9900K @ 5Ghz 1.2V guide Gigabyte Z390 Master

Soldato
Joined
18 Dec 2008
Posts
6,307
Location
Liverpool
I do, I'm just a little worried that I may have to up the dynamic vcore up some more to get stability as I'm at +0.030. I am a complete novice with all this and although I get the chips are tough etc I obviously want to get the temps down as low as possible.

Do you think it would have any benefit going push/pull with the radiator mounted up top? Just a friend mentioned the 3 high speed silent wings 3's don't have the pest static pressure and I do have 3 spare pure wings 3's that I swapped out on the rad when I got it.
 
Soldato
Joined
18 Dec 2008
Posts
6,307
Location
Liverpool
Euuuuurghhhh just had to add another +0.005 to the Dynamic Vcore, been solid for days and then blue screened on ebay! Doh!

Event viewer shows a kernel power error.
 
Soldato
Joined
19 Apr 2003
Posts
13,513
Euuuuurghhhh just had to add another +0.005 to the Dynamic Vcore, been solid for days and then blue screened on ebay! Doh!
Yep, your addicted/afflicted - whichever you want to call it.

You'll be setting up your custom loop before then end of the year :D
 
Soldato
Joined
18 Dec 2008
Posts
6,307
Location
Liverpool
Lol yeah! It's funny cause I can already see how this is going to play out... So I'll get the basics for a cpu loop, with an extra rad (for future proofing) I'll think I can make it look prettier with angled fittings, I'll track down a water block for my Gigabyte GeForce GTX 1080Ti Gaming OC BLACK. I'll want some RGB, maybe some dye to change the look of the loop. Oh and lets not forget custom sleeved cables to go with my new PSU. Then when it's done, it'll be hard line then maybe even glass while I try to push 5.1 or maybe even 5.2 lol and so on and so on.
 
Soldato
Joined
3 Oct 2007
Posts
12,085
Location
London, UK
Lol yeah! It's funny cause I can already see how this is going to play out... So I'll get the basics for a cpu loop, with an extra rad (for future proofing) I'll think I can make it look prettier with angled fittings, I'll track down a water block for my Gigabyte GeForce GTX 1080Ti Gaming OC BLACK. I'll want some RGB, maybe some dye to change the look of the loop. Oh and lets not forget custom sleeved cables to go with my new PSU. Then when it's done, it'll be hard line then maybe even glass while I try to push 5.1 or maybe even 5.2 lol and so on and so on.

Yep that is where I was 8 months ago. Did my first a custom loop on just my CPU in an old Coolermaster Cosmos tower, was immediately addicted and over Xmas built this lol. I've just updated from that pic with the Master and 9900k. Good job I don't have kids or they'd be living on baked beans while I feed my addiction :D

IMG-8176.jpg
 
Soldato
Joined
19 Apr 2003
Posts
13,513
Lol yeah! It's funny cause I can already see how this is going to play out... So I'll get the basics for a cpu loop, with an extra rad (for future proofing) I'll think I can make it look prettier with angled fittings, I'll track down a water block for my Gigabyte GeForce GTX 1080Ti Gaming OC BLACK. I'll want some RGB, maybe some dye to change the look of the loop. Oh and lets not forget custom sleeved cables to go with my new PSU. Then when it's done, it'll be hard line then maybe even glass while I try to push 5.1 or maybe even 5.2 lol and so on and so on.
Look forward to your project log - sounds like you've got it mapped out completely in your head already.

Yep that is where I was 8 months ago. Did my first a custom loop on just my CPU in an old Coolermaster Cosmos tower, was immediately addicted and over Xmas built this lol. I've just updated from that pic with the Master and 9900k. Good job I don't have kids or they'd be living on baked beans while I feed my addiction :D
Kudos, especially considering it was your second ever custom loop.
 
