Buying decision

[ehume]

I do heatsink reviews for a couple of online mags. I just reviewed the NH-C14S. When I turned it so it's heatpipes went vertical, I gained 5c of cooling. Now, however when I try to run the NH-D15 and NH-D15S, they throttle my CPU, even at 4.4GHz. The NH-C14S was doing in the high 80's at 139+ Watts. The other two are throttling even on Watts in the low 130's. Also, I see the imprint of IHS edges on the contact surfaces of the two D-type heatsinks. Diagnosis: I have dented the CPU's IHS. It is now useless for reviewing.

I am faced with 2 alternatives. Here they are:

1. Buy another Devil's Canyon. Advantages: I already have the mb, ultralow profile RAM and Win 8.1-64 for it. Disadvantages: It is now 2 generations behind. It is getting older. I will have an orphan CPU.

2. Buy a Skylake system now. Advantage: It is the current generation. No chip would be orphaned -- the chip with the sag is still a fine chip; I could give it to my daughter. I already have an unused Win-7. Disadvantages: It is an early stepping. I would be buying a new mb ($) with an immature BIOS. I would have to buy RMA twice -- in-store RAM for tomorrow, ULP RAM on Amazon (Micro Center does not have ULP RAM).

I had intended to wait for Kaby Lake. By then I could take my time, get 16GB of ULP DDR4 RAM for it. The BIOS will be mature. The MB will be 2d gen. But the i7 4790k was ruined recently. I was unable to get anywhere near my earlier numbers. I tried multiple remounts. Pfah!

MC is an hour away, so tomorrow is my day to go. What are your collective thoughts?

 

[Yuriman]

You other alternatives: Lap the IHS until it's flat, OR pop the lid

 

[MongGrel]

You other alternatives: Lap the IHS until it's flat, OR pop the lid

This.

I've never popped the lid on a chip myself, but if you can't lap it out sounds like a good one to try it out on.

Depends on if it's a chip viable to do it on, not sure on those.

 

[SOFTengCOMPelec]

In theory, if you modify the Intel chip, by de-liding it or lapping it. It may change the dimensions enough to be unfair to the review process.

Also lapping it, may improve its (the cpus) thermal properties. But that would mean that your review results, were NOT typical of the original cpu.

E.g. You incorrectly think that it cools a lot better with the new heatsink. But really it is because you lapped the cpu.

 

[ClockHound]

I vote for getting another 4790k. Technically, it's 2 gens behind, but in reality it's only one, Broadwell was just a 14nn desktop pipe cleaner. And with a new platform your previous reviews won't correspond with future ones.

Wait until DDR4 prices drop/BIOS matures before going Skylake. Will you even be able to get a mythical 6700k tomorrow at MC?

Or do the unthinkable and go for door #3, a 5820k/X99/DDR4 combo.

Or get both. A 4790k and a Sklyake combo and redo all your tests and post a big 22nm, little 14nm comparo.

Sorry, I'm not really helping. Maybe bring your magic penny to MC tomorrow and flip for it. ;-)

 

[ehume]

In theory, if you modify the Intel chip, by de-liding it or lapping it. It may change the dimensions enough to be unfair to the review process.

Also lapping it, may improve its (the cpus) thermal properties. But that would mean that your review results, were NOT typical of the original cpu.

E.g. You incorrectly think that it cools a lot better with the new heatsink. But really it is because you lapped the cpu.

This. Also, I don't have the time to delid, pound the IHS flat, lap the inside of the IHS, then relid. Lapping the inside of the IHS is too fine for me to pick up. I've never even been able to paint; so much for fine work.

I'm thinking now that I don't even have time to set up a new system. I can buy the new CPU (they still have them; I checked), re-OC and be up and running tomorrow night to test the heatsink that arrived today. I'll put it up against the NH-D15.

I don't see any of you rushing to advocate for the Skylake option.

Edit: cute, ClockHound.

 

[Yuriman]

If I were in your shoes, I'd probably just take the chip to some 400 grid sandpaper until the nickel is gone, then shine it up with some 600, 800 and 1200. Don't even bother to take it off.

EDIT: Your own CPU, lapped, will probably be more in-line with your older results than a new un-lapped CPU.

 

[ClockHound]

I don't see any of you rushing to advocate for the Skylake option.

