Question New i9-9900K NH-U14S - Prime95 - High Temps

[MrN1ce9uy]

Looking for feedback on what you would do, or what you think I should do.

System - https://pcpartpicker.com/list/xkTyw6

This system is for a customer. They will use it as their main PC; which could include anything but mostly working with Microsoft SQL and maybe some video editing.

I tested with Prime95 and it reached a max of 93C after 15 mins. And that's at 4.7GHz w/ multi-core enhancement disabled and running around 1.2V.

Do you think this is a concern? Even with the NH-D15 I feel like it only gets 3-4C cooler temps. That would still leave me around 90C.

 

[Barty1884]

I'm assuming you're using a current version of Prime95, including AVX instruction sets?
While that does create additional heat.... I believe SQL is/can be an AVX workload, so would be a potentially accurate workload for the customer.

Is that 4.7GHz all core? Or a 5GHz with -3 AVX offset?

IIRC, De8auer and others delidded these things after launch, found the die to be uneven and, although Intel moved back to solder....it's not the greatest.
Unless you want to start getting into delid territory, which doesn't sound like a great idea for a customer.... I don't think there's much you can do here.

 

[MrN1ce9uy]

I've used both AVX and w/o AVX,

*4.7GHz w/o AVX = 93C
*4.4GHz w/ AVX = 91C

*all core speeds.
**ambient temp is 71F.

letting it run for 15 minutes.

 

[Phaaze88]

Does the customer have to have a 9900k?
Starting with Intel's hyperthreaded 8 core models going up, liquid really is the only way to go with these heaters. A 360mm AIO is what I'd get for it - a 280mm could work as well.
I have the 7820x delidded along with a D-15S with 3 NF-A14 IPPC 3000 pwm(I did 3 just for the hell of it), and just trying to push for a 4.5ghz OC on this thing gets uncomfortably toasty - around 90C on Realbench's stress test.
I imagine it'd be worse on Prime.
I plan to do full custom loop later this year.

 

[MrN1ce9uy]

I'm assuming you're using a current version of Prime95, including AVX instruction sets?
While that does create additional heat.... I believe SQL is/can be an AVX workload, so would be a potentially accurate workload for the customer.

Is that 4.7GHz all core? Or a 5GHz with -3 AVX offset?

IIRC, De8auer and others delidded these things after launch, found the die to be uneven and, although Intel moved back to solder....it's not the greatest.
Unless you want to start getting into delid territory, which doesn't sound like a great idea for a customer.... I don't think there's much you can do here.

Does MS SQL utilize all cores/threads?

 

[Barty1884]

I've used both AVX and w/o AVX,

*4.7GHz w/o AVX = 93C
*4.4GHz w/ AVX = 91C

*all core speeds.
**ambient temp is 71F.

letting it run for 15 minutes.

That does seem pretty high for 4.7GHz w/o AVX at 1.2V. You're confident 1.2V is actually what's being delivered to the CPU?

Does MS SQL utilize all cores/threads?

Honeslt,y I'm not sure.

 

[MrN1ce9uy]

That does seem pretty high for 4.7GHz w/o AVX at 1.2V. You're confident 1.2V is actually what's being delivered to the CPU?

That's what HWInfo64 is showing. Max is 1.24V and average is 1.2V with it dipping to 1.19V a good part of the time.

 

[Barty1884]

That's what HWInfo64 is showing. Max is 1.24V and average is 1.2V with it dipping to 1.19V a good part of the time.

I'm at a loss then, honestly. The cooler is pretty good, for what it is.... but those temps are little on the high side for 4.7GHz no AVX and ~1.2V.
Just to be clear, those are "max" temps.... Are they actually sitting at max a lot of the time? Or are those simply occasional spikes?

 

[MrN1ce9uy]

Well,.. I'm running it again and it's been 8 minutes (still running) and the temps are between 77C - 85C with a spike up to 87C max. It seems like last time it was further spikes gradually up to 93C. Let me see..

After 10 mins that one core is hanging around 87C most of the time. I'm expecting git to spike up again.

88C after 13 mins.

18 mins and it's back down to hanging around 85C with 88C max.

Idk, I guess that's about normal?

 

[Barty1884]

Yeah, that sounds more like it. Still warm, of course.... but that's the nature of the beast with these chips.

Discounting 'spikes' if only occasional & brief, gives you a better idea of a "real world" temperature maximum.

 

[mdd1963]

The U14S appears to be roughly about half the total size (stacks and surface area for heat dissipation capabilities) of a D15....

It not throttling is about the best you could hope for with a single small heat sink stack and a single fan...the fans are so quiet, you could try running it full speed to see if that helps, but, I imagine it already is running max as speeds get to 90+C)

(I'm amazed it is handling the 9900K as it is with no throttling, frankly)

 

[MrN1ce9uy]

The U14S appears to be roughly about half the total size (stacks and surface area for heat dissipation capabilities) of a D15....

It not throttling is about the best you could hope for with a single small heat sink stack and a single fan...the fans are so quiet, you could try running it full speed to see if that helps, but, I imagine it already is running max as speeds get to 90+C)

(I'm amazed it is handling the 9900K as it is with no throttling, frankly)

Can't find the mounting bracket to my R1 Universal, but I am in the process of mounting my NH-D15S on there. I'll post back when I test it.

 

[MrN1ce9uy]

Okay, I'm running Prime95 with the NH-D15S (w/ 2 fans) and it's better, much better than my first run with the U14S, but only as I expected, about 3C cooler than the last run with the U14S.

Operating temps are about the same hanging around 78-85C, but the max is only 85C so far at 7 mins.

13 mins and still the same.

14 mins went to 86C.

My conclusion: Operating temps are practically the same, with max spikes lowered by 2-3C.

*Note: I had to turn the NH-D15S around upside down so it would fit with 2 top fans mounted in the case. I wonder if that will affect cooling because it should be shifted toward the top, now it is shifted toward the bottom of the case.

Also, I found the Cryorig R1 mounting bracket. I might install it tomorrow.

 

[MrN1ce9uy]

So the R1's backplate has what seems like a lot of wiggle room after screwing down the 4 pillars that hold it in place. The whole bracket lifts up and down about 1-2mm.

But, if you look at this video at @00:29 and slow it down to half speed, you can see him lift each side of the bracket up about the same amount of room that mine has.

View: https://www.youtube.com/watch?v=lPnQfQR0tbQ

 

[MrN1ce9uy]

Okay, testing the Cryorig R1 Universal right now.. looking very good!

All of these tests were Small FFTs btw.

3 mins in and temps hanging around 67-72C, spiked up to 78C though.

5 mins temps hanging around 74-76C, spiked to 79C.

6 mins back down to previous temps at 3 mins.

Spiked to 80C around 8-9 mins. Back to hanging around 67-73C.

10 mins 81C on a different core, and 2 other cores have spiked to 80C. But average temps are low, 68-73C.

Oh... at 15 mins it jumped up to 88C.

16 mins 100C.

Well... idk. Vcore was at 1.21V.

So disappointing.

*Edit: Do you think the wiggle room on this cooler caused that?
**Something is wrong, trying P95 again it goes to 95C and climbing in just 15s.

 

[MrN1ce9uy]

I swapped back to the NH-D15S with 2x NF-A15 1500RPM fans. I removed the two stock Phanteks fans I had on top so I could place the D15S right-side-up.

Prime95 Small FFTs for 20mins reaches a max of 85C, with average temps between 75-83C.

I'm going to wrap it up and ship it out.

Thanks for the input.

 

[MrN1ce9uy]

The U14S appears to be roughly about half the total size (stacks and surface area for heat dissipation capabilities) of a D15....

It not throttling is about the best you could hope for with a single small heat sink stack and a single fan...the fans are so quiet, you could try running it full speed to see if that helps, but, I imagine it already is running max as speeds get to 90+C)

(I'm amazed it is handling the 9900K as it is with no throttling, frankly)

I'd say comparing it as half the heat dissipation capability is somewhat inaccurate. Sure, it has double the heat spreaders, but both coolers have the same 6 heat pipes and will give much the same average temps.

On another note, I think this is ridiculous the amount of cooling needed for the i9-9900K. I would rather recommend the Ryzen 7 3700X instead.