Soldato
Joined
18 Dec 2008
Posts
6,307
Location
Liverpool
I'm getting there, just trying to decide whether a 420 and 280 radiator will actually fit :D

Would have liked a mono block for the board but doesn't seem like anyone makes them. Doing a lot of reading and watching youtube videos at the moment.
 
Associate
Joined
2 Jun 2019
Posts
1
I had to make a account to thank you! I have been following gigabytes guide to 5ghz and it was stable at 1.32 but it was pumping out a lot of heat! So I did your guide and had to add +.030 and I think I am stable it goes up to about 1.29 under load maybe a touch higher and the temps are so much better! In prime95 small fft it’s 5c better and in cinebench its 10c cooler after 3 runs and also has a higher cinebench score! Super happy I am hoping it is stable. More testing to come. At 5ghz I got a cinebench score of 2212 with other apps closed and cinebench set to high priority. Pretty happy with these settings. This should be more known cause it makes the 9900k way easier to control. Power output is lower and temps at about a similar voltage as a static overclock. Have you found any new tips. I have the aorus master
I might try for 5.1 and see what it needs



I feel when it comes to overclocking, people think they should increase as many settings as possible, which produces more heat from the motherboard and CPU, which is not great, especially for a 24/7 OC.

I use an Offset and speed shift for when I don't need all the performance (Typically when web browsing, so this will also help with electricity costs, and also prolonging the life of your CPU, and your Motherboard VRMs, Phases, etc.

So most guides will show you to use a set voltage and to turn off any power management, as well as running high currents and high load line calibrations etc.

So by doing this, my 9900K @5Ghz would be stable at 1.26v and will hit up to 85c+ with AIDA stress test using a AIO H100i. - then the motherboard VRMS hitting high temps, and generally stressing the system.

Now we will do the opposite here, and help keep everything cool and quiet.

Adjust all the settings like so.

(VAXG Phase Control only has options: Auto - High Performance - Extreme Performance.)

Fy3BV46.png.jpg

Here we will set the CPU core to normal which will unlock the Dynamic Vcore setting. (Depending on your CPU, you may need to increase by +0.010v till your system is stable.

I have mine set at 0.000v

Ltr1hRY.png.jpg

Set the CPU ratio to 50
Set your Dram speeds to what your kit is, unless you know what is a stable overclock for your Dram
Then we will head into (Advance CPU Core settings).

ONWY1an.png.jpg

Scroll down to near the bottom and enable Intel Speed Shift technology (Speedstep)

KBKkFy1.png.jpg

Save and exit. Now run a stress test.

Stable at 5Ghz with 1.2V - VRMS not peaking 40C - So we have a nice solid cool n quite overclock, which is more efficient than the stock settings with turbo.

HGIf3cb.png.jpg

And when you don't need all that power, the CPU frequency and Voltage drop off.

VIlXhz4.png

You're welcome. :)
 
Associate
Joined
7 Apr 2006
Posts
292
Location
USA
Just so you guys are aware, the OP's settings of "cool and quiet" have nothing to do with needing a lower stable voltage during stress tests at 5 ghz. Neither do any of the power limiting options have any effect either.
Assuming you guys are correctly monitoring VR VOUT (do not pay attention to the Vcore 1 and 2 sensors), the reason for being able to run at a lower VR VOUT than before (usually around 20 to 40mv lower) is directly from the decrease in loadline calibration. A tighter (higher) LLC, which gives less vdroop (less *sustained* voltage drop at load) has a drawback of causing higher transient undershoots and overshoots at load also, and the more current/power you pull, the larger those will be. The undershoots are what wind up dictating your actual minimum voltage you need at load. At a lower LLC, the undershoots are less, so your VR VOUT you see will be closer to your true minimum you need. Note that undershoots and overshoots happen in microseconds and cannot be registered on VR VOUT; you can look at buildzoid's recent oscilloscope loadline video as well as Elmor's recent posts on the OCN Z370/Z390 thread to see what they look like.

