explain this "needle" application of superglue? Do you just dip the pin in superglue and slide it on the edges of the ramsinks?

[Eastcoasthandle]

Read this thread (http://www.xtremesystems.org/forums/showthread.php?t=125377&page=2) someone was able to get it working on your card.

[Philly_Boy]

Thermal epoxies like arctic silver adhesive give pretty good transfer when applied well. BUT the big downside is that it is ususlly permanent. On smooth surfaces like a mosfet or CPU, you can usually just freeze spray the heatsink and break the connection, but with rough surfaces like BGA memory IC, it's stuck pretty good.

There are many other thermal adhesive/epoxies that give similar results as Arctic Silver adhesive, but I personally think (with no measured proof) think AS5 epoxy is the best.

Interface thickness is the main issue with epoxies and I usually clamp down the heatsink with C clamps with moderate pressure while it is curing.

Thermal interface tape and similar products are only good for ease of use, less mess, and to lower manufacturing costs. For performance, it isnt very good unless there is a substantial gap between the heat producing surface and heatsink. But usually in these cases, they are using fiberglass reinforced mat type materials or elastomer gap fillers, but still they dont perform very well... Just good enough to let the thing work safely.

With adequate air flow, the memory on the 8800GTX could run all day long perfectly safe with no heatsik attatched at stock volts, but if we can get the temperature lower then we can get the clocks higher at the stock volts. If we use the absolute best interface we can find, then we maximize our chances for better clocks and protect against overheating when volt modding.

I highly recommend using a larger heatsink than the MC14 on the TMPS IC. My MC14 on that IC runs about 65C with decent airflow above it. If the heatsink runs that warm with airflow then the actual die must be around 70 to 80C which I say is a little too warm for 24/7 use. Yes it is designed to run hot, but anything running that hot all the time wont last too long.

As is common knowledge, the thinner the interface the better, higher clamping pressures give the best results. Anything you can do to maximize both of these will give you the best results, even if it is regular old silcone based thermal grease. cheap Silicone grease performs miles better than any thermal tape.
I had a thought today while driving. Why not lap the bottom of the MC14's (obviously remove the white adhesive patch first) with 220 or 300 grit until the bottom is smooth and flat. Then apply a square or T411 tape and clamp to the mem module. The next time I have to rip a card down, I am going to do just that. Compare the temps of the non lapped card's MC 14's to the ones that are lapped. I am also going to experiment with a larger copper heat sink on the TMPS chip. Stay tuned. . .

[big poppa pump]

I had a thought today while driving. Why not lap the bottom of the MC14's (obviously remove the white adhesive patch first) with 220 or 300 grit until the bottom is smooth and flat. Then apply a square or T411 tape and clamp to the mem module. The next time I have to rip a card down, I am going to do just that. Compare the temps of the non lapped card's MC 14's to the ones that are lapped. I am also going to experiment with a larger copper heat sink on the TMPS chip. Stay tuned. . .

Been there, done that! Read my post in this thread.

[Philly_Boy]

Been there, done that! Read my post in this thread.
How fast is the lapping process per ram sink?

Does the spuer glue you apply with the sewing needle leave any residue on the memory module?

Do you apply the glue just in the middle of the ram sink or at the corners or both?

I am in the end stages of my G80 water block review so I'll need to apply the ram sinks once more then take them off then apply them once more permantly. I don't want to accumulate a lot of superglue residue by having to do this multiple times.

[IanY]

The question is how to remove the ramsinks after superglue is applied, even in tiny amounts in the corners. Won't there be nasty blobs of residue that would be impossible to clean off the ram ?

I did the superglue thing in the old days of old 3DFX Voodoo cards and used huge cpu heatsinks on the video cards. Those were like impossible to take off without freezing first, even with tiny bits of superglue. And after the heatsink was off, it was impossible to scrape off the residue without nasty acetone nail polish remover.

[big poppa pump]

How fast is the lapping process per ram sink?

On an average it takes about anywhere from 15-20 minutes per ramsink. I used 400/600/1000/2000 to smoothen down my ramsink completely.


Does the spuer glue you apply with the sewing needle leave any residue on the memory module?

No, it doesn't. I applied Ceramique on the ramsinks as well as the memory modules and then used the tip of the sewing needle to just dab a very very tiny amount of superglue on the ramsinks itself. The amount of superglue is so miniscule that even a small twist of the ramsink will pop it out. At the same time it is also strong enough to prevent the ramsink from falling off once the card is installed. Most of them even stayed on after I somehow accidentally hit them with my tubing when installing it. However extreme force may knock them off and its back to square one.

The good part of the this method is that there is absolutely no residue left on either the memory modules or the ramsink. The glue gets mixed up well with the ceramique and makes a good bond. The best is that there is no risk of ripping off the memory modules from the card when removing the ramsinks. All it takes is a quick twist and the ramsink pops off.


Do you apply the glue just in the middle of the ram sink or at the corners or both?

Either way is fine. I prefer all four corners since this gives a better hold and bonding.


I am in the end stages of my G80 water block review so I'll need to apply the ram sinks once more then take them off then apply them once more permantly. I don't want to accumulate a lot of superglue residue by having to do this multiple times.

