Discussion in 'Overclocking & Hardware' started by Ratzz, Aug 23, 2017.
Yep bad stuff all right.... why it's best left under the IHS
Direct die LM means I can maintain a 120W draw on a setup designed for 75W GPUs, and a 75W draw on the CPU on its setup designed for 45W/55W, so, swings and roundabouts...
Thanks for the videos, Ratzz.
I have a bunch of 3770Ks crunching numbers for https://www.worldcommunitygrid.org/and I've intended to delid them for quite a while. I delidded a spare CPU using an old-fashioned double-edged razor blade, and managed to not delid any fingers at the same time. It was an easier job than I thought, actually, but when a Rokit tool came up 4 sale her, I bought it. I have some Coollaboratory Pro and Ultra that I bought from Germany, but it's unused to date because I've wanted more info on applying the metal and re-lidding. Your vids go part of the way.
Q1: What supports the IHS when you put it back in place - the die pressing on the underside of the IHS, or the rim of the IHS resting on the CPU PC board? I'd want it to be the die, to minimise the thickness of the TIM layer. Have you ever checked, eg with garden-variety soft non-metal TIM used as a marker?
Q2: The last stage in your vids was you leaving the CPU on the clipboard for the silicone to set a bit. What's the problem with putting the CPU straight into a motherboard? If you did that, the motherboard clamp would ensure that the IHS was in the right spot. If you also put on the heatsink, it would squeeze out any excess metal if the IHS bowed upwards in the middle when the motherboard socket clamp was locked down.
Q3: Re the silicone that you use - does it give off acetic acid (vinegar smell)? If so, it's corrosive and electrically conductive and is not recommended around electricity. I have a tube of Selleys Wet Area silicone that doesn't smell and I would probably use that. It's not as strong as the other stuff and it's opaque white, so any if you leave any excess it would be very obvious.
A guy in a shop said that there's no way he'd use a metal TIM because they are very toxic, so I did some DuckDucking. (Start here: https://en.wikipedia.org/wiki/Liquid_metal --> http://www.indium.com/TIM/solutions/liquidmetal.php ). It seems that unlike mercury, the metals used in these TIMs are not highly toxic, though I'd prefer not to test that. If left in the open they probably give off metal vapour, as Mercury does. Resealing the IHS right around, rather than just at the corners, may stop most of the vapour loss.
Also, these TIMs are usually not pure gallium, but are eutectic mixtures of several metals.
Unlike you gamers and LN2 jockeys, I'm not after 5GHz+ overclocks unless they come easily on low Vcore (1.25 or less, preferably under 1.23) and low temps. Running with all cores at 100% load 24/7, CPUs die or if you're lucky degrade if they run hot or with high Vcore. I prefer high-end air cooling to water ... have 3 x lapped TRUE Rev.-nothings with LGA11xx/1366 adapters, 1 x TRUE Ven X (needs the wet'n'dry), 1 x Megahalem in use, plus a Cryorig and an NH-15 waiting for use. At the moment, speeds are limited by temps, even though all Vcores are well under 1.2V. Electricity co$t is a killer - rock on die-shrinks!
Hi mate, thanks for the input.
Q1. The IHS is supported by the PCB. There is less than .5mm separating the die from the IHS, as evidenced by the hardened TIM that comes ex factory.
Q2. When the clamp is tightened onto the CPU on the board, the IHS will slide across the top of the die. The motherboard clamp makes contact with the IHS and pushes it forward unless the silicone prevents it from doing so. I've tried preventing this movement by pushing against the IHS with a finger, but this doesn't work. Setting the silicone a bit first makes the clamp slide across the IHS rather than pushing the IHS across the die.
Q3. Yes, its smelly clear silicone. I'm not aware of any conductivity or corrosion issues, I've been doing this a long time and had no problems. I still have a number of delidded chips that were done a few years ago and still work perfectly. If you aren't confident with this, most people use black Sikaflex for that 'original' look, I've just never got around to buying any
I wouldn't try to eat Gallium based liquid metal, but IMHO it isn't dangerous to use for this purpose at all.
The use of liquid metal has multiple effects as you would see in the screenshots. Not only does it reduce temperatures, but in doing so it allows higher clocks at lower voltages. I don't game much, nor have I ever used LN2. My interests are getting maximum results from ambient cooling, and I only do it for fun. I enjoy the competitive nature of getting the most out of a chip while still retaining its daily use, which of course you can't do with LN2. I've considered going with a phase change setup before, but the simplicity of water or high end air suits me better. Phase change is too noisy for starters !!
