Adjusting voltages @ 9900K and its temperature adventures

[SirKeldon]

Hello everybody, it's been a while since I overclocked/tuned the CPU for the last time, kinda like 3-4 years ago, so excuse any possible mistakes.

I had to renew my old system prior to its failure and purchased this 9900K like 2 weeks ago, had to go Intel to avoid any issues when running the Hackintosh that I need to work. Despite having Coffee Lake I've followed the Kaby Lake CPU OC Guide published at ROG's forum and served me quite well, I didn't want to OC too much, just tune the voltages/turbo a little bit cause it was throwing a lot of Auto voltage. I want to share the results with you so at the end, I can see if my temperatures are normal for my cooling system.

I managed to semi-stabilize a 4.9GHz (Sync All Cores @ x49) at 1.24VCore adaptive voltage (1.22-1.21 with VDroop) and AVX offset at 1, stability tests passed:

- 15 min with ROG RealBench Stress (8GB of RAM) - Max Temps: ~65ºC to ~70ºC
- 15 min with ROG RealBench Stress (16GB of RAM) - Max Temps: ~67ºC to ~72ºC
- 10 min with Intel XTU CPU/Memory - Max Temps: ~55ºC to ~60ºC
- 10 min with AIDA64 CPU - Max Temps: ~60ºC to ~63ºC
- 3DMark TimeSpy - didn't record the temps
- 3DMark Fire Strike Extreme Test - Max Temps: ~60ºC to ~62ºC (max in GPU was 54ºC)
- 4xCinebench R20 runs: - Max Temps: ~64ºC to ~68ºC
- Several Geekbench runs

My LLC is at Level 5 (being 7 the most higher offered to me at my MoBo) so I don't heat up too much while at load and acquiring a good balance in VDroop (tested 4 and the drop was too high, used also 7 when determining the voltage at manual mode but temp spikes were really crazy)

- Minimum temperature inside the MoBo ~26ºC to ~29ºC
- CPU tends to idle ~26ºC to ~29ºC and GPU ~28ºC to ~30ºC
- Ambient Temp around 23º to 25ºC

So my two questions are:

1 - Are my temps within a normal range for this kind of CPU with my sort of cooling? Here's a picture of my watercooling loop (EK Quantum Kinetic D5->EK 1080 GPU Block->Swiftech MCR220-XP ->Swiftech Apogee XT->Corsair Hydro R5 Series->back to res)

2 - If my temps are kinda high in your opinion, could it be caused because my block is an oldie Swiftech Apogee XT? I know the disposition is arranged for the same kind of die that lied underneath my previous 2500K & 4790K and now this little 9900K but the pins design can be a little bit obsolete (it's a block from 2009) and specially the extra doubled cores are joining the game ... saw the new designs with fins and more routed paths without that high restriction. Theoretically my block was designed to hold up till a i7-920/950 so 130/40W and that's the limit I set in my BIOS for Max Power Limit. Right now i'm not gonna spend more money since I threw a lot buying the new components but in the upcoming months or 1 year I'll have to change the GPU so I was thinking if to invest into a new CPU block and what would be your recommendations, if that's the problem.

Thanks for the patience to read all this explanation, cheers to everybody

PS: Case is the Thermaltake Core V51 so all meshy on the front, top and bottom so plenty of good airflow. Also, the rear-exhaust fan is now rear-intake fan following a friend's advice, improved my mid-M2 temperatures as well as the RAM and I really hope is helping the VRMs too.

 

[JackCarver]

Temperatures seem good to me for this cpu. My temps are in the same region and I have only a 6 core cpu which produces not that much heat as yours. Mine is also water cooled with an Aqua Computer block.

Edit:
Are those rads two 240 rads?

max in GPU was 54ºC

My RTX 2080 is about 53C here.

 

[SirKeldon]

Temperatures seem good to me for this cpu. My temps are in the same region and I have only a 6 core cpu which produces not that much heat as yours. Mine is also water cooled with an Aqua Computer block.

Edit:
Are those rads two 240 rads?

My RTX 2080 is about 53C here.

Thanks for the response man! Yeah, they're indeed 2x240 rads, GPU is going to an old Swiftech MCR-220XP from 2011 (same date I got the CPU block) and CPU is being cooled at a new Corsair Hydro Series XR5 (purchased along the new components i threw into the build)

I think my GPU could be lower but I screwed it up at the mounting and didn't notice till the end ... that I switched the GPU inlet/outlet, did a quick search on my specific model (MSI Sea Hawk 1080 EK X) and ppl was telling it didn't made a big difference so I decided not to redo it.

The CPU temps were "alerting" me a lil bit too much and decided to share and my thoughts were on the oldie block so thanks again for your input

EDIT: And I just noticed this is my 666th post, kinda funny i'm talking about high temps

 

[JackCarver]

Yeah, they're indeed 2x240 rads

Then with this setup nothing to worry about, I would say. I have also two rads (360/240 from EK) and my temps are nearly the same. In stress testing the cpu with prime95 (Version 26.6 without AVX) about 60-66 degrees, if I use OCCT and AVX/AVX2 then a bit higher so 70-75 degrees, but with AVX2 and small dataset my cpu draws about 190W. GPU mostly stabilizes at about 50-53 degrees in stress testing with 3D mark or Unigine Superposition. In demanding games like Metro Exodus fairly the same.

