i9 9900K at 5Ghz running hot

[Julebrusen]

I bought my new setup a week ago.

 

Processor: Intel Core i9-9900K

MB: ASUS ROG STRIX Z390-F GAMING, S-1151

RAM: Corsair Vengeance LPX DDR4 3600Mhz 16GB

Cooler: Corsair Hydro H115i RGB Platinum

 

Running 5G setup from BIOS to get 5Ghz.

 

Had a Corsair H100x first, and temps during gaming was around 73degrees, and Cinebench R20 temps were around 91-93, with no thermal throttling.

Since this I bought a H115i to get better cooling on the OC setup. When installing the H115i i got idle temps in windows around 45 degrees.

And when running cinebench, CPU went to up to 100 degrees(seen in Intel ETU) before I aborted.

 

I removed the thermal compound that came with the cooler and added Noctua NT-H1, with same temperatures.

I put the old cooler back on, and it now has the same higher temperatures as the H115i.

 

Is there something im missing or is the 5G setup in the BIOS wrong?

 

 

 

[Julebrusen]

CPU stress test in ETU has temps at 70 degrees, with slight thermal throttling

[Princess Luna]

Well you motherboard is notorious for being awful at overclocking something like the i9... and if you overclocked with voltages set on auto it is very likely you're pumping too much electricity for your AiO to keep up with, VRM's may also be suffering hard as they don't have as much air flow going through them.

 

You have bought both a bad motherboard and a bad cooling solution (yeah AiO's are just overpriced and only serve for the aesthetic appeal) for a good 5ghz OC.

 

What you can do is ensure the AiO's functioning right, that it's pump is not dead, "install" it properly with a good thermal paste and then OC your processor manually with fine tuning of the power states and voltages.

[Julebrusen]

Awesome!

I tried using the AI tool on the MB first with varying results. So i tried the 5G profile in the BIOS, and it ran with no thermal throttling at 93 in cinebench.

Just weird that the bigger cooler should get it hotter.

 

I do not have the skills to OC my system all by myself yet, so guess im going back to 3.6Ghz.

[Sorenson]

 

4 minutes ago, Julebrusen said:

Awesome!

I tried using the AI tool on the MB first with varying results. So i tried the 5G profile in the BIOS, and it ran with no thermal throttling at 93 in cinebench.

Just weird that the bigger cooler should get it hotter.

 

I do not have the skills to OC my system all by myself yet, so guess im going back to 3.6Ghz.

From my experience the overclocking profiles in the bios will jack up the voltages to get the clock speed. I personally never raise the stock voltages and have had great results overclocking on intel. I recommend enabling sync all cores or MCE (what ever is called on your motherboard) and manually setting the boost clock starting at like 47. Run a 20 minute stress test if it works bump it up to 48 and so on. Once you have a frequency you are happy with that isn’t running too hot just leave it. Keep in mind that the difference between an all core boost of 4.7ghz and 5.0 is very small, so don’t over do it if you’re new to overclocking.

[Julebrusen]

I will try this, thanks for the help. 

 

I just find it weird it ran good before i changed coolers. Max temp i had was 93 in cinebench and no thermal og power throttling. 

[Julebrusen]

I loaded default settings in bios, restarted and tried the boost clock at 48 and XMP 1 for the RAM. Stresstest got me 85 degrees, and the cinebench maxed out at 96 degrees.

I still think this is too high, but no thermal or power limit throttling. I'll try it here and see if it is unstable.

 

But since the problem is now with both coolers, is there anything i could have broken or bent?

All pins in connectors to fans and pump is good and all are connected. Tightened the pump to the CPU in a cross pattern and all.

 

Baffled

[drago95]

2 hours ago, Julebrusen said:

I loaded default settings in bios, restarted and tried the boost clock at 48 and XMP 1 for the RAM. Stresstest got me 85 degrees, and the cinebench maxed out at 96 degrees.

I still think this is too high, but no thermal or power limit throttling. I'll try it here and see if it is unstable.

 

But since the problem is now with both coolers, is there anything i could have broken or bent?

All pins in connectors to fans and pump is good and all are connected. Tightened the pump to the CPU in a cross pattern and all.

 

Baffled

Maybe and only maybe you need to set the votage manually...............unless you need a bios update

[Tristerin]

Do you have an air bubble trapped in the pump/blocks?

 

Also, an AIO is going to struggle with that.  I went from an H100i V2 to 600mm of rad space to contain the temps on my OC.

[Julebrusen]

Might be, need to read up on OCing. Bios is up to date.

 

Airbubble sounds like a possible thing. Im guessing rad to the top so air gets to that? 

[Turtle Rig]

5 hours ago, Princess Luna said:

Well you motherboard is notorious for being awful at overclocking something like the i9... and if you overclocked with voltages set on auto it is very likely you're pumping too much electricity for your AiO to keep up with, VRM's may also be suffering hard as they don't have as much air flow going through them.

 

You have bought both a bad motherboard and a bad cooling solution (yeah AiO's are just overpriced and only serve for the aesthetic appeal) for a good 5ghz OC.

 

What you can do is ensure the AiO's functioning right, that it's pump is not dead, "install" it properly with a good thermal paste and then OC your processor manually with fine tuning of the power states and voltages.

What is the TJ MAX on this thing ?

[Mister Woof]

Probably 100 like the rest

[xg32]

1 hour ago, Julebrusen said:

.

i'm actually surprised you got it to run at 5ghz with that board and a h115i, asus z390 are notorious, the vrm is probably hitting 100C too along with the cpu, I'm more worried about that than the cpu itself, it'll just shut off, i know @Princess Luna already pointed this out, but it's REALLY bad, i can't stress that enough, i give the motherboard about 18-24months. It's preferred if you point a fan directly at the vrm. I'm gonna guess it's running at 100-110C~.

 

The h115i starts struggling above 135w or so,, i had to really tinker with it even on a 8600k, if ur 9900k really is running 5ghz all core, it'd be around 175w. 

I suspect AIOs have a slight flaw in that they can't move the heat fast enough on small dies, (they are fine for GPUs), it's really just for aesthetics.

[K0NG]

On 9/13/2019 at 1:56 AM, xg32 said:

The h115i starts struggling above 135w or so

what?
the Cooling is not the problem.
the terribly thick die (more than twice as thick as a 8700k) and the very thick solder on top makes the CPU that hot.)

my 8086k ran at 5.4 Ghz at 1.48v for benching and it was delidded.
my h115i in the balanced mode (like 50% pump speed) kept the CPU at below 70°C (average) in Aida64 Extreme.  (CPU Power draw was at around 190W at the time)

delidding and sanding down the die makes an 9900k 10-15°C cooler with ease.

[Zberg]

Stock BIOS profiles notoriously run way too much voltage and super hot.  My mobo's stock 5.0 all core causes CB runs to hover right around 100 degrees (and vCore spikes to 1.41 a lot).  My manual OC at 5.0 all core (currently locked at vCore 1.34)  hovers in the mid to high 80's with stress testing (noctua D15).  A manual OC would probably do much better.

[Zberg]

7 hours ago, K0NG said:

what?
the Cooling is not the problem.
the terribly thick die (more than twice as thick as a 8700k) and the very thick solder on top makes the CPU that hot.)

my 8086k ran at 5.4 Ghz at 1.48v for benching and it was delidded.
my h115i in the balanced mode (like 50% pump speed) kept the CPU at below 70°C (average) in Aida64 Extreme.  (CPU Power draw was at around 190W at the time)

delidding and sanding down the die makes an 9900k 10-15°C cooler with ease.

This is bad information.  Check the gamers nexus video.  Not only is it more difficult than delidding non soldered processors with higher risk to dmg the die, the gains are also less.  I think they got 2-3 C in their tests.  Have seen others with repeated similar results (never very significant, certainly not 10-15 C).  The risk benefit on the 9900k makes it very not worth.  

[Falkentyne]

There is very minimal risk as long as you use a proper delid tool with a guarantee (Rockit guarantees their delid tool won't damage your processor or they will replace it).  Rockit 89 or Der8auer tools work well.  The chips are machine soldered (you can tell by how all samples have the same solder removal leftover pattern after a delid), so there is very little variance.