I've never lapped a chip or cooler in my life.
I'm currently using Ryzen, so I have no need to delid on soldered AMD chips. I'll get back to Intel and delidding again though one day when I am bored
I hope this is of use to you, and I'm glad you found my thread interesting !!
Also.. #post 4669 just made my day, thanks !! Not sure whether to be honoured or nervous !!
Thanks for yet more useful info. Your reply to Q1 means that removing the original black silicone and replacing it with the soft squishy gel that is fresh silicone goo means that the clearance between the die and IHS will be reduced. I think Intel arranged for such a thick layer of TIM (0.5mm you say) to allow the TIM to cope with sideways expansion & contraction due to heating and cooling. In the delidding thread at Xtremesystems it was reported that polymer-based replacement TIMs lost their benefit over Intel's muck after a month or 2. My theory is that thinner layers of polymer TIM can't cope with the shearing movements. Liquid metal probably can, and it's a better conductor in the first place.
And your explanation re allowing the silicone to partly set before installing the CPU will probably save me making some mistakes.
Re Post 4669 - I'm a bit slow ... even BF couldn't see what was funny about your pic for a while. Nick is due to my eyesight problems - keratoconus plus developing cataracts, but I'm not actually blind.
Alert: Blind Freddie is following you
can I use any ole thermal paste or need a specific one, going to delid 8700k. No real need to glue back down eh?
You can use any thermal paste and you will see an improvement of several degrees at least. Use a liquid metal instead, and you will see between 10°-20° at least. Its really that good.
Reglueing the IHS is not necessary, but highly recommended. Liquid metal is electrically conductive, keeping it properly contained is necessary. It also makes your chip just like any other, removable, transportable, and easily handled. The new glue will not be as difficult to remove as the old, so it will not be as difficult to delid it as it was the first time from that point on. UnlesI you use Sikaflex of course, in which case time it will be nearly or as totally bonded as it was before delidding. Again, not really an issue as you can just delid it the same as you did the first time.
I personally would be using liquid metal, and reglueing it. I can see no logical reason for taking any other path.
Thanks metal+glue it is! Any suggestions for both?
I got a 3d printed delidder for a 7700k I was going to try, will have to see how it goes on the 8700k. (If I can find it!)
I use Thermal Grizzly Conductonaut, basically because I can buy it locally. Pretty confident brand wouldn't matter, the other big player is Coollaboratory Liquid Ultra, but not sure where you would buy it.
Whatever you do, don't forget the Liquid Electrical Tape on the small gold contacts on the PCB under the IHS. Get some liquid metal bridging those contacts, power it up, and presto, dead CPU. That being the main thing you need to be careful about with a liquid, electrically conductive TIM. The stuff grips to the die and IHS like a snot on a wall if applied correctly, but you can't be too careful !!! You can get a small tube like the one I use at Jaycar for a few gold coins.
I use ordinary bathroom silicon to glue down the IHS, but if Blind Freddie is correct this might not be ideal. Personally I have chips I delidded 5 years ago and used clear silicon on, never had a problem. Many people use Sikaflex, its probably the closest to 'original' being a similar compound and black.
I only use silicon because it works, I have it handy and I'm too lazy/tight to bother going and buying some SIkaflex. That, and Sikaflex doesn't come in a small tube that I am aware of.
I'd be interested to know how you go with the 3D printed 7700K delidder on your 8700K. I've not done an 8700K yet, I've been playing with Ryzen in recent times, so not much delidding happening here , but I've had a few people inquiring if I would do 8700K's for them. It would be nice to know if my tool is still useful, or if I'll need to get another.
7700 and 8700 have the same lid. So you shouldn’t have any issues.
Think I posted in another thread the Sika product I use. Totally worth resealing to keep everything where it belongs. I only leave my bench chips lose so it’s easy to change pastes for cold. Daily resealed is perfect.
But, you can leave lose if you want. I’d suggest not removing the old silicon if you go this route though. It will help protect the pcb, and holds the lid better when you clamp so it doesn’t slide under pressure.
know the name off top of your head? id rather seal it tbh. Interesting point re: liquid electrical tape ratzz thanks probly saved my ass!
Is there anyone in Canberra who wants to do for me? for a fee
This going to be enough paste?
Cool, thanks. Nobody has been able to answer this for me to date, and I haven't tried an 8700K as yet to find out for myself.