And I just noticed this is my 666th post, kinda funny i'm talking about high temps

LOL

 

[SirKeldon]

I think RealBench stress test and Cinebench uses AVX instructions cause i saw the downclock to 48x and the other temps seem also about the same.

In demanding games like Metro Exodus fairly the same.

I just have to test the high-demanding games as well, that's what I use 3DMark for 30 mins between bench and stress to test that kind of environment at first, but with work and social life is a no-go, I just played a lil bit Cities Skylines w/everything maxed out at 2560x1080 and ofc the temps were around 45ºC on CPU and 40ºC on the GPU, when I do have time I have to test Path of Exile, Assetto Corsa and GTAV and I'll do an update
Cheers again

 

[DR4G00N]

Yeah, your temps look fine. The Apogee XT is a good block regardless of how old it is.

Since both your cpu & gpu are watercooled I suggest making all the radiator fans exhaust out of the case if they are not already, should bring temps down a touch because it's not recirculating air from one rad to another. It will bring down the temp inside the case also.
I have my rig set up like this and it works well.

 

[JackCarver]

I just have to test the high-demanding games as well

The Witcher 3 is a good game to test your GPU cooling solution, as it is very demanding. And one of the best games also. Firestrike is a nice test to get info if your GPU's VRM cooling is good enough as it heats up the VRM very good. Also the hint to take the rear fan as rear intake fan was a good advice. Overclocker der8auer tested several arrangements of fans/radiators and measured CPU/VRM temps and with a rear intake fan, instead of outtake, the VRM temps went down at about 10 degrees compared to before.

I suggest making all the radiator fans exhaust out of the case

It's not a radiator fan but the rear fan should be intake to get better VRM temps. The best VRM temps would be if the top rad is also intake and front outtake, but if they are all outtake and the rear fan intake, this is also sufficient.

 

[SirKeldon]

Yeah, your temps look fine. The Apogee XT is a good block regardless of how old it is.

Since both your cpu & gpu are watercooled I suggest making all the radiator fans exhaust out of the case if they are not already, should bring temps down a touch because it's not recirculating air from one rad to another. It will bring down the temp inside the case also.
I have my rig set up like this and it works well.

My oldie Swiftie ... I love it, cooled 3 diff generations of these Intel since 2011 and it's still going strong, glad to hear it's a good block even nowadays, new ones with fins though, wouldn't benefit a lil bit more the temps? Or it will be not really noticeable?

Top radiator is in pull configuration exhausting air out while front one is in push getting air in, both are powered by Corsair ML 120 Pro fans, at the bottom there are two Be Quiet Pure Wings 2 pushing air in and the rear fan is same Be Quiet, and as I mentioned is pushing also in. Same friend that recommended that last switch did it to adjust my current setup to produce more possitive pressure and exhaust the most of the heat through the top.

But maybe in the next maintenance i'll try to switch the front ones as well so thanks for the tip!

Also the hint to take the rear fan as rear intake fan was a good advice. Overclocker der8auer tested several arrangements of fans/radiators and measured CPU/VRM temps and with a rear intake fan, instead of outtake, the VRM temps went down at about 10 degrees compared to before.

Kewl!!! I didn't know about those tests, I just followed my friend's advice that I explained to DR4G00N but it's really good to read!!!

 

[DR4G00N]

My oldie Swiftie ... I love it, cooled 3 diff generations of these Intel since 2011 and it's still going strong, glad to hear it's a good block even nowadays, new ones with fins though, wouldn't benefit a lil bit more the temps? Or it will be not really noticeable?

Top radiator is in pull configuration while front one is in push, both are powered by Corsair ML 120 Pro fans, at the bottom there are two Be Quiet Pure Wings 2 pushing air and the rear fan is same Be Quiet, and as I mentioned is pushing also in. Same friend that recommended that last switch did it to adjust my current setup to produce more possitive pressure and exhaust the most of the heat through the top.

But maybe in the next maintenance i'll try to switch the front ones as well so thanks for the tip!



Kewl!!! I didn't know about those tests, I just followed my friend's advice that I explained to DR4G00N but it's really good to read!!!

A newer block would cool better but is it worth $100-150 to to drop the cpu temps by 5 to 7c when your temps are good already? If your 9900K was closer to 90c+ then maybe it would be worth looking into.

 

[SirKeldon]

A newer block would cool better but is it worth $100-150 to to drop the cpu temps by 5 to 7c when your temps are good already? If your 9900K was closer to 90c+ then maybe it would be worth looking into.