With the amount of superglue that needs to be applied, you shudn't even be worried about any residue. Plus as I said earlier, apply the AS5 or ceramique first to the memory modules/ramsinks and then apply the superglue. This will prevent any kind of residue formation.

Again to all, you dont need to apply excessive amounts of superglue. All you need is to wet the tip of the needle and put that small drop on each corner.

[skindogg]

I did manage to fit an old zalman NB47J on the TPMS IC with MC14s on the ram and MC21s on the mosfets along with the MCW60. Just epoxied with AS thermal epoxy. I was a bit surprised it stayed on considering the small surface area of it. I seriously doubt any tape, even T411 would hold it on there. I just put a couple nickels on each heatsink and 4 quarters on the zalman while curing.

I know i'm not extreme, but I added this to my corsair nautilus, which was only cooling my E6400@3.2ghz. Temps didn't go up at all, and now my 8800GTX is running at ~51-52C@600/900. Time to push it up.

On a side note, I got that G80 cooling kit from Petra with the new MC21 heatsinks. Petra rocks, but everyone in here knew that already. Thanks for the pen!

[Philly_Boy]

How fast is the lapping process per ram sink?

On an average it takes about anywhere from 15-20 minutes per ramsink. I used 400/600/1000/2000 to smoothen down my ramsink completely.


Does the spuer glue you apply with the sewing needle leave any residue on the memory module?

No, it doesn't. I applied Ceramique on the ramsinks as well as the memory modules and then used the tip of the sewing needle to just dab a very very tiny amount of superglue on the ramsinks itself. The amount of superglue is so miniscule that even a small twist of the ramsink will pop it out. At the same time it is also strong enough to prevent the ramsink from falling off once the card is installed. Most of them even stayed on after I somehow accidentally hit them with my tubing when installing it. However extreme force may knock them off and its back to square one.

The good part of the this method is that there is absolutely no residue left on either the memory modules or the ramsink. The glue gets mixed up well with the ceramique and makes a good bond. The best is that there is no risk of ripping off the memory modules from the card when removing the ramsinks. All it takes is a quick twist and the ramsink pops off.


Do you apply the glue just in the middle of the ram sink or at the corners or both?

Either way is fine. I prefer all four corners since this gives a better hold and bonding.


I am in the end stages of my G80 water block review so I'll need to apply the ram sinks once more then take them off then apply them once more permantly. I don't want to accumulate a lot of superglue residue by having to do this multiple times.

With the amount of superglue that needs to be applied, you shudn't even be worried about any residue. Plus as I said earlier, apply the AS5 or ceramique first to the memory modules/ramsinks and then apply the superglue. This will prevent any kind of residue formation.

Again to all, you dont need to apply excessive amounts of superglue. All you need is to wet the tip of the needle and put that small drop on each corner.

Thanks for the advice. I'll use this method as I have to apply new MC14's to my 8800GTX's. I will remeasure the temps when I am done this and report my findings here.:toast:

IanY - Does the acetone easily remove the super glue residue from the memory modules? What's your method? I think I have to break out the acetone for some residue on a few memory modules.

*shakes his head wistfully* Sheesh, the paths we take in the name of getting the last from our rigs. . .

[IanY]

I am generally concerned whenever I take acetone to any electronics. What I do is dab the nail polish remover on a make up pad, which I am sure your lady friend has. Those tend to leave a tiny bit of lint, but are better than cotton ball fuzz, which does wonders for anti-static precautions. heh

Dab the acetone onto the residue and let it sit for 30 seconds. Gently use a razor blade or *gentle* and sharp box cutter to scrape off the residue. A blunt box cutter would introduce a danger of taking out the whole memory chip lol

That nasty stuff usually comes off with the first try, or else repeat process. Make sure that the acetone is removed completely from the memory before it dries up or it'll look nasty.

Alternately, any pharmacy sells alcohol swab pads and/or medical swabs for removing sticky stuff (dunno what its called.. nasty stuff like the glue from band aids left on skin).

[Philly_Boy]

I am generally concerned whenever I take acetone to any electronics. What I do is dab the nail polish remover on a make up pad, which I am sure your lady friend has. Those tend to leave a tiny bit of lint, but are better than cotton ball fuzz, which does wonders for anti-static precautions. heh

Dab the acetone onto the residue and let it sit for 30 seconds. Gently use a razor blade or *gentle* and sharp box cutter to scrape off the residue. A blunt box cutter would introduce a danger of taking out the whole memory chip lol

That nasty stuff usually comes off with the first try, or else repeat process. Make sure that the acetone is removed completely from the memory before it dries up or it'll look nasty.

Alternately, any pharmacy sells alcohol swab pads and/or medical swabs for removing sticky stuff (dunno what its called.. nasty stuff like the glue from band aids left on skin).
Thanks. . .I think I have one card that has a bit of residue on two chips. I think the gf has those foam make-up aplicators. I'll try those.

[phelan1777]

I used much the same method, though I only used glue on opposite corners, 2 per sink and it holds well.

Ceramique and Super Glue.

[serialk11r]

Will acetone really do bad stuff to electronics? In the DI/LN2 section the sticky says to put nail polish all over the mobo!!!