A messy kitchen, 9900K UK edition, 85F but chilly inside

[Charlatan 68]

On 11/24/2018 at 5:07 PM, Rob Ainscough said:

Tight fit in my limit SimPit space:

 

Blimey!  Be cheaper to have your own cessna than all that gear 😉

[shivers9 1,174]

Very nice looking setup Rob. I showed that Pic to my wife and suggested that she has time in the evenings and weekends for a couple more jobs to supplement her day job and I could get a pit just like yours. Now I am going to miss a whole day of simming just to replace that bedroom door!!

[Guest]

10 hours ago, westman said:

i pass on this CPU

The 9900K runs hot, too hot ... the soldered IHS was a mistake and is clearly NOT able to transfer heat well ... no idea why they went that direction ... hopefully not because of some young YouTuber's.

Disappointed with ASUS and the Z390 ... don't think I'll go this direction in the future, stick with X series quad channel, more PCEi lanes or move to AMD.  Why are manufacturers wasting resources providing LN2 specific options (dip switch, UEFI settings, etc.) for the pointless exercise of freezing that isn't sustainable except for one very exotic solution from EVGA.  LN2 community represents about 0.0001% of buying consumers (actually they probably don't even buy most of what they freeze, get it for free) ... I would prefer these manufacturers spend time on quality components/designs for the majority use case.

The ASUS AI features are primitive and in some cases very dangerous to the components ... and for what?  So they can come up with an OC that is worse than what most can do manually?

Think it's time for MB manufacturers to get back to the basics ... spend more resources on design/quality/tolerances and less on useless fluff.  I'd happily pay for something that actually works rather than voodoo AI and LN2 options ... but I don't want to pay for zero practical use case LN2 options.  Charge the 0.0001% use case LN2 folks a fortune for custom LN2 MBs and stop making the rest of us pay for it.

Anyway, fortunately my PC setup is very interchangeable and easy to swap out components ... and I can't see myself sticking with this CPU/MB for long.  Sorry to disappoint anyone following this thread.

I'll post my my P3D results later ... but so far they aren't looking much better than my 7900X and I'm having issues running SLI/NVLink at 8 X 8 rather than 16 X 16 that I could on the 7900X ... so no NVLink/SLI numbers.

And the Time Spy numbers really should have been closer to 26000 mark IMHO.

Per Ady's suggestion, I might try single loop using the Koolance to do it all (CPU, VRM, GPUs) ... I think my quick disconnects should make that an easy job to change ... but I honestly don't think that will help much as the problem isn't cooling, it's heat transfer from CPU just not being efficient.

Apologies for being so negative, but it is what it is.

Cheers, Rob.

EDIT:  On the positive side, the MB does have some very nice visuals and a boot /error status display that looks almost holographic ... it's a very pretty MB.

[Guest]

1 hour ago, Charlatan said:

Blimey!  Be cheaper to have your own cessna than all that gear

I wish that were the case ... and I wish I had the time to fly at all hours of the day/night.

 

50 minutes ago, shivers9 said:

Now I am going to miss a whole day of simming just to replace that bedroom door!!

Haha ... you could always tell her the PC can be used as a portable AC unit also  🙂

Cheers, Rob.

[fghdgdfdfgfgf 111]

7 minutes ago, Rob Ainscough said:

 Sorry to disappoint anyone following this thread.

What?!? I think most here appreciate your effort. I think we were all pulling for you and Intel and Co., but the quest was to learn and we learned.

[BrettT 63]

Rob,

I just appreciate the candid opinion on it.   You always approach things very scientifically so I typically find your results reliable and presented in a way that I can relate to.  This insight will likely either have me hold onto my 6850K @4.2 or go for an 8700K/8086K and shoot for 5.0 or take things one part at a time with the video card upgrade most likely being next (presently a 1080 powering a Vive with a Vive pro on order).  Thanks for the insight.

Thanks,

 

Brett

[martin-w 3,234]

9 hours ago, Rob Ainscough said:

The 9900K runs hot, too hot ... the soldered IHS was a mistake and is clearly NOT able to transfer heat well ... no idea why they went that direction ... hopefully not because of some young YouTuber's.

 

 

The soldered IHS is not a mistake Rob. It's superior to the method they were using before, namely Dow Corning paste and silicone glue. Given the heat generated by this chip, solder was their only choice. If they had stuck with the TIM and glue method they used previously the temps would have been higher. The lowest temps would have been with the strategy used by enthusiasts... liquid metal and barely any, or no,  glue. But of course, Intel will never adopt that method. So no, I don't believe it's related to "what YouTuber's have said". 