Maybe now as my temps seem normal as you both guys say, it's not worth as you mention, but within a year and to be future-proof I think it's not a bad idea to look up for a new CPU block and replace a 10 year-old one, specially if the temp drop is that good, really from 5ºC to 7ºC? I think is a real good improvement (i'm kinda addicted to see lower temps as possible, guilty as charged ) and as I'll probably switch GPU as well, 100 more bucks won't hurt that much. Thanks again! =)

EDIT: One more doubt, I used to run a EK-DDC bay pump the past 5 years ... I went D5 cause I wanted less noise and same kind of performance, I know they're not as powerful regarding head pressure but has more flow in terms of liters per hour, I'm doing tests on 75% and 85% (between 3500 and 4200rpm) depending on the profile, would you keep it that way? Lower?

 

[JackCarver]

Hard to say how much you would benefit from a new CPU block. One thing which gives you definitely better cooling capabilities for your whole loop would be bigger rads. If it's possible to build into your case.

That's my CPU waterblock:

Aqua Computer cuplex kryos NEXT CPU Water Block Review

The cuplex kryos NEXT from Aqua Computer is an extremely customizable water block with features including an integrated display with a 32-bit ARM MCU, the ability to fit the block to your particular CPU, an industry-first physical vapor deposited titanium coating, and excellent performance...

www.techpowerup.com

 

[oxrufiioxo]

Your temps seem pretty good to me...

What I see after 30m of CPU/FPU aida package gets pretty warm but core to core temps aren't that bad...

 

[JackCarver]

EDIT: One more doubt, I used to run a EK-DDC bay pump the past 5 years ... I went D5 cause I wanted less noise and same kind of performance, I know they're not as powerful regarding head pressure but has more flow in terms of liters per hour, I'm doing tests on 75% and 85% (between 3500 and 4200rpm) depending on the profile, would you keep it that way? Lower?

I have also a D5 pump, it runs at 3300 rpm, controlled via the Aqua Suite Software. So yes I would say 75% pump Speed will be enough here.

What I see after 30m of CPU/FPU aida package gets pretty warm but core to core temps aren't that bad...

Yes core temps are good and the iCue Software is really nice here. Didn't know that it can also monitor DRAM and GPU temps

 

[oxrufiioxo]

Yes core temps are good and the iCue Software is really nice here. Didn't know that it can also monitor DRAM and GPU temps

Yeah, I like it works really well on my Ryzen system as well. Only issue with it I have is it bugs out if you try to run other monitoring software with it.

 

[SirKeldon]

Your temps seem pretty good to me...

What I see after 30m of CPU/FPU aida package gets pretty warm but core to core temps aren't that bad...

Thanks, third good temps, that kinda confirms the block is still holding good, also pump speed as the circuit disposition is doing a good job.

Yup, your temperatures are pretty similar to the ones I've got after 30 mins of RealBench which also stresses FPU AFAIK, I'll try more tests tomorrow though. Are you using AVX offset btw? Even if the answer is no, what VCore are you dropping down while performing intensive FPU/AVX instructions?

I have also a D5 pump, it runs at 3300 rpm, controlled via the Aqua Suite Software. So yes I would say 75% pump Speed will be enough here.

Mine is controlled via Q-Fan at the BIOS but yup, 75-80% is the sweet spot at least now.

 

[oxrufiioxo]

Yup, your temperatures are pretty similar to the ones I've got after 30 mins of RealBench which also stresses FPU AFAIK, I'll try more tests tomorrow though. Are you using AVX offset btw? Even if the answer is no, what VCore are you dropping down while performing intensive FPU/AVX instructions?

No offset, voltage fluctuates from 1.26-1.28 my min stable vcore but I am pretty sure I got a decent chip it does 5.1 at 1.33v avx but temps get out of control in the 90c range.


My ambient is around 19-20C while benchmarking btw even an increase of 3-4c makes a noticble impact on temps.

 

[GorbazTheDragon]

Tried lowering the LLC a little?

Generally adding LLC doesn't increase stability much above the first few settings as LLC only pulls up a constant voltage droop, if the current draw is very erratic the LLC cannot compensate fast enough.

Should drop you a few degrees...

 

[SirKeldon]

Tried lowering the LLC a little?

Generally adding LLC doesn't increase stability much above the first few settings as LLC only pulls up a constant voltage droop, if the current draw is very erratic the LLC cannot compensate fast enough.

Should drop you a few degrees...

I've tried but the temps were basically the same, besides that, the LLC4 gave me some really extra VDroop that with LLC5 I managed to stabilize, it's just -0.02V diff when on hard-load, LLC6 and LLC7 are both overshooting.

---

I've also waited some more days before posting again to see if I'm stable while working (basically intensive JS development on VSCode and several projects running on Node as well as comm software -teams, skype, slack-, spotify and A LOT of browser tabs ) and the mild gaming experience that I'm still demanding from this beast, just some CS:GO, Cities Skylines and the most GPU intensive one, Age Of Empires II HD

All of this worked well, both on MacOS and Windows, and I did some more intensive stability tests and they're also running pretty well (still working to get AVX offset 1 working, I'm getting best stable results at AVX2 though) ... but the best part it was to follow the advice DR4G00N pointed, I converted my front rad fans intake into exhaust ones and the whole machine is running fresher ... the temps in the long sessions (stress testing/working/gaming) did improve quite a lot not just in the CPU/GPU department but also in the PCH/MoBo/M2 one, so i'm really glad, thanks a lot man!!!