 

However what only a few delid guides tell you is that after delidding, you *MUST* also sand down the edges of the IHS underside!  About 0.2mm will do.  That's because after the solder is removed, and even after the Intel silicone is completely removed (required), the die->IHS z-height is going to be larger since the solder accounted for this (even despite the Intel silicone), which will cause sub-optimal contact and not enough pressure.  Sanding the IHS edges down 0.2mm, and then following this up (important too!) by buffing (not polishing) both the underside of the IHS and the CPU die with 1500 to 2000 dry sandpaper!  Why?  Liquid metal LOVES a flat, but non smooth surface--it greatly improves adhesion and ability to spread. mirror finish or polished surfaces are very bad for liquid metal.

 

Then relid and you'll get at least an 8C improvement.

[PowerBaller]

Ruling out all the thermal interface problem, it really depends on your voltage setting, which is actually the outcome of your silicon lottery.

 

If you are lucky enough to draw a good chip that can stabilize at 5.0 Ghz below 1.285 V (assuming you are not achieving this by tuning BLK frequency), it shouldn't be this hot , not even while you are testing the small FTTS in prime95. Since your motherboard is an Asus, their AI suite has a function to predict how good the chip is, ( shown as "Setup Parameter"). For a 9900K, a good chip is usually above 95%, the best i ve seen is 132%.

 

Find the minimum bootable voltage, increase it by 0.02 with TPU sight, and start vary the voltage  by small increment until you find the sweetspot, where your Prime95 small LTTS wont lose half of the threads immediately. This will be a reference state for your future tuning.

 

Power delivery wise, Z390 F may not be able to feed an OC 9900K well beyond 5.0Ghz

[Turtle Rig]

1 hour ago, PowerBaller said:

Ruling out all the thermal interface problem, it really depends on your voltage setting, which is actually the outcome of your silicon lottery.

 

If you are lucky enough to draw a good chip that can stabilize at 5.0 Ghz below 1.285 V (assuming you are not achieving this by tuning BLK frequency), it shouldn't be this hot , not even while you are testing the small FTTS in prime95. Since your motherboard is an Asus, their AI suite has a function to predict how good the chip is, ( shown as "Setup Parameter"). For a 9900K, a good chip is usually above 95%, the best i ve seen is 132%.

 

Find the minimum bootable voltage, increase it by 0.02 with TPU sight, and start vary the voltage  by small increment until you find the sweetspot, where your Prime95 small LTTS wont lose half of the threads immediately. This will be a reference state for your future tuning.

 

Power delivery wise, Z390 F may not be able to feed an OC 9900K well beyond 5.0Ghz

Yes Linus peeps always talked about this how ASUS vrms sux for 9900k OCing to 5Ghz.  I knew all this so when we got my old man his machine I got a Gigabyte Aorus z390 and off boot up its already to go to 5ghz depending on load, without touching BIOS.  It finds the memory incorrect at 2133 but that is fixable in BIOS.  That is only setting.  She runs great.

[PowerBaller]

Just now, Turtle Rig said:

Yes Linus peeps always talked about this how ASUS vrms sux for 9900k OCing to 5Ghz.  I knew all this so when we got my old man his machine I got a Gigabyte Aorus z390 and off boot up its already to go to 5ghz depending on load, without touching BIOS.  It finds the memory incorrect at 2133 but that is fixable in BIOS.  That is only setting.  She runs great.

Aorus stacks quite a bit of their cost in CPU power delivery, for the same price compare to Asus , MSI and EVGA. But we all know Gibabyte bios are hideous, as well their memory controller. A 1.40V is not able to stabilize a B Die Trident Z at 3600Mhz on Aorus master, while the same setting on maximus Xi hero stabilize them at 4000Mhz CL16.

 

I am avoiding Gigabyte motherboards for better OC experience. ( Asus' T-Topology OC 4 ram module better, while MSI and assumed EVGA's Daisy Chain OC better with 2 ram , at Dimm 2 and 4.)

