Reasonable drop for liquid metal?

[mackerel]

In short, I just applied Conductonaut to a 6700k (stock, 4.0 GHz all core turbo, 1.25v-ish), cooler 120mm AIO.

Before core temps were 80, 76, 78, 73C. After was 72, 69, 71, 68, giving differences of -8, -7, -7, -5C. Same stress test was performed, in the form of 8 threads of 64k in-place FFT Prime95 29.2. It was left to settle for 17 minutes in each case with the 120mm AIO on it. Ambient was around 27C at the time of the "after" results, I forgot to check on the "before" but it might have been slightly cooler as it was earlier in the day before the sun kicked in.

Note the "before" is delid with MX-2 paste I think. Is 7C about right ball park here? Are those reporting bigger temp differences comparing pre-delid?

 

[EarthDog]

Most results over 10c or so, other variables snuck in there. Drops seem fine to me.

 

[mackerel]

I just wanted to check as there are 3 conditions:
1, pre-delid
2, delid, conventional compound
3, delid, liquid metal

I'm comparing 3 to 2, not 3 to 1. It's been too long I can't remember what I got going from 1 to 2.

 

[EarthDog]

Typically the highest values would obviuosly be from 1 to 3.

 

[Sentential]

Can't give you much to work with but I have a laptop 6700HQ with liquid metal applied directly to the core and HSF and my load temps are 65/66/66/64 and idle temps are 26/25/24/24 and that is at 1.22v so I'd imagine yours should be quite a bit lower given prior to LM I was in the high 80s low 90s.

 

[trents]

In my experience, delidding with using conventional pasted like MX2 under the hood will give about half the temp drop that using liquid metal TIM will. Liquid Metal will usually give a 10c-20c drop. So you're right there.

 

[mackerel]

Can't give you much to work with but I have a laptop 6700HQ with liquid metal applied directly to the core and HSF and my load temps are 65/66/66/64 and idle temps are 26/25/24/24 and that is at 1.22v so I'd imagine yours should be quite a bit lower given prior to LM I was in the high 80s low 90s.

You're giving me ideas now... I could do my laptop also... it is also a 6700HQ. But not today...

 

[Sentential]

You're giving me ideas now... I could do my laptop also... it is also a 6700HQ. But not today...

For laptop or cooling confined environments liquid metal, specifically Conductonaut like what you got has the higher Indium content does wonders for keeping heat at bay. Go look at notebookcheck forums for a guide and be sure you are also buying Super33 electrical tape and any appropriate Fujipoly thermal pads for the VRM and memory. I know it's expensive but if you do it right and not rush you'll only have to do it once.

 

[TimoneX]

Stuff is runny though and might not be the safest thing around for a mobile gadget. You'll get similar results and less risk with something like liquid ultra...which is more viscous.

 

[Sentential]

Stuff is runny though and might not be the safest thing around for a mobile gadget. You'll get similar results and less risk with something like liquid ultra...which is more viscous.

That's what the Super33 is for, and honestly it really isn't that bad. Once the solvent/suspensions boil off it's actually pretty thick.

 

[mackerel]

Thought I'd do "before" temp measurements on my laptop before eventually repasting it. Encountered rather interesting throttling I've not noticed before. If I run anything AVX it seems to throttle down to around 2 GHz compared to base of 2.6, and non-AVX tests are above 3 GHz. If I look at hwinfo64 I see all sorts of current/power throttles activating. I wonder if this is the CPU itself, or the bios set rather protectively? Even with that limiting, I got hottest core to 69C with P95. Windows power profile is set to max performance, and I don't recall any settings in bios but will take another look. It also has a H-chipset so no OC as such, but I still wonder if I can relax some of the limits somehow.

 

[trents]

I think this is the bios being protective. After all, it is a laptop. It's normal for even a desktop bios to have an AVX offset of about -2x by default. Besides, how often do you actually use AVX?

 

[mackerel]

OCCT was also triggering that throttling, but now on a retry it doesn't seem to be. Hottest core now up to 80C. Also P95 is no longer throttling either. I have a hunch as to what was happening. When I started earlier, the laptop battery was low. I've heard before that it may go into a restricted power state when when plugged in if the battery was low. Now the battery is almost full, it is running at full power.

Also, before Skylake-X I haven't encountered AVX offset before. Can't recall when it was added. Was it Kaby Lake? Everything else I have is Skylake or earlier. On desktops, I probably spend 95%+ power on time running AVX loads if not gaming or other tinkering. I don't normally do it on the laptop.

 

[trents]

Makes sense.

 

[mackerel]

I kinda suspected temps on the desktop where I applied the liquid metal was getting hotter again. I just left it running P95 with same settings as before, and it was hottest core 77C coolest core 70C. Difference in test are I'm now using P95 29.3, and ambient is probably cooler. Liquid metal isn't supposed to go off right, so I have to wonder what's happening. Oh, I've also updated the bios, but the same OC has been applied. I'm half debating taking off the waterblock and see what the liquid metal is like. I'm still a little concerned I put it on too sparingly, as the 1st attempt was far too little, and the 2nd one I have now seemed better.

 

[Sentential]

I kinda suspected temps on the desktop where I applied the liquid metal was getting hotter again. I just left it running P95 with same settings as before, and it was hottest core 77C coolest core 70C. Difference in test are I'm now using P95 29.3, and ambient is probably cooler. Liquid metal isn't supposed to go off right, so I have to wonder what's happening. Oh, I've also updated the bios, but the same OC has been applied. I'm half debating taking off the waterblock and see what the liquid metal is like. I'm still a little concerned I put it on too sparingly, as the 1st attempt was far too little, and the 2nd one I have now seemed better.

Could be any number of things. LM is permanent and does not dry, especially conductonaut, the temperature rise could be any number of things including ambient temp. As for the right amount you want to use the same amount you would if you were going to paint the core with a normal TIM. Basically paint the core with LM until the core has no visible bald spots. It might be hard to see so try moving your head different angles to make sure its completely covered.

I've had conductonaut on my laptop at least as long as you have on your desktop and if anything my temps have gone down now up.

 

[mackerel]

Thinking more, I can't guarantee my fan configuration is the same as when I originally applied the liquid metal. Due to not having the right size fan, I swapped things around when doing the 7800X build, and I think LM was applied before that. So airflow might be lower. If I can find the right settings, I'll try increasing the fan speed on the rad. By hand, it feels a bit slow, so I think there is more potential to be taken there. Using a thermal camera the exhaust area was around 36C after running for a long time, and I feel it could be closer to ambient than that. With traditional pastes, I like to go in heavy, which is obviously not a good idea with liquid metal.

Regardless of that, the core difference is still a bit of a concern. 2 cores are more hotter than others which could be suggestive of mounting variation.

 

[Sentential]

Thinking more, I can't guarantee my fan configuration is the same as when I originally applied the liquid metal. Due to not having the right size fan, I swapped things around when doing the 7800X build, and I think LM was applied before that. So airflow might be lower. If I can find the right settings, I'll try increasing the fan speed on the rad. By hand, it feels a bit slow, so I think there is more potential to be taken there. Using a thermal camera the exhaust area was around 36C after running for a long time, and I feel it could be closer to ambient than that. With traditional pastes, I like to go in heavy, which is obviously not a good idea with liquid metal.

Regardless of that, the core difference is still a bit of a concern. 2 cores are more hotter than others which could be suggestive of mounting variation.

That sounds far more reasonable, again as for the LM I think people have overblown the whole "only use a little" I used a ton on mine and a ton on the Vega56 I have. Temps are great and I'm not at all worried about it spilling since I've sealed the socket area. I don't use any less when applying LM than I do traditional paste. The difference is I don't let it spread on its own and use a Qtip to make sure there's an even layer on the die.

The last time I tried doing the whole "maybe I can make the temps just a liiiiiiiitle bit better" and thought there wasn't enough pressure on my Vega56 I ended up shearing a screw so I've slowly started to learn that the more you mess with something the more likely it is you'll break it. I chipped a 8800GT doing the same thing so just be careful.

 

[Tír na nÓg]

I use coolaboratory on my laptops since my first de-lid (3770K), and average temp drop on load was 15 to 20c on the 5 of them. Did not notice any increase in temp over time (checked after thorough can cleaning).

Regarding the 7c difference between core, there is nothing to sorry about IMO.

I have done dozens (literally!) of remounts and it never changed anything. I think is due to the fact that some cores are physically closer to the IMC.

What I noticed is the higher the OC, the bigger the difference.

E.g.: on the rig in sig, there are 1/2c delta@stock, 4/[email protected] and 6/[email protected]. This has been consistent on all my Intel PC's for the 4 past years.

 

[mackerel]

That sounds far more reasonable, again as for the LM I think people have overblown the whole "only use a little" I used a ton on mine and a ton on the Vega56 I have. Temps are great and I'm not at all worried about it spilling since I've sealed the socket area. I don't use any less when applying LM than I do traditional paste. The difference is I don't let it spread on its own and use a Qtip to make sure there's an even layer on the die.

With conventional pastes I've usually gone for "too much is better than too little" but the extra risk involved with LM means it is more difficult. You still don't want too little, but too much is now more of a bad thing too. It is a fine line, and I'm not sure I'm treading along it optimally.

What I noticed is the higher the OC, the bigger the difference.

E.g.: on the rig in sig, there are 1/2c delta@stock, 4/[email protected] and 6/[email protected]. This has been consistent on all my Intel PC's for the 4 past years.

This makes sense, as there is a thermal gradient, so as power increases, differences will also. I think my temps are higher than desired even at stock, so something is under-performing. Still haven't tried setting a faster fan speed yet. Always a tricky balance between cooling and noise.