The Internal Load Line preset of "Power Saving" simply sets the AC and DC values to 0.4 mOhms and 1.3 mOhms. The DC loadline is only used for power reporting via VID (does not affect actual CPU supply voltage), e.g. VID * Amps=CPU Package Power.
AC Loadline is the CPU Supply voltage, which uses a baseline of the CPU's default VID (at a certain multiplier, which stops scaling at the highest 2 core turbo ratio), and then is biased boosted up by the AC Loadline mOhms resistance value. This does not affect overshoots or undershoots and has nothing to do with "Loadline calibration." This total biased setting is then sent to the VRM as the "target voltage", and then that target voltage is then dropped via vdroop based on the Loadline Calibration setting.

You can see the original target voltage via VID, if you set DC loadline manually to 0.01 mOhms (Value 1). This will remove "VID" droop from VID and show the VRM value. VR VOUT will show the value after VRM Vdroop is applied (Loadline calibration).

Note that very few CPU's will be stable with zero offsets with Auto/Normal vcore and an AC Loadline of 0.4 mOhms at 5 ghz, and LLC set to low. This requires a very good CPU sample. And is almost guaranteed to fail prime95 small FFT (29.8 build 3-5) with AVX.
 
Associate
Joined
14 Jan 2013
Posts
77
Just so you guys are aware, the OP's settings of "cool and quiet" have nothing to do with needing a lower stable voltage during stress tests at 5 ghz. Neither do any of the power limiting options have any effect either.
Assuming you guys are correctly monitoring VR VOUT (do not pay attention to the Vcore 1 and 2 sensors), the reason for being able to run at a lower VR VOUT than before (usually around 20 to 40mv lower) is directly from the decrease in loadline calibration. A tighter (higher) LLC, which gives less vdroop (less *sustained* voltage drop at load) has a drawback of causing higher transient undershoots and overshoots at load also, and the more current/power you pull, the larger those will be. The undershoots are what wind up dictating your actual minimum voltage you need at load. At a lower LLC, the undershoots are less, so your VR VOUT you see will be closer to your true minimum you need. Note that undershoots and overshoots happen in microseconds and cannot be registered on VR VOUT; you can look at buildzoid's recent oscilloscope loadline video as well as Elmor's recent posts on the OCN Z370/Z390 thread to see what they look like.

The Internal Load Line preset of "Power Saving" simply sets the AC and DC values to 0.4 mOhms and 1.3 mOhms. The DC loadline is only used for power reporting via VID (does not affect actual CPU supply voltage), e.g. VID * Amps=CPU Package Power.
AC Loadline is the CPU Supply voltage, which uses a baseline of the CPU's default VID (at a certain multiplier, which stops scaling at the highest 2 core turbo ratio), and then is biased boosted up by the AC Loadline mOhms resistance value. This does not affect overshoots or undershoots and has nothing to do with "Loadline calibration." This total biased setting is then sent to the VRM as the "target voltage", and then that target voltage is then dropped via vdroop based on the Loadline Calibration setting.

You can see the original target voltage via VID, if you set DC loadline manually to 0.01 mOhms (Value 1). This will remove "VID" droop from VID and show the VRM value. VR VOUT will show the value after VRM Vdroop is applied (Loadline calibration).

Note that very few CPU's will be stable with zero offsets with Auto/Normal vcore and an AC Loadline of 0.4 mOhms at 5 ghz, and LLC set to low. This requires a very good CPU sample. And is almost guaranteed to fail prime95 small FFT (29.8 build 3-5) with AVX.

Then how the hell do we get a stable adaptive voltage OC...
Christ i've been doing this all week and no matter what, as soon as I pop open P95... I get a failure eventually.
TBH I'm thinking I'll just be using realbench and cinebench from now on, I won't ever be stressing my cpu as much as p95 does and it is beggining to feel like a pointless endevour.
 
Associate
Joined
2 Dec 2017
Posts
45
Location
Leeds
Sorry to bring a old post up.

The new F10 bios as been redesigned for the Aorus Master and it looks weird any one messed with it to get 5ghz like on the old bios?

Lance
 
Associate
Joined
28 Sep 2018
Posts
2,242
@Falkentyne good to see you on here. I moved to offset from fixed turbo llc and p95 non avx went from average of 1.265 to 1.215. What a difference. 10 less amps and 15less watts of heat (current iout and power pout)
 
Associate
Joined
17 Nov 2019
Posts
3
Hello guys.

I have the same board with F9 bios but just can't get 5.0 ghz stable with Noctua D15. I followed few youtube how to's tutorials including the gigabyte manual but it just seems that d15 on 5ghz is not adequate cooling solution. None the less I keep my I9 stable on all cores 4.7 ghz, 1.23V. To be honest I didn't even try other frequencies, it was 5 ghz or nothing. Idle temps are low 30's, gaming not passing 65 degree celsius. The closest I got to 5.0 ghz was 1.29V with AVX offset 3, but than it crashed in aida 64 benchmark, while in prime was pumping 93 degrees and within few hours some workers stopped working and crashed. With more voltage I passed 100 degrees and stopped the test. I didn't want to cook my I9. Tests done: Realbench 8 hours straight, no crash. Prime 95 with avx stable for 8 hours, not passing 92 celsius. OCCT also 4 hours, not a single crash. Intel burn test few hours, no crash. Aida 64 for several hours,also no crash. My question is next for the experienced overclockers, will I see any difference between 4.7ghz and 5ghz in gaming performance or general day to day PC performance? I mean it's only 300mhz difference. Is it worth to push that 300mhz from 4.7 to 5.0 for day to day use and gaming with additional temperature increase?

Thank you.
 
Associate
Joined
28 Sep 2018
Posts
2,242
Hello guys.

I have the same board with F9 bios but just can't get 5.0 ghz stable with Noctua D15. I followed few youtube how to's tutorials including the gigabyte manual but it just seems that d15 on 5ghz is not adequate cooling solution. None the less I keep my I9 stable on all cores 4.7 ghz, 1.23V. To be honest I didn't even try other frequencies, it was 5 ghz or nothing. Idle temps are low 30's, gaming not passing 65 degree celsius. The closest I got to 5.0 ghz was 1.29V with AVX offset 3, but than it crashed in aida 64 benchmark, while in prime was pumping 93 degrees and within few hours some workers stopped working and crashed. With more voltage I passed 100 degrees and stopped the test. I didn't want to cook my I9. Tests done: Realbench 8 hours straight, no crash. Prime 95 with avx stable for 8 hours, not passing 92 celsius. OCCT also 4 hours, not a single crash. Intel burn test few hours, no crash. Aida 64 for several hours,also no crash. My question is next for the experienced overclockers, will I see any difference between 4.7ghz and 5ghz in gaming performance or general day to day PC performance? I mean it's only 300mhz difference. Is it worth to push that 300mhz from 4.7 to 5.0 for day to day use and gaming with additional temperature increase?

Thank you.

Simple answer is that P95 avx small fft will pull about 180amps on a 9900k at 5ghz. Your D15 simply can't handle that.

OCCT large/avx2 is a good test for a couple of hours. x264 stress test 10+ loops is also good.
 
Last edited:
Associate
Joined
17 Nov 2019
Posts
3
Simple answer is that P95 avx small fft will pull about 180amps on a 9900k at 5ghz. Your D15 simply can't handle that.

OCCT large/avx2 is a good test for a couple of hours. x264 stress test 10+ loops is also good.

Does that mean that my OC at 5 ghz is not stable if it doesn't handle P95 avx small fft with D15 and is it worth upgrading to AIO liquid cooling solution for additional 300mhz push from 4.7 to 5.0 to see any noticeable gain in performance?
 
Back
Top Bottom