TR Forums

Finally cracked 4.6 GHz on my i7-4790K using Cool Laboratory Liquid Ultra between the die, heat spreader, and in between the heat spreader heatsink. I tried getting to 4.7 GHz it almost looked like it might happen but stopped when I had the vcore at 1.4 still wasn't stable gave up trying. However CPU temperature never exceeded 80c when I tried 4.7 GHz with the vcore at 1.4 which says something for the heat conduction capabilities.

Bought it from OutletPC currently discounted from $19.87 to 10.98 was $10.56 went I ordered.

Screenshot from stress testing, the heavy multitasking test was the one it kept blue screening with WHEA hardware until I nudged the vcore up from 1.2 to 1.225.

Correction: I ran the newest version of RealBench 2.43 system hard locked had to bump the voltage to 1.23.

So what no one wants to try there luck with applying liquid gallium as a thermal compound on die, and for the heatsink?

Well I was wrong my chip can hit 4.7 GHz, turns out I did need to drop the Uncore to 4.5 GHz to make it stable.
Tinypic

Tried 5.0 GHz just to see, it was sorta of stable once I bumped the voltage to 1.45 but heat, and power consumed was alarming.
Backing down to 4.9 GHz wasn't much better so I went for 4.8 GHz with the voltage at 1.35, looks like a success to me.

yeah, i tried the original liquid pro, it was amazing stuff. very different in application and performance compared to regular thermal paste. if you've got steady fingers, it's totally worth it. if not, say goodbye to your motherboard, lol.

glad they found a version that's slightly easier to work with.

It comes out of the syringe looking like regular paste compound BUT once you start spreading it out it takes on the appearance of liquid metal. The fun part was trying to get it to stick to the silicon die all it wanted to do was stick to the applicator brush. Once I managed to get some sticking to die the rest of it went on. I almost managed to get it over in the row of capacitors on the one side of the die.
The easy part was spreading it over the heat spreader.

biffzinker wrote:

Correction: I ran the newest version of RealBench 2.43 system hard locked had to bump the voltage to 1.23.

So what no one wants to try there luck with applying liquid gallium as a thermal compound on die, and for the heatsink?

It makes absolute sense to use the best TIM available between the IHS and the die along with between the IHS and your main HEATSINK!
Plus the Gallium never dries out.

JUST MAKE SURE YOUR HEAT SINK HAS A ALL COPPER BASE AND YOUR POSITIVE IT WILL NOT GET ON ANY ALUMINUM SINCE GALLIUM EATS ALUMINUM FOR BREAKFAST LUNCH AND DINNER!!! Very Dangerous to the aluminum!

Recently did my 6600k with Thermal Grizzly Conductonaut (their metal paste). The difference in temperatures was unbelievable as well as how much it evened out temperatures under load across the cores. I"m talking like 20c at least. Used regular paste on the IHS as I've heard of heavy discoloration and difficulty removing off copper that isn't nickel plated.

The heatsink base is solid copper block with nickel plating same as the heatspreader is copper with nickel plating so no worries.

The heatsink I'm using is Scythe Kotetsu (SCKTT-1000)

Edit: I see both the Cool Laboratory Liquid Ultra, and Thermal Grizzly Conductonaut have "Made in Germany" on them. Wonder why Germany is popular for liquid metal compound?

vargis14 wrote:

JUST MAKE SURE YOUR HEAT SINK HAS A ALL COPPER BASE AND YOUR POSITIVE IT WILL NOT GET ON ANY ALUMINUM SINCE GALLIUM EATS ALUMINUM FOR BREAKFAST LUNCH AND DINNER!!! Very Dangerous to the aluminum!

Didn't know it attacks steel though per wiki.

Gallium attacks most other metals by diffusing into their metal lattice. Gallium, for example, diffuses into the grain boundaries of aluminium-zinc alloys or steel, making them very brittle. Gallium easily alloys with many metals, and is used in small quantities as a plutonium-gallium alloy in the plutonium cores of nuclear bombs, to help stabilize the plutonium crystal structure

Gallium Wiki

With a thermal conductivity of just 40.6W/(m-K), gallium is not your dream TIM material. Heck, it's probably not even as good as Vegemite.

Silver: 429W/(m-K)
Copper: 401W/(m-K)
Gold: 318W/(m-K)

All from Wiki

The numbers here disagree somewhat with Wiki's numbers, but the graph clearly shows the supremacy of the top three thermally-conductive elements.

