delidding the 7740X

[batboy]

I have the water loop back together and I'm leak testing. I have an older laptop and this HDD seems so loud and slow after getting used to that Samsung 960 Pro.

Neb, I tried the vice method, but when I started applying force, I chickened out.

I did the razor blade IHS removal method to a couple P-4 way back when (think they were the first with the cap).

 

[Nebulous]

Depending if the PCB is thin and warps under pressure, I wouldn't do the vise method. The Haswell/Refresh's PCB was thick so it was pretty easy. But the PCB for the Skylake/Kaby Lake's are a little thin so using the vise method isn't recommended. Those Delid tools are great to have. I hope they have a redesign top or bottom for the existing tool so ppl will be able to delid the Skylake-X/Kaby Lake X chips without buying a whole new tool.

Also there were horror stories of ppl clamping down too much force on waterblocks with certain chips and warping the pcb enough to kill the chips.

 

[batboy]

Yep, the Kaby Lake-X looks big and feels heavy until you remove the IHS. the PCB/core part is light and flimsy.

I got a new personal best in PCmark10, that moved me up to #3 globally. I'm only 40 points from taking first place.

I switched out the water block and put new hoses on, plus cleaned the radiator and flushed everything out. The XSPC Raystorm block looks great and goes well with my system. It's a good block too, but really pretty much equal to my Swiftech Apogee GTZ. I don't have room to upgrade my 2X120 rad to a triple, so I plan to add a single radiator to go with the double.

Ok, I need to do some research on what to do with this CPU now that I delidded it. I got the IHS off late last night and then worried that I killed the CPU. That's why I hurried and put it back together. I keep hearing about this liquid metal. Is that the thing to do?

 

[JLK03F150]

You should get the desktop image and run Cinebench R15. I'm guessing you could earn a few points for the Benching Team.

http://www.overclockers.com/forums/showthread.php/784399-Marathon-Season-VI-September-Cinebench-R15

 

[batboy]

Ok, back to delidding.

I was so focused on removing the IHS, that I didn't have a good plan for what to do once it was off. A casual glance at info seems like the preferred route is clean it up, put liquid metal (insert fav brand name) onto core and on the inside middle of the IHS. The you have options of resealing it or putting the IHS on and letting the socket hold it in place or going naked (leaving the core unprotected).

But, liquid metal is electrically conductive and especially the Skylakes, some of those CPUs have circuitry on the PCB near the die. Most people make no mention of it, I saw some say coat the surrounding area with clear nail polish first. Sounds like liquid metal has risks.

What, if anything, should I do to obtain max high performance?

[JLK, I downloaded Cinebench, but it was the wrong version. I'll try again later when I get time.]

 

[mackerel]

Liquid metal certainly has its risks, but it also is the highest performance interface. How badly do you want it? You could play safe in the short term, stick with conventional compounds until you decide.

I've only applied liquid metal to one CPU so far, and I was so scared of over-doing it, I didn't put enough on 1st time, hotter than ever. Put some more on, to give visible wetness, and temps dropped as expected, beating anything I saw with conventional TIM. Now to see how it does in the longer term...

 

[||Console||]

I have done a few chips in the past using the razor , but with some of the chips with caps (or whatever they are ) I would rather use the tool .

Alot of ppl use the nail polish trick . 10C with as5 seams normal what most ppl get , then another 5-10 with liquid metal . One thing to remember with Liquid metal you will have to reapply more often as the heating and cooling of the ihs will Pump it out over time .

 

[batboy]

I tried thinking of a permanent solution, but the only thing that came to mind was Arctic Silver epoxy. I used this to attach heatsinks to RAM chips, clock generators, MOSFETS, etc. (and still have some). That would be totally permanent and perhaps drastic. On the other hand, I bet it would work.

Right now, this thick layer of AS5 I applied to the die has me back to where I was and then some. I got a new personal best today that should move me into the #2 spot globally for PCmark10 (only 10 points behind #1).

 

[Mr.Scott]

Ice water your rads and you'll have that #1 spot.

 

[||Console||]

Ice water your rads and you'll have that #1 spot.

+1
Have done and works good =)

 

[batboy]

Well, it just so happens I have a new single rad on the way right now. I plan to mount it in the front panel at the bottom of the case. If I leave the hoses a little long, that ice water trick is very doable with the side panel off (and it would be right before the waterblock). My loop: res > pump > single rad > CPU block > double rad > res.

 

[Sentential]

Ok, back to delidding.

I was so focused on removing the IHS, that I didn't have a good plan for what to do once it was off. A casual glance at info seems like the preferred route is clean it up, put liquid metal (insert fav brand name) onto core and on the inside middle of the IHS. The you have options of resealing it or putting the IHS on and letting the socket hold it in place or going naked (leaving the core unprotected).

But, liquid metal is electrically conductive and especially the Skylakes, some of those CPUs have circuitry on the PCB near the die. Most people make no mention of it, I saw some say coat the surrounding area with clear nail polish first. Sounds like liquid metal has risks.

What, if anything, should I do to obtain max high performance?

[JLK, I downloaded Cinebench, but it was the wrong version. I'll try again later when I get time.]

Yes liquid metal is dangerous and it takes balls of steel to use. I’ve dedicated myself to only using LM and it’s nerve wracking everytime. So far I’ve done 2 laptops a GPU and a CPU and I’m starting to trust myself with it more. The trick is to use electrical tape, specifically Super33 and seal around the die so the LM has no where to go and worst case sits in the moat you made between the resisters and die.

Where LM makes the most difference is it’s ability to heat soak not so much super low temps. For example the laptop I’m using has a 10C* delta from idle to load. On the Vega I just did it takes almost 10 minutes to go from 26C* at 300rpm to 65C* at 1900rpm on the stock cooler. Hell even at +50% power on Vega I can hold 65C* at 3000rpm (75%) most reviewers said they couldn’t hold 80 at 100%.

That’s the beauty of LM is that when you get a load spike literally nothing happens and it takes forever to transfer heat into the block

 

[batboy]

In these articles, I'm seeing what can only be described as a mess when you remove a sink/block after it's had liquid metal on it. Apparently, it's a pain to cleanup. There are claims of it damaging water blocks and heatsinks (don't use on aluminum). But, in this situation it's used under the IHS, so it should never come in contact with anything else. Doesn't this stuff have a low melting point?

EDIT: Ok, I get it, I hope others reading this get it too. This technique of delidding and using liquid metal is an extreme measure for maximum performance... like what I'm doing, trying to break a world record. So kids, don't try this at home. There are definitely risks.

 

[Sentential]

In these articles, I'm seeing what can only be described as a mess when you remove a sink/block after it's had liquid metal on it. Apparently, it's a pain to cleanup. There are claims of it damaging water blocks and heatsinks (don't use on aluminum). But, in this situation it's used under the IHS, so it should never come in contact with anything else. Doesn't this stuff have a low melting point?

That is correct it is hard as hell to get off which again is why you use tape because if it spills you can remove it without cleaning. When you install liquid metal it needs to be treated as a one time permanent application like epoxy. LM never expires and never needs to be reapplied and in my case I plan on never removing the stock coolers on my Vega56 or the laptops.

Yes it has a low melting point and at room temperature it is more semi-solid than liquid

 

[batboy]

Here's a chart I ran across out there in cyber space. Wish I'd seen this before I buttoned the 7740X back up.



AS5, like I used, consistently outperformed the Intel TIM by about 3 degrees on average in this test. The liquid metal shows even better results. Arctic Silver thermal paste also seems to improve a little bit more after an initial break in period (we used to call it "burn-in"). I used to think that was a myth, but over the years I have personally witnessed the phenomena on several occasions. For now, I think that I'll leave the AS5 under the IHS alone.. I can always change it later... or sooner if the high temps return.

That said, the two most recommended liquid metal seems to be Coollaboratory Liquid Ultra and Thermal Grizzly Conductonaut. Both seems to have a good track record of impressing reviewers I guess.

I looked through my stash of thermal paste and most are generic white stuff (silicon I think). I have 2 partial tubes of AS5 and one partial AS3 (from about 2002 or so). I don't remember where this came from, but I have a new, never been used tube of Thermalright thermal paste that looks like a fancy package containing generic white paste.

If I were to get a bit of liquid metal when I order next time--which is better--Coollaboratory Liquid Ultra and Thermal Grizzly Conductonaut?

 

[Sentential]

Conductonaut by miles. It’s blend is far more stable and has higher metal content. General rule of thumb is that Conductonaut is 20C colder than stock paste. My understanding is that CLU gets a bad rap because it is mostly Gallium in suspension with some other chemicals that flash out with heat and time and what you’re left with looks similar to Intel’s solder since Gallium has a higher melting point. Whereas TRC has mostly Indium content with no suspension materials that flash off which makes it slightly more difficult to work with but it doesn’t dry like CLU does.

It’s difficult to describe the way it behaves so my suggestion as has been written in guides is first use Super33 electrical tape to seal the CPU just in case and second test fire the stuff on a surface other than the CPU so you get a feel for the material before you use it. Remember LM is more or less permenant so make sure you got it right cuz it ain’t coming off. That said once you’re accustomed to using it you won’t go back.

 

[batboy]

I'll add Conductonaut to my next Amazon order. What did we do before Amazon? They just make it way too easy to order stuff any kind of thing.

 

[batboy]

I'm still waiting for official confirmation, but I submitted a score for what I believe is a World Record in the PCmark10 benchmark. Wow, never had a world record before. Had plenty of hardware cups in the past though. This "Old School" computer I built is one fine machine. I currently have won 5 gold hardware cups, 2 silver and 1 bronze. Yippee!

PCmark10 is one long and tough benchmark. It takes about 25 minutes to complete. I moved up to third place two days ago and then nailed down second place yesterday. But, today was like I hit an invisible wall. What's left? Memory. I already have it running at the full rated speed of DDR4-3600 with tight timings. I did get it to boot at DDR4-3733 once, so I played with it for a while and figured out how to get it to run at DDR4-3733. A little more DIMM voltage and a bump in voltage of these two settings: CPU VCCIO Voltage and CPU System Agent Voltage. That really seemed to be the key for stabilizing the RAM.



I might not need to take this 7740X apart again (re-delid). The Arctic Silver 5 is doing good, maybe not as good as liquid metal, but way better than the Intel TIM. Like I figured, the AS5 needed some burn in time to set up. It seems to keep getting better, whereas the Intel TIM just kept getting worse. When I put the IHS back on, I used a super thin layer of adhesive and a slightly heavy layer of AS5. So, the IHS should fit a tiny bit closer to the die and the AS5 will hopefully stay in place better (I think). Old school can still git 'er done!

EDIT: That screenshot is not what I was running when I got my new personal best. I just put that out to show what I was able to do with the RAM. For my PB, I had the CPU OC at 5.3 gig with those same RAM timings.

 

[Nebulous]

Nice work batz! I haven't benched in years except for stability when I'm finding my best 24/7 clock. Hmm..wondering what my PCmark10 benchmark scores would look like with my current setup?

 

[Mr.Scott]

Don't get too attached to that sub.
No sub test details in screenshot and no FM link for top 20 score.
It is invalid.