9900K Direct Die Cooling

[Noel 4,486]

I understand the idea of direct to die cooling.  Doesn't 9900K come w/ a soldered IHS?  If so, there is no point in delidding one unless doing direct to die cooling, correct?

If you need another 10 to 15 degrees C cooler consider installing a thru-the-wall air conditioner and install it so that its output blows into or at your PC and its intake ports.   The benefits of this approach is the entire PC case and contents cool down quite substantially including mainboard, DIMMS, CPU/GPU etc.  Plus, when it's time to upgrade the box you don't need to do anything put place the new box where the old one lives.   I get about 10-15C lower temps and i don't even route that air output directly into an open case which I thought might not be wise due to static electricity from higher air flow.   I've done this now for around 12y and still run my old SB-E at 4.42 GHz w/ HT enabled for the past nearly 6y.

Edited January 6, 2019 by Noel

[martin-w 3,234]

15 hours ago, Noel said:

Doesn't 9900K come w/ a soldered IHS?  If so, there is no point in delidding one unless doing direct to die cooling, correct?

 

 

Hi Noel.

The 9900K is soldered, yes. The solder is soft and in a thick layer. It can be delidded quite easily with something like a DieMate 2. the solder can be removed  with a blade. Once delidded, and the thick layer of solder replaced with a thin layer of liquid metal TIM, there can be a 5 to 10 degree reduction in temp. So yes, there is a temp reduction as a result of standard delidding, reinstalling the IHS and cooling in the usual manner. 

In terms of the direct die mounting, it's another 5 to 10 degrees reduction in temp, compared with an already delidded CPU. So compared with the stock, soldered CPU, the total temp reduction as a result of direct die mounting "could be" anywhere from 10 to 20 degrees as an estimate. 

Personally, I wouldn't bother accept for fun, and on the condition I could afford a new CPU if there were any issues.

 

Quote

If you need another 10 to 15 degrees C cooler consider installing a thru-the-wall air conditioner and install it so that its output blows into or at your PC and its intake ports.   The benefits of this approach is the entire PC case and contents cool down quite substantially including mainboard, DIMMS, CPU/GPU etc.  Plus, when it's time to upgrade the box you don't need to do anything put place the new box where the old one lives.   I get about 10-15C lower temps and i don't even route that air output directly into an open case which I thought might not be wise due to static electricity from higher air flow.   I've done this now for around 12y and still run my old SB-E at 4.42 GHz w/ HT enabled for the past nearly 6y.

 

You get 10 to 15 degrees less. With my 8700K, I get 15 degrees less as a result of delidding, and super quiet thanks to the awesome NH-D15S. 😊 Many get more like 20 degrees less. Having said that, I admire your creative approach to cooling.

I recall you went to such extremes as a result of high ambient temp where you live. Did I recall that correctly? 

 

Quote

I've done this now for around 12y and still run my old SB-E at 4.42 GHz w/ HT enabled for the past nearly 6y.

 

What was your CPU temp, under load, prior to your creative air con cooling Noel? Did you really need to?

[Noel 4,486]

The big improvement Martin over delidding and all of that is that THE ENTIRE BOX is cooled, mainboard, dimms, GPU.  Not just the CPU.  And further, the same cooling solution works for all future PCs, which can't be said for other solutions.  And super quiet is overrated for flight sims if you care for realism!   Flying in real aircraft is amply noisy!  And I use Sony studio monitor headphones anyway so can't tell the A/C is running when it is.

During the warm summers here ambient in the house gets up to 80F.   I never run the A/C full on, not because it's any louder, but because it isn't needed.   IF I run the A/C at full on at peak load flying my main core, and I am hyperthreaded for 6 cores, is around 50C.   It would be right at 70C at this ambient, for a 20C decrease in temp.   I adjust the A/C to allow the main thread to run at around 60C is all.  When ambient is around 70 or less I either don't use A/C or just use its fan-only setting.

Did I really need A/C w/ your recommended DH-14?  Well, maybe if I wanted my SB-E to last 6y overclocked to 4.42Ghz hyperthreaded!

Quite frankly Martin I'm amazed more people don't do what I'm doing.  It's both simple to do, durable, and costs very little to operate overall.   Plus you can't beat it to offset the hot GPU & CPU temps to keep the room cooler.

I find it hard to fathom delidding a 9900K off of its soldered IHS adds a whole lot w/o direct to die cooling.   I'd have to see several controlled tests to believe the numbers are anywhere near your stated claim of 5-10 degrees.

We were victims of the Camp Fire in Paradise, California and are moving out to Colorado soon.  I'm taking my PC and will transport it FLAT so that I don't need to remove the Noctua.  I did strap it in, suspend it, for when the case is upright, but don't want to take any chances.  I am trying to make the beast last until it won't go another minute because it's still good enough to run P3D V3.4 buttery smooth.

