Expected delid-relid results with CLU on OC'd i7-6700K?

[BonzaiDuck]

So, this is my understanding of things since Sandy Bridge days.

Ivy-Bridge TIM brought us the attempts and results of delidding as tested and reported by IDontCare.

Then, either with Haswell or Devil Canyon Haswell-refresh, Intel touted a new "polymer" TIM that was better or more efficient.

I assume this latter material is used for Skylake.

If you replace it with CLU and then restore the processor cap, what should be expected as a temperature improvement for 6700K @ ~4.7Ghz and the likely thermal wattage?

Perhaps I might have better posted on "CPUs and OC'ing," but the issue or question addresses the relative results of CLU versus stock in centigrade temperature.

 

[YBS1]

I can make an uneducated guess. I have my delid tool sitting here along with some PETG hardline and associated tooling but haven't torn into the NCase yet (I forgot the glue for the collars), I'm going to get around to it at some point this month. I don't know just how much effect it will have on the stable and benchable clocks allowed but I know for certain that TIM is a massive thermal bottleneck. Even under a custom loop now the 6700K at similar clock speeds as compared to the 5960X immediately jumps up to about 12C higher than the 5960X under load. Now the loop my 5960X is in is anything but ordinary but I honestly don't think it would make even a two degree difference if the 6700K was being cooled by the same equipment, there is a bottleneck and I don't think it would matter how much water you threw at it, it's going to run in the mid 70s under load until that goop is gone. I'm guessing I'll see a drop to about 60 to 64 degrees @ 4.8GHz under load once I complete the delid.

 

[BonzaiDuck]

I can make an uneducated guess. I have my delid tool sitting here along with some PETG hardline and associated tooling but haven't torn into the NCase yet (I forgot the glue for the collars), I'm going to get around to it at some point this month. I don't know just how much effect it will have on the stable and benchable clocks allowed but I know for certain that TIM is a massive thermal bottleneck. Even under a custom loop now the 6700K at similar clock speeds as compared to the 5960X immediately jumps up to about 12C higher than the 5960X under load. Now the loop my 5960X is in is anything but ordinary but I honestly don't think it would make even a two degree difference if the 6700K was being cooled by the same equipment, there is a bottleneck and I don't think it would matter how much water you threw at it, it's going to run in the mid 70s under load until that goop is gone. I'm guessing I'll see a drop to about 60 to 64 degrees @ 4.8GHz under load once I complete the delid.

And for my assurance -- you say that you're replacing the IHS with CLU installed between it and the die? I just wanted to make sure what you were saying.

I've been looking at one "true" AiO, two different "customizable" AiO kits (and you can guess probably right), and a couple "custom" water-cooling kits. There's another thread probably still on the first-page pile discussing AiOs, air, water and options for what I took to be a noob OP. RussianSensation offered voluminous posts, pictures, graphs -- circles and arrows on the back of each one. At one point, he pronounced that Air on a 6700K was still the most cost-effective choice, touting many air-coolers I had seen -- and someone introduced a list with some coolers I hadn't looked closely at before.

And I think I can beat an EKWB Predator 240 with a retail-box original i7-6700K, just by adding a beefier case exhaust and adding a pusher fan 140mm to . . . a certain cooler. The cooler is marketed as a "passive" heatpipe tower, but it has a ducting solution a benchmark separate from a no-duct configuration. Before I order an AiO, EK or Swiftech or custom kit, this is a $70 outlay and I want to try it. The benchmark comparison with the EKWB AiO showed the latter with only a 5C edge -- with the stock TIM on the EK and the heatpipe configuration, a limp Noctua exhaust and no pusher fan on the latter. At either 4.7 or 4.8, I'm hoping the load temperatures don't exceed 70C. I think the TCASE spec on those chips is about 68C. So that's a pretty cool standard to meet, in my subjective opinion. It could turn out better than that.

A benchmark comparison I looked at also suggested mid-70s for a 4.6+ OC without TIM replacement or other enhancements. If the TIM replacement means a drop of 6C or more as you suggest, then this could be my BIG 6700K NO-WATER COP-OUT. Which -- I might rule as "success."

 

[YBS1]

And for my assurance -- you say that you're replacing the IHS with CLU installed between it and the die? I just wanted to make sure what you were saying.

I've been looking at one "true" AiO, two different "customizable" AiO kits (and you can guess probably right), and a couple "custom" water-cooling kits. There's another thread probably still on the first-page pile discussing AiOs, air, water and options for what I took to be a noob OP. RussianSensation offered voluminous posts, pictures, graphs -- circles and arrows on the back of each one. At one point, he pronounced that Air on a 6700K was still the most cost-effective choice, touting many air-coolers I had seen -- and someone introduced a list with some coolers I hadn't looked closely at before.

And I think I can beat an EKWB Predator 240 with a retail-box original i7-6700K, just by adding a beefier case exhaust and adding a pusher fan 140mm to . . . a certain cooler. The cooler is marketed as a "passive" heatpipe tower, but it has a ducting solution a benchmark separate from a no-duct configuration. Before I order an AiO, EK or Swiftech or custom kit, this is a $70 outlay and I want to try it. The benchmark comparison with the EKWB AiO showed the latter with only a 5C edge -- with the stock TIM on the EK and the heatpipe configuration, a limp Noctua exhaust and no pusher fan on the latter. At either 4.7 or 4.8, I'm hoping the load temperatures don't exceed 70C. I think the TCASE spec on those chips is about 68C. So that's a pretty cool standard to meet, in my subjective opinion. It could turn out better than that.

A benchmark comparison I looked at also suggested mid-70s for a 4.6+ OC without TIM replacement or other enhancements. If the TIM replacement means a drop of 6C or more as you suggest, then this could be my BIG 6700K NO-WATER COP-OUT. Which -- I might rule as "success."

Actually I'm going to try Thermal Grizzly Kryonaut. I'll see how that goes, I'm certain it won't beat CLU, but it's said to be very good. I don't have any doubt you'll see a minimum of 6C drop in temps, if I only see that, I'll actually be disappointed in the delid.

 

[BonzaiDuck]

Actually I'm going to try Thermal Grizzly Kryonaut. I'll see how that goes, I'm certain it won't beat CLU, but it's said to be very good. I don't have any doubt you'll see a minimum of 6C drop in temps, if I only see that, I'll actually be disappointed in the delid.

Kryonaut has also been recommended to me with a pointer to a benchtest review comparison. It may have depended on the test-bed: I was curious how it rated with ICD, and the recommended review showed Kryo about 1C better, while another review showed ICD as 0.7C better. But you can't use ICD between IHS and processor die, or advised not to do so. Kryo still falls about 2C short of CLU. Of course it would depend on the test-bed, but the benches I'd seen were for a 3930K (140W), a 4770K @ 4.5 Ghz, and a 6700K.

The real issue, as we discuss here, is the difference between the Intel TIM and either Kryonaut or CLU.

And you could probably use CLU between the heatsink-base or waterblock and the IHS. But it makes the cooler or waterblock difficult to remove. I wouldn't be able to say whether there's risk for the relidded processor cap. I actually might replace the ICD with CLU if the cooler and CLU-relidded CPU show good results.

YOu may be disappointed; I may be merely satisfied. I think I can squeeze out an extra 3 to 4C with my fan and duct options, so add 6C to that and it's nearly a 10C reduction. Maybe not worth the effort or expense (I had Silicon Lottery prepare my 6700K), unless that edge matters for something. Silly Lottery shows a range of between 8C and 18C. Since they explicitly say they relid their work unless otherwise told, that range probably covers it.

I may know for sure in a week's time. Parts arrive Friday. I'll want to put it together in time to meet any RMA requirements.

 

[ClockHound]

CLU is proven not to pump out. Haven't seen the evidence that Kryonaut will not. Kryonaut is a great TIM, but CLU is the de-lidder's safe choice.

Both are better than Intel TIM, but the stock TIM isn't horrible. What is horrible is the variable amount between the IHS and the die. It's a TIM-thickness lottery. SL does good work. You're winning the silicon AND the TIM-thickness lottery. Enjoy!

 

[BonzaiDuck]

So, for any interested, here's an update. SL replied to my e-mail with Core-Tj load values on my processor using an H105 cooler -- before the delid and after the relid. It looks like a 12C reduction in the package temperature.

Now, for the record, I was going to keep quiet about this until I put the system together and began testing with screen captures as I OC'd. I picked the TR Le Grand Macho, following two comparison reviews that included an NH-D15S and an EKWB Predator 240 -- for which I was just about willing to shell out ~$200.

With the TR rubber-duct between the Macho and a limp Noctua 120mm exhaust fan, the Macho matched the D15S while the EKWB showed 5C lower load temperatures.

The Macho weighs in at 900 grams without any fans hung on it, or 80 grams less than the bare D15S. However, TR issued an advisory at their web-site of reports that coolers exceeding 500 grams might deform the PCB substrate of the Skylake CPU. I would expect the real risk for a cooler of >= 900g, but they distributed a specially-designed plastic shim or spacer to all their resellers: it is tapered from about 0.3mm to 0.7, or something proportional to those dimensions. It distributes the pressure from the cooler mount evenly across the CPU PCB, as opposed to the two pressure points of the latch plate without the shim.

I think I can get another 3 to 5C just by adding a fan similar to the Macho RT bundle and using an iPPC 3000 fan for the ducted rear exhaust. So I'm wondering if I could beat an EKWB Predator 240 (stock lid and TIM for 6700K) by 10C degrees.

