Keeping an i7-4790S powered NAS cool as possible. [Delid?]

[iunlock]

I swapped out the i3 that came with my QNAP TVS-671 NAS with an i7-4790S, which is the highest I could go CPU wise for the TVS line. (65w TDP, max)

I've used MX-4 and reapplied it multiple times already because I haven't been happy with the Temps. (3/4th the size of a pea in the middle / dot method. This was the perfect amount as I saw it fully thinly covered when I took the heat sink off.) I really think I have a borked tube.

Transcoding 1080p media with 10+ simultaneous streaming and running VM's keep my NAS busy 24/7.

Here are my temps:

Ambient Room Temp: 72F
Idle Temp: 42-43C
Max Load Temp: 74C

* The threshold for the i7-4790S is 71.35C.

I know the idle temps aren't too bad, but in a very limited / tight NAS environment, I could use all the coolness I can get.

Inside is just an aluminum base heat sink with a "L" shaped copper piping that has radiators at the other end that sits behind two weak fans. (3000rpm max.)

Note: Other i7-4790S NAS owners that did the swap like me are idling at 35-37C and only peak at 62C under max load.

So for the heck of it I'm going to delid the CPU and use:

CoolLabs Ultra on the die -> IHS.

Gelid Extreme for the IHS -> hest sink.

* Because my heat sink plate is aluminum, CoolLabs Ultra is a no go...as it'll pretty much eat through it.

Another thing is that the heat sink doesn't sit 100% flush as one corner of it is slightly lifted compared to the others when fully seated. (Just "slightly") This is why I figured something more paste like (Gelid Extreme) would be ideal, even if I could have used CL Ultra had my heat sink not been aluminum.

Any thoughts? 74C at max load bugs me...since I'm limited on space and options, I'm thinking a delid + some good quality TIM is my only and best option.

I'm open to any ideas....

Thanks.

 

[eton975]

Any way to realign your cooler? Delid's a bit risky.

 

[dave_the_nerd]

Is that temp core temp or the casing/IHS temperature? The 71.35C number of for casing temp, core would be hotter.

Anyway, maybe should have stuck with a -T CPU at 45w TDP? The i3s and Pentium CPUs that ship stock with those units only have, like, 54w TDP mostly, and I'd be surprised if they ever come close to generating that much heat, especially the ones without Turbo.

 

[JeffMD]

Sounds like you are trying to make the heat transfer from the cpu to heatsink faster when the underlying problem is you have a heatsink that sucks. I was unable to find any details on the HSF used in this unit so I would just point out adjusting two things. a) better/faster fan(s) for the hsf fins and b) undervolting the cpu sacrificing a little bit of performance for lower temps.

 

[sdifox]

Nothing will change until you seat the heatsink properly. File if you have to.

 

[iunlock]

Thanks guys.

I agree with the heat sink having to seat properly first. I may have made it sound worse than it really is. After the last two times I've changed out the TIM, I had flashed my LED light between the small gaps on each side and this is when I had noticed that the one 'corner' didn't "seem" to have seated all the way flush. However, ever so slightly. I had just noticed a little bit of the TIM kind of bubble up in that corner...again every so slightly....I know the word bubble sounds terrible, which it is I know...but that's the best way that I can describe what I saw.

Too bad I couldn't take a picture of it as I've tried, but the gap was so slim in the first place that this was almost impossible to do.

I believe that the heat sink is making fairly even contact with the IHS, but I had just made mention of this to paint a complete picture for yall to understand all the variables at play here.

I would rather not have to change out the fans. I've adjusted the fans manually to medium and high and although that brings the temps down, I'd prefer not have to keep those blowing at those rpm's all the time. One for the noise and two...well just a preference as I rather would like to nip this issue at the bud, hence wanting to delid....

1. I'm still trying to confirm exactly what metal my heat sink plate is...I'm 90% sure it's aluminum, which sucks because CLL Pro or Ultra would eat through aluminum. This is why I've opted for using Gelid Extreme on the IHS->Heat Sink and CLL Ultra on the die->IHS.