Do we know what's on the horizon for next gen 6-8 core Intel CPU yet?

Edited January 7, 2019 by Noel

[martin-w 3,234]

3 hours ago, Noel said:

The big improvement Martin over delidding and all of that is that THE ENTIRE BOX is cooled, mainboard, dimms, GPU.  Not just the CPU.  And further, the same cooling solution works for all future PCs, which can't be said for other solutions.  

 

 

Sorry about all the quoting Noel, and I'm playing devils advocate a bit with my comments. 🙂

 

And what was the temperature of the board, RAM and GPU prior to your air con solution? What I'm getting at is that most of us over cool our PC's. I'm trying to get a handle on whether I myself would deem the temps prior to your air con solution an issue or not. 

 

Quote

the same cooling solution works for all future PCs, which can't be said for other solutions

 

I would disagree, delidding is feasible on all CPU's with TIM rather than solder. And pretty much all coolers can be used on alternative CPU's.

 

Quote

  And super quiet is overrated for flight sims if you care for realism!   Flying in real aircraft is amply noisy!  And I use Sony studio monitor headphones anyway so can't tell the A/C is running when it is.

 

Most people don't want that sort of realism though. They want to enjoy the sim in comfort without excessive noise. Plus of course they don't just engage in flight sim, they play other games too. But at the end of the day, the main thing is that you yourself are happy with your set up and it's right for you. 

 

Quote

During the warm summers here ambient in the house gets up to 80F.   I never run the A/C full on, not because it's any louder, but because it isn't needed.   IF I run the A/C at full on at peak load flying my main core, and I am hyperthreaded for 6 cores, is around 50C.   It would be right at 70C at this ambient, for a 20C decrease in temp.   I adjust the A/C to allow the main thread to run at around 60C is all.  When ambient is around 70 or less I either don't use A/C or just use its fan-only setting.

 

80F is what, 26C? Pretty much the same as in the UK, on a hot summers day. We get quite a few of them these days. We don't have air con here, except in business premises. Our issue is keeping warm in the winter, so just central heating. My approach is to take into consideration the max temp I am likely to see in the summer, and overclock until I ma at a safe CPU temp based on that max ambient, summer temp.

 

Quote

Did I really need A/C w/ your recommended DH-14?  Well, maybe if I wanted my SB-E to last 6y overclocked to 4.42Ghz hyperthreaded!

 

D15S, not D14. I wasn't suggesting you adopt the D15, just pointing out how much the drop in temp is with delidding and the D15 compared with your air con solution. 

 

Quote

Quite frankly Martin I'm amazed more people don't do what I'm doing.  It's both simple to do, durable, and costs very little to operate overall.   Plus you can't beat it to offset the hot GPU & CPU temps to keep the room cooler.

 

Well to be honest Noel, there are a number of reasons why they don't.  Many live in cooler climates were it's simply not necessary, and a standard cooling solution, either a high end air cooler or AIO is perfectly adequate. In addition, if living in a warmer climate, air con is usually installed in the house anyway, and ambient temp is lower enough to not be an issue. They wont be dropping their system temps as low as you are, by blasting cold air directly into the PC, but to be honest, I'm still not sure that's necessary. The other point of course, as alluded to above, is that what I do, and I'm sure others, is to take note of the max ambient temp I am likely to see in the summer. I then overclock with this temp in mind, and make sure I have enough margin to be in the safe temp zone on a hot summers day. Overclock is then lower, but as you know, overclocking is generally linear, so a few hundred megahertz lower overclock is minimal in terms of frame rate. So my preference in your position, would be to overclock lower rather than use the air con solution. 

 

Edited January 7, 2019 by martin-w

[martin-w 3,234]

3 hours ago, Noel said:

 

I find it hard to fathom delidding a 9900K off of its soldered IHS adds a whole lot w/o direct to die cooling.   I'd have to see several controlled tests to believe the numbers are anywhere near your stated claim of 5-10 degrees.

 

 

 

Quite a few have delidded the 9900K and achieved these results. As I said, the solder is applied very thick, so not surprising that replacing it with a THIN layer of liquid metal drops temps. See video below re temps... 7:59 in video, a 10 degree drop in temp with liquid metal instead of solder. In addition, the 9900K die is also very thick, 0.42 for 8700K and 0.87 for 9900K, thus, lapping that down so it is thinner drops temps further. Lapping the die temps at 11:22 in video. 

 

 

 

Edited January 7, 2019 by martin-w

[martin-w 3,234]

Quote

We were victims of the Camp Fire in Paradise, California and are moving out to Colorado soon. 

 

Oh my god, sorry to hear that Noel. Was your home damaged? That must have been very scary.

[Noel 4,486]

Nope, our house wasn't damaged directly beyond the trees and vegetation around it, and some modest smoke damage.  Only about 5% of all homes survived the fire, so we now live in the middle of our very own Hiroshima!  But we're fine, thanks.

[Noel 4,486]

9 hours ago, martin-w said:

 

 

Quite a few have delidded the 9900K and achieved these results. As I said, the solder is applied very thick, so not surprising that replacing it with a THIN layer of liquid metal drops temps. See video below re temps... 7:59 in video, a 10 degree drop in temp with liquid metal instead of solder. In addition, the 9900K die is also very thick, 0.42 for 8700K and 0.87 for 9900K, thus, lapping that down so it is thinner drops temps further. Lapping the die temps at 11:22 in video. 

 

 

 

No way I'd remove a solid connection w/ metal solder for this outcome.  I will be running 9700K w/o HT so won't need this with my universal-fit A/C solution ;o)   

I believe more heat = accelerated wear, all else being equal.  This applies to all components in the box, that very hot expensive GPU, and all the rest.

[martin-w 3,234]

10 hours ago, Noel said:

No way I'd remove a solid connection w/ metal solder for this outcome.  I will be running 9700K w/o HT so won't need this with my universal-fit A/C solution ;o)   

I believe more heat = accelerated wear, all else being equal.  This applies to all components in the box, that very hot expensive GPU, and all the rest.

 

No, nor me. Not much point for 10 degrees. If it was a easy as TIM, then I might, but there is a bit more effort with solder, carving it off with a blade and then sanding the die. 

Yep, from my understanding of component degradation I would agree with that, re heat. RAM is rarely an issue re temp of course. The point re CPU degradation and heat is that its hard to estimate just how much degradation would be sustained. Personally though, in my experience, I don't think 10 to 15 degrees or so is that relevant. I would "guess" we would have to be dealing with temps at 90, daily, to shorten the lifespan to a degree that it failed just after the warranty period. If it's not throttling back its regarded as within the normal range, and designed to last the 3 year warranty period or more.

I would say most CPU's that are up at 80 degrees or less, should last well past the warranty period and still be going strong by the time most enthusiasts are ready to upgrade. In fact I had an old i7 920 quite a few years ago that was a great overclocker, I sold it to a friend. It's still going strong after, oh god, must be 8 years at least. Been overclocked pretty high right from new. No sign of degradation. 

So yes, I agree that in theory a cooler CPU undergoes less degradation and should, theoretically, have a longer lifespan. But I think we over estimate how fast degradation takes place. And under estimate how warm a CPU can be without significant degradation. 

Edited January 8, 2019 by martin-w

[martin-w 3,234]

10 hours ago, Noel said:

Nope, our house wasn't damaged directly beyond the trees and vegetation around it, and some modest smoke damage.  Only about 5% of all homes survived the fire, so we now live in the middle of our very own Hiroshima!  But we're fine, thanks.

 

Glad to hear it Noel. Coming from the UK of course, we have no experience of what you have gone through. Must have been a nightmare.

Surprisingly we do get quite a few tornado's in the UK. Little mini tornadoes, something like 30 per year are recorded. They rarely cause much damage though. Flooding of course, now more common. And believe it or not the occasional earth tremor when something moves in the North Sea, or old mine workings shift. I have experienced a few tremors over the years, the last one bad enough to wake my family up and swing open wardrobe doors. But who am I to talk about earth tremors when you live in California... 😀

[Noel 4,486]

1 hour ago, martin-w said:

I would "guess" we would have to be dealing with temps at 90, daily, to shorten the lifespan to a degree that it failed just after the warranty period. If it's not throttling back its regarded as within the normal range, and designed to last the 3 year warranty period or more.

So yes, I agree that in theory a cooler CPU undergoes less degradation and should, theoretically, have a longer lifespan. But I think we over estimate how fast degradation takes place. And under estimate how warm a CPU can be without significant degradation. 

Back in the old days when you could expect a major improvement in throughput about the time the 3y warranty ended I can see not caring if the thing died at 3y.   But since improvements in CPUs have been relatively modest in terms of P3D I decided to aim for durability.  6y out running an SB-E w/ HT enabled at 4.42Ghz for just a few months short of 6 years, and a lot of hours per day!   I'm not so sure I'd have the same experience had I not keep the CPU down around 60C or less all of these hours, but we'll never know for sure, granted.  This being said I would bet significant $$ if you took 10,000 SB-E chips all running at 4.42Ghz continuously you would absolutely demonstrate a significant difference in longevity w/ ones running at 60C over 80C.  I've had to bump up core voltage this past year up to 1.325v for this clock speed, and for the first few years it took only 1.305v, which I take a possible indicator of degradation.