The point is though, if they had stuck with the old method, namely Dow Corning paste and glue, temps would have been crazy, limiting stock frequency but we enthusiasts would have had an easier time delidding and achieving the best temps possible.

 

Quote

Why are manufacturers wasting resources providing LN2 specific options (dip switch, UEFI settings, etc.) for the pointless exercise of freezing that isn't sustainable except for one very exotic solution from EVGA.  LN2 community represents about 0.0001% of buying consumers (actually they probably don't even buy most of what they freeze, get it for free) ... I would prefer these manufacturers spend time on quality components/designs for the majority use case.

 

Exactly! It's bonkers. But then motherboard manufacturers often seem to make weird decisions. With the latest generation of boards, they have been messing about with Turbo again apparently. Violating Intel's Turbo Duration.

https://www.gamersnexus.net/guides/3389-intel-tdp-investigation-9900k-violating-turbo-duration-z390

 

Quote

The ASUS AI features are primitive and in some cases very dangerous to the components ... and for what?  So they can come up with an OC that is worse than what most can do manually?

 

Strange you should say that. Many including der8auer are impressed with it. Might be just your immature BIOS.

 

Quote

Anyway, fortunately my PC setup is very interchangeable and easy to swap out components ... and I can't see myself sticking with this CPU/MB for long.  Sorry to disappoint anyone following this thread.

 

If you intend to just scrap it... why don't you delid it? Quite a few have now with success. Some have ground down the die as well to reduce thickness, one of the issues re heat build up. 

Edited November 26, 2018 by martin-w

[J0nx 48]

I replaced my 8700k with a 9900k over the weekend and am not seeing much improvement in anything except for encoding and other multi-thread apps. I bought my cpu locally and have until middle of January to return it. On the fence at the moment. I quite like the improvements all around in apps that can take advantage of it but my 360 AIO is barely keeping it under control at 1.35v running AVX instructions. Just not sure I like the chip $600 worth though...

Edited November 26, 2018 by J0nx

[fghdgdfdfgfgf 111]

5 hours ago, martin-w said:

 

If you intend to just scrap it... why don't you delid it?

Scrap it Rob!  One mans trash is another mans treasure ,,,right? Delid the flipping garbage can and dump it ....dump it all!!!!

I'll be rifling through your garbage over the next few weeks  .....imagine finding a couple of 2080ti's, a Maximus XI Formula, 9900K (with a lil mayo on it), 4400 c18 memory  ......it would be a flipping bonanza in that garbage can! ....So, Rob, addy please. I promise only to be in the can on garbage nite and only after the lights are out. 🙂

[Guest]

6 hours ago, martin-w said:

If they had stuck with the TIM and glue method they used previously the temps would have been higher.

I think you're missing the point, IHS, paste, delid, etc. are 300Mhz improvements if you're lucky?  The heat is still there, as the die process gets smaller and smaller (10nm next) it's going to be even harder and harder to transfer heat quickly and efficiently.  If one wants to remove heat quickly one needs surface area regardless of the IHS, paste, delid, etc. etc. ... moving forward the surface area is going to be smaller and smaller.  The solution to this problem needs something more than what YouTubers are "reacting" to ... the hammer and chisel method doesn't work when measurements are in nano-meters.

36 minutes ago, FunknNasty said:

it would be a flipping bonanza in that garbage can!

And I thought it was just raccoons last night :)

Cheers, Rob. 

[AlphaInfinity 13]

Rob,

 

Thank you for all the information on this new processor! I was also underwhelmed by the lack of PCI lanes. I planned on running dual M.2, SLI 2080TI's as well as a 10Gb interface card.  This whole architecture seems like a step down from the X299 platform. Intel really tried to wow us all with that 5ghz turbo speed but it seems like everybody is having problems with heat. Based on the information you provided, I am now considering just upgrading from the 7820x to the 7920x or 7940X and extending the life of my existing machine.  Thoughts?

Edited November 26, 2018 by AlphaInfinity

[martin-w 3,234]

"" think you're missing the point, IHS, paste, delid, etc. are 300Mhz improvements if you're lucky?  The heat is still there, as the die process gets smaller and smaller (10nm next) it's going to be even harder and harder to transfer heat quickly and efficiently.  If one wants to remove heat quickly one needs surface area regardless of the IHS, paste, delid, etc. etc. ... moving forward the surface area is going to be smaller and smaller.  The solution to this problem needs something more than what YouTubers are "reacting" to ... the hammer and chisel method doesn't work when measurements are in nano-meters.""

 

I'm not missing any point. The tests have been done. Delidded and liquid metal installed is significantly cooler than the solder.