2. Should I lapse the top of the IHS with Gelid Extreme? Or just leave it be?

3. Should I just go with the die direct to the Heat Sink and forget the IHS plate? (While using Gelid Extreme in this case if my heat sink plate is indeed aluminum.)

Very limited space...

Here are some pics of the bottom of the heat sink plate when I was testing the amount and dot method to make sure it covers well...This seems like it's aluminum yea?

 

[iunlock]

Should I just go bare die to heat sink and forget the IHS?

If so, I'd have to make sure the heat sink plate goes down far enough to make good contact with the die. (Mod the screws to lower it?)

I assume Gelid Extreme would be best for die->heat sink? Since Cool Labs Pro or Ultra would eat through the aluminum heat sink plate.

I'm all ears....

Thanks guys

 

[sdifox]

Should I just go bare die to heat sink and forget the IHS?

If so, I'd have to make sure the heat sink plate goes down far enough to make good contact with the die. (Mod the screws to lower it?)

I assume Gelid Extreme would be best for die->heat sink? Since Cool Labs Pro or Ultra would eat through the aluminum heat sink plate.

I'm all ears....

Thanks guys

you need thermal paste because that is the main conductor of heat. Metal surfaces are very irregular on the microscopic level and the paste fills them out.

I am assuming there is a fan assisting that radiator mounted outside the case?

 

[iunlock]

you need thermal paste because that is the main conductor of heat. Metal surfaces are very irregular on the microscopic level and the paste fills them out.

I am assuming there is a fan assisting that radiator mounted outside the case?

Yes, there's two fans on the back that sits in front of the radiator.

As for the thermal paste, indeed! I did mention using Gelid Extreme between the die and heat sink if I were to go without the IHS all together.

I'm curious if going direct die->heat sink (with thermal paste! Of course! ) would be more beneficial to aid in maximum overall cooling in my case due to the nature of the NAS environment being so limited.

 

[JeffMD]

Wait, you don't want the trouble of messing with the fans, but you are A-OK with prying the top of the cpu off and dropping a heat sink on top of its fragile core?

I can see delidding for things like water cooling and maximize over clocking. Doing it because the factory heat sink/fans don't seem to be cooling well enough is just...

 

[sdifox]

Yes, there's two fans on the back that sits in front of the radiator.

As for the thermal paste, indeed! I did mention using Gelid Extreme between the die and heat sink if I were to go without the IHS all together.

I'm curious if going direct die->heat sink (with thermal paste! Of course! ) would be more beneficial to aid in maximum overall cooling in my case due to the nature of the NAS environment being so limited.

err, no, the casing act as heat spreader.

 

[Ketchup]

If you are hitting the throttle only in the rarest of circumstances, I honestly wouldn't worry about it. They are make to do this, and in other scenarios (like laptops), do so quite often.

 

[iunlock]

Wait, you don't want the trouble of messing with the fans, but you are A-OK with prying the top of the cpu off and dropping a heat sink on top of its fragile core?

I can see delidding for things like water cooling and maximize over clocking. Doing it because the factory heat sink/fans don't seem to be cooling well enough is just...

Yes I'm A.OK with delidding. I take it you haven't done it before? Well it's not too hard.

I'm not messing with the fan because that's not the core issue.

If you have a home with a faulty heating unit, would you really approach it with just messing with the fans and ignoring the heating unit? That doesn't help.

I've delidded all my gaming rigs i7-4790K's. ...it's not the first time at the rodeo. I also have an extra i7-4790S laying around just in case, but I highly doubt it. It's not a big deal.

 

[iunlock]

err, no, the casing act as heat spreader.

I know...was just throwing the idea out there.

 

[iunlock]

If you are hitting the throttle only in the rarest of circumstances, I honestly wouldn't worry about it. They are make to do this, and in other scenarios (like laptops), do so quite often.

I really think this chip itself is from a bad batch as my other i7- 4790S in my QNAP TVS-871 model idles at 38C and barely visits 60C's under load. So I'm thinking there's definitely an issue with the stock TIM on this die with the heat issue.

 

[JeffMD]

If you have a home with a faulty heating unit, would you really approach it with just messing with the fans and ignoring the heating unit? That doesn't help.