That chart misses diamond.

http://hyperphysics.phy-astr.gsu.edu/hb ... thrcn.html

No offense, but you can do WAY better than 4.6-4.7GHz on a 4790k de-lidded with Liquid Ultra/Pro. I ran mine @ 5GHz and 1.4V for months before deciding to drop it down a bit to reduce voltage and turn down my radiator fans. The key is uncore speed. 4.5GHz is completely unnecessary. Try 4GHz and see if you can't push that core up another 100-200MHz at the same voltage, or even turn down the voltage.

Ryu Connor wrote:

That chart misses diamond.

http://hyperphysics.phy-astr.gsu.edu/hb ... thrcn.html

Hmm, I see a niche industry here. Spend $1,500 on the best CPU and $1,000,000 on a pure flawless diamond TIM shim. You'll sell millions.

I'll stick with the Vegemite. At least I know where to get that.

Man made diamond in the very fine particulate matter for you need for a TIM is cheap.

http://www.amazon.com/IC-Diamond-24-Car ... d+24+carat

I already tried having the core at 5.0 GHz, and uncore at 4.0 GHz problem was this chip wants more than 1.4 vcore it only starts to stabilize at 1.43 and it was iffy at that. The total power consumption jumps up to 180 watts with temps at 80-85c with it going at 5.0 GHz.

I tried stepping down to 4.9 core/4.0 uncore wasn't much better still wants around 1.35-1.375. Th sweet spot for this chip seems to be 4.8 with the vcore at 1.32-.33 the uncore clocked at 4.5 doesn't cause any instability/heat issues unless I set it to 4.6.

Been working on figuring out if I can get the vcore down to 1.3 at 4.8 GHz.

1.3x @ 4.8 is a pretty good sweet spot for long term use, that's where I'm at now after about a year of 1.4 @ 5.0. I mis-read your initial post and thought you were still running 1.4V, btw, I see that you're at a much better voltage now.

Ryu Connor wrote:

Man made diamond in the very fine particulate matter for you need for a TIM is cheap.

http://www.amazon.com/IC-Diamond-24-Car ... d+24+carat

Except it has been known to cause visible pits, and etching in the heatsink base when the compound is cleaned off.

Been using it for years. Never seen that behavior. FUD.

Pits would require huge diamond chunks and it doesn't have that. That concept doesn't even pass the smell test.

Carbon staining is the worst thing it does.

I'd also note that every Clevo reseller uses/offers it.

That's a huge number of systems.

http://www.dickblick.com/items/27014-7154/

23.75K gold leaf. It's malleable, so it'll adapt to any surface inconsistencies. Wonder how many layers you need to get a good interface (and one that doesn't involve goo)?

http://www.amazon.com/Genuine-Silver-Le ... B005H84FOM

Silver leaf here.

http://www.amazon.com/Speedball-Mona-Ge ... B001U3X39G

For giggles, copper leaf.

It won't be cheap, but this might make for a very interesting TIM test.

Ryu Connor wrote:

Been using it for years. Never seen that behavior. FUD.

Pits would require huge diamond chunks and it doesn't have that. That concept doesn't even pass the smell test.

Carbon staining is the worst thing it does.

I'd also note that every Clevo reseller uses/offers it.

That's a huge number of systems.

Perhaps I mixed up which diamond compound was causing the pitting issue, and this wasn't the one responsible. I heard about it from either XtremeSystems or Overclock.net don't remember which was reporting it.

I've seen the reports too, many years ago. It was IC Diamond people blamed, but it just didn't make sense in the context of the pitting.

I'm not even sure you could managed to extrude something the size of the little pits there were pictures of out of the tubing.

An outside contaminant made more sense, but nobody in those few threads wanted to hear that discussion.

IC says they use a less than 40 micron particle of diamonds. I've often wonder if that mean something technically smaller than that, but regardless. That's really, really small. It's the edge of visibility as described by this linked picture. If the IC site has a typo and it's <.40 micron (like Arctic Silver particles), then it's even more infinitesimal.

People also freaked over the carbon staining, which yeah, it will do. But you're not exactly staring at the heat spreader of the CPU, so who cares.

Also Clevo is a huge ODM. If they're not sweating the issue, then I guess it's up the customer to decide if they're idiots or if it really isn't an issue/cause for concern.

Settled on running my i7-4790K at 4.9 GHz since I lasted posted about trying for 4.8 GHz with the Vcore at 1.3.
In case anyone was interested how it scores running at 4.9 here is some results from Geekbench, and Cinebench.


Geekbench Browser