Liquid metal and die lapped coolest - 13 degrees cooler than stock

Liquid metal - 8 degrees cooler than stock

Solder warmer

Conventional paste and glue even warmer.

Intel had two methods at their disposal. Dow Corning paste and glue or solder. They chose the coolest... solder. This was not a mistake.

 

 

Edited November 26, 2018 by martin-w

[Guest]

5 minutes ago, martin-w said:

I'm not missing any point. The tests have been done. Delidded and liquid metal installed is significantly cooler than the solder.

Significant cooler at what frequency?  HT ON, AVX 0?  Maybe I missed something, but I haven't seen anyone running HT ON and AVX = 0-2 Cache at 4.6 or higher at 5.5 Ghz delidded and/or otherwise that isn't hitting 100C during high loads (AVX included in load testing) ... except when frozen.  

You need to think outside the box and dump the hammer and chisel method ... YouTubers need to approach Intel from a very different angle ... squirming over 300Mhz just isn't worth the effort.  How's Intel going to solve it's heat problem going forward?  Heat to frequency curve is exponential regardless of delid, liquid metal, removing layers around the CPU, etc. etc.

But hey, I'm open ... would love to see a video of a 9900K not running hot with above configuration.

Cheers, Rob.

[martin-w 3,234]

Quote

Significant cooler at what frequency?  HT ON, AVX 0?  Maybe I missed something, but I haven't seen anyone running HT ON and AVX = 0-2 Cache at 4.6 or higher at 5.5 Ghz delidded and/or otherwise that isn't hitting 100C during high loads (AVX included in load testing) ... except when frozen.  

 

Not relevant to the point I was making. I didn't say anything about 5.5 GHz at under 100 degrees, or any temp or frequency. Not once did I make any frequency or temp claim Obviously there comes a point where no matter what the die/IHS interface happens to be, heat still builds up. We don't possess Area 51 alien tech. 

My point was that a thin layer of liquid metal has been demonstrated, a number of times, (see video) with the 9900K, to be superior to Intel's thick layer of solder. Thus, if my recollection is correct, and you intend to scrap this chip, delidding with liquid metal is an option that is superior to solder and could drop your temp.  Your mileage will vary dependant on how far you push your overclock and obviously if you push the chip to insane levels it may result in zero improvement.  Same for lapping the die, silicon isn't  a particularly good material to conduct heat, thus, reducing the thickness of the die is a proven method to drop temps more. See video! Now if I recalled incorrectly, and you aren't scraping the chip, perhaps selling it on eBay, then it would clearly be madness to delid it. 

 

Quote

You need to think outside the box and dump the hammer and chisel method

 

I don't need to do anything of the sort. I don't own a 9900K and would never buy this CPU. However... as you are engaged in an endeavour to carry out extensive testing of this CPU and have (IIRC) said you will scrap this chip if it doesn't meet your expectations... the extra effort and risk of delidding this chip may be enlightening. Personally, if I were about to scrap  a chip, I wouldn't hesitate to test this particular scenario. As you said on page one... "I'm willing to fry the CPU and/or MB, so we'll see what I can do". 

 

Quote

How's Intel going to solve it's heat problem going forward?

 

The above is nothing to do with any of the points I was making. You said...

 

Quote

"the soldered IHS was a mistake and is clearly NOT able to transfer heat well ... no idea why they went that direction"

 

The soldered IHS is not a mistake, the reason they "went that direction" is because Intel only had two choices, Dow Corning paste and glue, or solder, solder is somewhat superior so the right choice. Intel knew the chip would be hot so chose the best solution they had available. Again... testing has clearly demonstrated that Intel solder is thermally more efficient than the old paste and glue they were using, thus it was the only choice they had. They are hardly going to use paste and glue when they know the chip is hot.  

Furthermore, Intel don't give two hoots about your endeavours to reach super high frequency at sub TJ Max temps. As long as the CPU runs at stock they are happy. Intel make absolutely no guarantees regarding your overclocking results.

And Rob... don't forget, this is a CPU that has a 95 watt TDP. That is Intel spec. How they designed it. It's motherboard manufacturers that are pushing the chip beyond that, which results in excessive temp when overclocking.  

 

Quote

squirming over 300Mhz just isn't worth the effort

 

A comment that I am famous for. And no, YouTubers aren't squirming over 300 MHz, they conclude its not worth the increased risk and effort for the average enthusiast.. But then they aren't scraping their CPU's, like I recall you said you would. 

 

 

.

 

Edited November 27, 2018 by martin-w

[martin-w 3,234]

deleted

Edited November 27, 2018 by martin-w