But, your heating unit is fine. On the other hand your house is cold because it isn't transferring the warmth to the air outside of the heating unit. So yes, I would very much look at the fan situation. I mean there might be an issue with the contact point as you may have surmised, but if the contact point is fine then the issues would be the air escaping from the heat fins fast enough. but w/e, youve done it before, delid to your hearts content. I'm interested in what you find.

 

[iunlock]

With the temps that I've seen so far and also compared to other fellow i7 swappers with the same NAS model, my temps are abnormally high under load.

My idles are okay, but still I've noticed from a spreadsheet that we all share (the i7 swappers) that they all idle 2-3C lower than I do when we have the same ambient room temp. However, the idle isn't what's important here.

Also on the spreadsheet contains all the temps of when they were under 50%, 80% and 90+% CPU load. Not one of them reaches remotely close to 74C like I have. Also, keeping in mind that this is a NAS environment so with its uber compact nature, every bit of cooling helps.

I've also swapped out my TIM on the IHS->Heat Sink 3 times and it didn't help at all. (Used MX4...I seriously that's a borked tube...it's absolute rubbish and it's brand new!)

This is how I've concluded to the fact the chip itself either has:

A.) Crappy TIM on the die from the factory, which it does...

B.) Due to the slob job of the glue that they put around the IHS to hold the IHS down to the chip itself, the die may not be making a solid contact with the IHS. Or an even contact I should say. This has been a common problem with this line of chips.

C.) Although I do believe the top of my IHS is convex (rounded), I don't think this is the root cause. I'd say 99.5% of the surface area of the IHS is making contact with the Heat Sink, it's just that I did notice a little convex going on with one of the edges, but ever so slightly.

I'll be preforming the delid today and will be using CoolLab Liquid Ultra on the die and Gelid Extreme on the IHS->Heat Sink.

Wish me luck LOL...not in the delid so much as that's something that you and I can do in our sleep, but in hopes of this doing the trick for my temps to go down more than just a few degress in my NAS. These things are super compact...I almost want to rig something like the SilverStone Tundra Series TD02-LITE closed loop cooler for giggles....

I'll keep you posted mate.

 

[XavierMace]

One word of caution I would give before putting to much faith in the spreadsheet. Unless you are absolutely certain everyone is using exactly the same testing methods you can't trust those numbers 100%. How long was the CPU at specified load level when numbers were given and how is that load being achieved? For example the AVX tests on Prime95 causes unrealistically high temps.

 

[iunlock]

I understand that. The figures were all from averages and a lot of them I talk to regularly. Despite the obvious variables that do exist, once you get over a certain point in temp that is different from the rest of the pack, it doesn't need much thinking to understand that something isn't right.

I've just successfully finished my delid and my idle temp is already 2-3C lower than before. I'll do some stress tests tomorrow....

 

[Yuriman]

I'm confused, reading this thread. Why are you concerned about 74c when maximum safe core temperature is 105c?

http://ark.intel.com/products/80808/Intel-Core-i7-4790S-Processor-8M-Cache-up-to-4_00-GHz

71.35c is "TCASE", which is the maximum temperature of the heatspreader. When you monitor temperatures though, you're not looking at heatspreader temperature, but core temperature. The heatspreader doesn't have a temperature sensor in it.

 

[iunlock]

Yes indeed. Unfortunately, the stock nas's software doesn't show anything but the CPU temp, system temp, and HD temps. As I've mentioned, I was basing my higher than normal heat temps against the averages of others who have the same set up as me. With the same ambient room temp and with the same under load figures, my NAS was much higher across the board.

I've just finished a stress test and the NAS now is running at a cool 45C under 80% load, where as before it was 74C at 80%. Something is obviously working....thanks to the CoolLab Liquid Ultra on the die and Gelid Extreme on the IHS->Heat Sink.

Thanks for the info though...much appreciated....pretty fascinating stuff.

 

[sdifox]

Glad to hear it worked out

 

[iunlock]

Thanks. Much appreciated.

 

[iunlock]

It's done...results are amazing. Couldn't be happier.

 

[JeffMD]

Yea.. that looks pretty ugly under that flip top. Is it normally dry and hard like that? seems like the wrong knock to it could break contact internally and never reform again.