[Zberg]

1 hour ago, Falkentyne said:

There is very minimal risk as long as you use a proper delid tool with a guarantee (Rockit guarantees their delid tool won't damage your processor or they will replace it).  Rockit 89 or Der8auer tools work well.  The chips are machine soldered (you can tell by how all samples have the same solder removal leftover pattern after a delid), so there is very little variance.

 

However what only a few delid guides tell you is that after delidding, you *MUST* also sand down the edges of the IHS underside!  About 0.2mm will do.  That's because after the solder is removed, and even after the Intel silicone is completely removed (required), the die->IHS z-height is going to be larger since the solder accounted for this (even despite the Intel silicone), which will cause sub-optimal contact and not enough pressure.  Sanding the IHS edges down 0.2mm, and then following this up (important too!) by buffing (not polishing) both the underside of the IHS and the CPU die with 1500 to 2000 dry sandpaper!  Why?  Liquid metal LOVES a flat, but non smooth surface--it greatly improves adhesion and ability to spread. mirror finish or polished surfaces are very bad for liquid metal.

 

Then relid and you'll get at least an 8C improvement.

Guarantee on a tool only counts for the delid not fucking up the CPU.  If you fuck it up sanding down the die, all bets are off.  It introduces a significant risk above and beyond normal delidding, and as GN showed, the gains were very small.  Would not recommend for anyone not very comfortable with this (aka 90% of the tech community)

[Falkentyne]

3 hours ago, Zberg said:

Guarantee on a tool only counts for the delid not fucking up the CPU.  If you fuck it up sanding down the die, all bets are off.  It introduces a significant risk above and beyond normal delidding, and as GN showed, the gains were very small.  Would not recommend for anyone not very comfortable with this (aka 90% of the tech community)

90%?

Perhaps you mean young kids, but some people have been doing hardware mods since the days of soldering clock crystals.

And by the way, no need to make a straw man argument with me.

I never even mentioned sanding *down* the die in my post!

 

I said that you have to BUFF The die (roughen it up a little so its not shiny) with 2000 grit sandpaper (which anyone can do with their fingertips) and the IHS underside, not SAND it.

I did say that you DO need to sand the IHS edges however to compensate for the solder removal.

 

Might want to re-read my post again.

[K0NG]

17 hours ago, Zberg said:

This is bad information.  Check the gamers nexus video.  Not only is it more difficult than delidding non soldered processors with higher risk to dmg the die, the gains are also less.  I think they got 2-3 C in their tests.  Have seen others with repeated similar results (never very significant, certainly not 10-15 C).  The risk benefit on the 9900k makes it very not worth.  

i never said that someone should delid and sand it down.

but a proper die sanding (about 0,1mm) does decrease the temperature massively.

[Zberg]

17 hours ago, Falkentyne said:

90%?

Perhaps you mean young kids, but some people have been doing hardware mods since the days of soldering clock crystals.

And by the way, no need to make a straw man argument with me.

I never even mentioned sanding *down* the die in my post!

 

I said that you have to BUFF The die (roughen it up a little so its not shiny) with 2000 grit sandpaper (which anyone can do with their fingertips) and the IHS underside, not SAND it.

I did say that you DO need to sand the IHS edges however to compensate for the solder removal.

 

Might want to re-read my post again.

I can re-read it until the cows come home.  Many pro's have tried their hands at this and gotten unimpressive results with the 9900k.  Whether or not every tech lover can do what you and I can, it involves risk, and the gains are just not impressive.  And to your comment, again the tools are guaranteed only for the delidding as I said.  Once you do anything else, whether its buffing, sanding (or whatever else you want to do that isnt delidding, pick a verb), whatever you do to alter it, it aint covered any more.  It was a misleading comment.

 

Other guys quoted 10-15, you said 8 easy.  Fact is that isnt the case.  You are incorporating time and risk for a few degrees probably, maybe 5.  Not worth.   

 

If it was actually a 10 degree decrease that would be one thing, but its not.  

[K0NG]

On 9/12/2019 at 8:16 PM, Princess Luna said:

(yeah AiO's are just overpriced and only serve for the aesthetic appeal)

 

 

My h115i beats my NH D15 by 15°C at the balance mode with low fan/pump RPM.

absolutely ridiculous assumption