That's because most of us North Americans have only seen a 6700k on the internet. Another week and I'll have the datasheet memorized and my virtual overclocks will be legendary.

Oh..and what new heatsink (that won't be available in North America for months) will you be testing? ;-)

 

[ehume]

That's because most of us North Americans have only seen a 6700k on the internet. Another week and I'll have the datasheet memorized and my virtual overclocks will be legendary.

Oh..and what new heatsink (that won't be available in North America for months) will you be testing? ;-)

Deepcool K2.

This is frustrating. I had, all lined up: NH-D14, NH-D15, Megahalems, Genesis and Silver Arrow. Based on quotes from a TR rep in a thread on SPCR, and now that I've had my chip sunk, I'm reluctant to mount the Silver Arrow on my new chip. Maybe I'll mount it on my old chip.

What's interesting is that on my older i7 860, I tested the NH-L12 and found something similar. It made for a long review -- it turned into an exploration.

 

[MrTeal]

Another alternative that I've always wanted to see reviews use is a resistive heat load. When I was evaluating different AOI coolers for performance I made tested using a piece of copper buss bar. It's an ~4" x 4" piece with the outside milled down so there's a raised 35mmx35mm section in the middle simulating a CPU IHS, and six 200W surface mount resistors on the opposite side to provide a known, controllable and repeatable heat load.

 

[ClockHound]

Deepcool K2.

This is frustrating. I had, all lined up: NH-D14, NH-D15, Megahalems, Genesis and Silver Arrow. Based on quotes from a TR rep in a thread on SPCR, and now that I've had my chip sunk, I'm reluctant to mount the Silver Arrow on my new chip. Maybe I'll mount it on my old chip.

What's interesting is that on my older i7 860, I tested the NH-L12 and found something similar. It made for a long review -- it turned into an exploration.

Oh noes....not the vertical vs horizontal mounting dilemma again. Is this behavior observed only with the low-profile pancake coolers?

 

[ClockHound]

Another alternative that I've always wanted to see reviews use is a resistive heat load. When I was evaluating different AOI coolers for performance I made tested using a piece of copper buss bar. It's an ~4" x 4" piece with the outside milled down so there's a raised 35mmx35mm section in the middle simulating a CPU IHS, and six 200W surface mount resistors on the opposite side to provide a known, controllable and repeatable heat load.

FrostyTech uses a heatplate method: http://www.frostytech.com/testmethod_mk2.cfm

Their tests correspond fairly well to 'on cpu' testing. One thing I have noticed tho, is how the direct-pipe coolers performed better generally at FrostyTech than other tests methods that use actual cpus.

 

[ehume]

Oh noes....not the vertical vs horizontal mounting dilemma again. Is this behavior observed only with the low-profile pancake coolers?

Noctua tells me that the NH-C14S I got was a prototype, so it might not be too representative of shipping units.

I did try the NH-D15S crosswise -- heatpipes vertical. Maybe 1c difference, at most. Not the 5c I saw with the C14S. So yeah, we now have two pancake coolers with big differences when they are mounted 90 degrees apart.

What I find interesting is that I am getting a prototype. Am IO getting other pre-shipping units? I know I got an NF-P14 redux that clicked. They told me it was a prototype, and the other units did not click. The replacement was silent. But that's two admitted prototypes.

 

[DrMrLordX]

Another vote for a hotplate standin for an actual CPU.

 

[ehume]

Another vote for a hotplate standin for an actual CPU.

You can get them from Intel . . . at a cost. Personally, I don't trust them.

 

[NTMBK]

Delid the 4790k and give it to your daughter, and buy a Skylake system for thermal testing.

 

[ehume]

Delid the 4790k and give it to your daughter, and buy a Skylake system for thermal testing.

I have to buy a replacement today. The Skylake will be expensive. It will be OK, but more like the original Haswell compared to Devil's Canyon/Kaby Lake.

I'm pretty sure now that I'll get another DC, and wait another year or two for Kaby Lake.

 

[ehume]

When I went to take the mounting hardware off, I found one loose finger screw. I tightened it, remounted the D15 with Arcticlear. Then tested again. Still very hot.


Then I got the arc clips off. One of the spacers was broken, shattered. Replaced it. Remounted the D15. Still hot. Tomorrow I'll replace it with the new i7 4790k I got this morning.