Howsoever it beats an EKWB Predator 240, it won't much matter if OC'd load temperatures for 4.6 to 4.8 Ghz peg at the mid 50s or mid 60s. I don't get much of anything for temperatures lower, or maybe 100 Mhz with any luck and more voltage. The result may well show a maximum below the TCASE spec for the CPU. If I can do that, why bother with water?

Or why bother with it at this moment?

 

[ClockHound]

Exactly. Big Air or Really BIG Water. Just put the tweener water down the drain. ;-)

 

[BonzaiDuck]

Yes. And if I continue reviewing water options over the next several months, the Stacker has been prepared to leave space for a ~295mm radiator or water system at front intake, and with the same exhaust strategy. If I eliminate front-panel USB3 and a little red toggle switch I installed next to those ports in the 5.25" spacer-screen around the 3.5" USB3, I can probably use a rad as big as 320. If I eliminate the front-panel SATA hot-swap bay, I can probably install a rad of 360mm.

My enthusiasm will depend a lot on the reported 12C delta for the CLU-relid, and whether I can squeeze a full 5C from a Macho pusher fan and the iPPC 3000 exhaust. No. I forgot -- I changed my mind this morning.

I was going to use the iPPC 3000 to replace a Gentle Typhoon AP-30 PWM 4,200 RPM unit in my original Sandy Bridge. Then, use the AP-30 with the LG Macho. I just really wish they'd given a different model name to it. Like the Grand 135, or the TR-E150. Or something that isn't so . . . glitzy-promotional.

I think I'll sand the nickel-plate off the base. That could be another 1 or 2C. It pains me -- the thought that I could put CLU on the bare-copper IHS and heatsink base, knowing that I could do mischief trying to break apart the gallium-copper amalgam.

Here's a question to consider, if it even makes sense or might resolve that problem.

Suppose you mixed some dry nano-diamond powder with CLU, applying it otherwise as directed. Would it be easier or harder to break apart?

I'll settle for ICD until I need to change something next year.

 

[ClockHound]

Jeez....sanding and amalgam? Sounds too close to dentistry to me.

I know you like to plan extreme and execute to the more extreme plan, but why not leave a few areas of improvement for future tweaks?

Another 1-2c is margin of error stuff...a bigger improvement might come from closely monitoring cooler intake temps versus ambient and optimizing intake airflow. The best of my case mods can achieve a 1-2c gradient between cooler intake temp and room ambient. Most stock cases can have a 10-15 degree higher intake temp. Your case is not most, of course. ;-)

And just call it the LGM.

 

[BonzaiDuck]

Jeez....sanding and amalgam? Sounds too close to dentistry to me.

I know you like to plan extreme and execute to the more extreme plan, but why not leave a few areas of improvement for future tweaks?

Another 1-2c is margin of error stuff...a bigger improvement might come from closely monitoring cooler intake temps versus ambient and optimizing intake airflow. The best of my case mods can achieve a 1-2c gradient between cooler intake temp and room ambient. Most stock cases can have a 10-15 degree higher intake temp. Your case is not most, of course. ;-)

And just call it the LGM.

Well, it became a habit with these tower-coolers. I actually patted myself on the back for not lapping the i5-3570K for my brother's rig. It was an accomplishment of restraint.

I have seen people -- seen firsthand -- put four fans in the side (120 or 140) and two in the front of their Stacker 830.

If you take time to seal off every other unvented or unducted exit/exhaust, and every vent that isn't part of your strategy, then you can also selectively reduce the number of fans.

My approach has been to channel air over specific components and exhaust it to keep the channels from mixing. The only exception in that setup involves the PSU: all surplus pressure and exhaust that doesn't go through the cooler, doesn't get sucked out by the barrel fan for the mobo-duct -- it all gets forced through the PSU.

I might be able to trim down to two intake fans. But these four don't make any noise . . .

 

[ClockHound]

If your sneaky ducts method avoids the issue of unwanted turbulence from the crossed air streams, then sure. Especially if you have all your ducts in a row.

Best case airflow diagnostic investment I've made is on a $10 indoor/outdoor digital thermometer. Measuring cooler intake temp vs room temp is a valuable tool to help find the optimum intake/exhaust config.

 

[BonzaiDuck]

If your sneaky ducts method avoids the issue of unwanted turbulence from the crossed air streams, then sure. Especially if you have all your ducts in a row.

Best case airflow diagnostic investment I've made is on a $10 indoor/outdoor digital thermometer. Measuring cooler intake temp vs room temp is a valuable tool to help find the optimum intake/exhaust config.

I think I had such a thermometer from Radio Shack. And I think it died. Or got lost.

On the "margin of error stuff," if it's a "real grain of rice" and not just measurement error, then they all add up and shift your error distribution in the downward direction.