Sikaflex 227. Giving away my trade secrets here lol
I always cover the pads. If you don’t have LET (since it’s like 20-30 a tin from jaycar) you can get away with using just the silicon or some clear nail polish. All work well to protect from shorting.
Make sure you apply the LM to the die and under the lid. Your temps won’t be as good if you only do the die.
When you’ve done as many chips as I have in the past year, it’s easy to see intel never change their ways.
Yep, I've noticed you've done quite a few. Since I started playing with Ryzen, the need has passed, but I've had quite a few people ask me. I've been non-committal since I wasn't sure if my tool was suitable and I haven't got around to buying one of those fancy things...
Any tips on how much thermal paste to use on the die and lid? And how much sikaflex?
I am worried I dont have the tools to do it the best way possible.. Maybe I should just see how it goes without for now and do it later once better equipped.
This kit looks very noob friendly!
If you're using liquid metal, just enough to coat the surface on the IHS and die. If you put either the CPU or IHS on its side, there shouldn't be enough that will cause it to run off the area you've applied it on. Here is one I recently did
I score the IHS with the locations of where the old thermal paste was with a blade before cleaning it away - you won't know where to apply the liquid metal when the IHS is clean otherwise. You can see it in the photo if you look closely.
After I apply liquid metal on the IHS, I tilt it diagonally and suck up the excess metal, from one of the corners, back into the tube with the syringe. The silicone is personal preference, I normally just do a light surface coat all the way around - any excess is going to get pushed off it when you clamp the IHS back down.
The Rockit 88 tool is great - I have done close to 100 delids now with it. That looks like a revision of the one I have (mine has white acetal relidding parts). The re-lid tool is great on the Rockit because it will reseat the IHS exactly where it was before.
Refer to earlier posts in this thread for video of this process as detailed below.
First, apply LET (liquid electrical tape) to any exposed gold contacts you can see on the PCB, under the area which would normally be covered by the IHS. This will give it time to dry while you apply the liquid metal.
Its a minuscule amount of the liquid metal. About a 1.5mm ball of it on each surface.
It will look and behave like a tiny ball of mercury, you'll wonder how the heck you are going to spread it.
You'll use the cotton bud supplied with the liquid metal, and push the stuff around (slightly firmly) onto the surface until you hear it start to make an audible squeaking noise - a bit like rubbing a mark off a piece of glass with metho if that makes sense? If you have failures, or you ever need to do it again, a standard cotton bud will do just as well as the fancy one that comes packaged with the liquid metal.
That's when the magic starts.. the slippery metal will start to stick. Continue to spread and rub it in, squeaking away until you have thoroughly spread it over the entire surface of the die, and it forms a nice rectangular shaped pool on the top of the die. The more you rub it, the better it will stick, and the better it will conduct. It is imperative that you take the time to make sure it sticks. Total time doing this.. about 2 minutes or so. Once you have applied it, assess the pool and make sure you are happy that there is a nice pool covering the entire surface of the die. If you feel it isnt' getting full coverage, you can add a very tiny additional amount of liquid metal.
Then do the same on the corresponding area of the IHS. I usually do a slightly larger area on the IHS, as of course it wouldn't be easy to make it exactly correspond to the die. Since the metal will stick to the surfaces once you've rubbed it in well, the IHS can then be inverted and placed over the die, and the two pools will become one when they meet.
I use a thin cable tie stuck into a cartridge of silicone and simply wipe a moderate smear along the four edges of the IHS. I'd do the same with a cartridge or tube of Sikaflex. That's all it needs, just a smear around the four edges. Too much would be too messy and too thick, not enough will be very obviously not enough. Use your own judgement, its pretty easy to judge.
Once I've smeared the silicon/sikaflex, I then simply drop it into place on the PCB and lightly clamp it for 20 minutes or so, and drop it into the motherboard socket and clamp it down with the CPU clamp, but if you wanted it would be even better if you wait longer before placing it onto the motherboard.
Try to ensure you place the IHS back into the same place it came from, as accurately as possible, before clamping.
It sounds horribly fiddly, but once you are doing it, you'll be surprised how easy it actually is, and it will be a lot easier to judge whether you have done it right than you think. You'll get it right the first time, if it looks right then it pretty much is.
Thanks guys you rock! Also found this which appears to be pretty fool proof! Thoughts on the copper IHS? I love the look and how easy it is to glue on.