 

[know of fence]

Having a convex/bulging heatsink contact surface is a completely terrible idea, but just like with PC cases every stupid idea will eventually find a way into production as a desperate attempt to be different or interesting. Maybe you should annotate your review with your actual experience and a dented lid.

Also have you ever tried a torque screwdriver for the various mounting solutions, or with the hundreds of CPU coolers you've tested have you ever come across recommendations for how much torque [Nm] should be applied to the screws.

 

[ehume]

Having a convex/bulging heatsink contact surface is a completely terrible idea, but just like with PC cases every stupid idea will eventually find a way into production as a desperate attempt to be different or interesting. Maybe you should annotate your review with your actual experience and a dented lid.

Also have you ever tried a torque screwdriver for the various mounting solutions, or with the hundreds of CPU coolers you've tested have you ever come across recommendations for how much torque [Nm] should be applied to the screws.

I agree. A terrible idea. It's a good thing the CPU was dented after the review was finished.

The torque screwdriver idea is interesting. I'll have to look at various solutions. I'm not sure that accurate torque screwdrivers can be had for reasonable cost (my wife complained when I spent $280 for a new CPU). In auto stores, there are torque wrenches. I've not heard of torque screwdrivers, which probably means that the ones that exist are expensive.

Of the various heatsinks, that I have seen, none look at torque. All of them look at pressure applied to the IHS. Around 2010, there was some controversy when Thermalright units allowed for a 70-pound pressure to be applied, exceeding specs (max 60 lbs). In my Silver Arrow, there is a hole where the center screw used to be. All of the mounts use either captive springs or the inherent springiness of the mount frame itself to apply pressure. And basically all of them have screw stops of one kind or another. That means screw until you can't go any further. The Noctua mount has a captive spring surrounding a captive bolt, that has internal threads for two frame-mounted bolts. There is a definitive stop, and the instructions tell you to screw down the tightening bolts until you hit the stops. That is typical these days.

 

[ehume]

Ah, the new CPU. With the same settings that produced 135W, the new chip is producing 121-122W, and the temps are only going to 78c. All this with identical settings -- Vcore reported as 1.212v, Vrin reported as 1.824v,m all cores at 4.4GHz. I used to run my old setup at 4.5GHz. I dropped to 4.4GHz when the D15 ran unexpectedly hot. I will probably return the system to 4.5GHz now. What a difference a year of new steppings makes.

 

[ClockHound]

Ah, the new CPU. With the same settings that produced 135W, the new chip is producing 121-122W, and the temps are only going to 178c. All this with identical settings -- Vcore reported as 1.212v, Vrin reported as 1.824v,m all cores at 4.4GHz. I used to run my old setup at 4.5GHz. I dropped to 4.4GHz when the D15 ran unexpectedly hot. I will probably return the system to 4.5GHz now. What a difference a year of new steppings makes.

178c - these later steppings are thermally tougher too. :biggrin:

Happy to hear you got a less volt thirsty and cooler chip.

 

[ehume]

178c - these later steppings are thermally tougher too. :biggrin:

Happy to hear you got a less volt thirsty and cooler chip.

Good catch. Thanks. )

 

[Idontcare]

There are a number of concerns raised in this thread which are worth discussing.

For starters, root-cause analysis on how your chip's IHS "got dented"? If you don't know how it happened the first time around, and don't develop protocols to prevent it from happening again, then this thread is going to become a bit of a "rinse and repeat" down the road once you bend the IHS on whatever replacement CPU you intend to purchase.

Second, you are concerned that a lapped CPU will result in cooling data that are invalid for comparison to non-lapped ("stock") IHS, but at the same time you acknowledge you are generating data with completely invalid (for the purpose of data comparisons) prototype HSF's?

You see what I'm getting at here? You have nothing in the way of ensuring you don't trash your next CPU's IHS, and you aren't generating apples-to-apples data anyways because you aren't being given retail HSF samples for your tests.

So...who cares if you lap your existing CPU IHS and just add one more disclaimer to the existing questionable cooling results?

 

[ClockHound]

So...who cares if you lap your existing CPU IHS and just add one more disclaimer to the existing questionable cooling results?

His daughter. She inherited the dented 4790k. And now you want to take it away.

His wife. How's he gonna explain $280 worth of fine grit Intel DC sandpaper? And how he made their daughter cry?

Stable testing environments and stable domestic environments. Pick one. :biggrin: