[Arctucas]

1. I setup everything manually AND it works on 8700K, BUT don`t work on 9900K
2. I explained that I setup 1.05 and 1.15 V (the same what I have on 8700K), also increased these values to 1.25V – nothing change!
3. I didn`t mention that I use XMP, I wrote “3733mhz with all custom timings (all timings changed regarding formulas, 100% stable system)” where is here XMP?

I don`t use XMP, I don`t activate XMP profile, I setup everything manually: frequency, voltages, timings etc.

Even more: I wanted setup from the scratch on 9900K, but more than 3400Mhz I can`t get.

OK, just a suggestion to find a starting point.

Good luck, hope you figure it out.

 

[PsyenceFiction]

1. I setup everything manually AND it works on 8700K, BUT don`t work on 9900K
2. I explained that I setup 1.05 and 1.15 V (the same what I have on 8700K), also increased these values to 1.25V – nothing change!
3. I didn`t mention that I use XMP, I wrote “3733mhz with all custom timings (all timings changed regarding formulas, 100% stable system)” where is here XMP?

I don`t use XMP, I don`t activate XMP profile, I setup everything manually: frequency, voltages, timings etc.

Even more: I wanted setup from the scratch on 9900K, but more than 3400Mhz I can`t get.

For the beginning - maybe you could start being less toxic when asking people in this forum for help.

Regarding your problem:
yes, it is surely possible. High memory overclocks can destabilize your CPU Ring clock and every CPU is different. It is possible to determine from the nature of "crashes" / when you get BSOD what is probably the cause.
Usually my experience is if you press hard-reset button on your case during this BSOD and after that it is doing nothing for a while, then suddenly PC shuts off and starts automatically again then it is probably unstable CPU ring clocks.
And you need to bump up the VCCIO voltage more or lower the CPU Cache/Ring ratio enough. If it restarts immediately without "shutting off between" then it is unstable Core clock = push more Vcore.
Be aware - VCore also affects Ring voltage, so if your 9900K runs on lower voltage then your previous 8700K then there is less voltage combined for the Ring than before and thus possibly unstable OC.

Last thing which comes to my mind is the 1T/2T Command Rate (CR). It is quite possible that the second CPU is not able to run 2T timings so smoothly as the first CPU.
When I tested some momory OC, I could reach stable overclocks at 1T CR with 7700K while on 8700K with even better board it was not able to run 1T even with considerably higher voltages. 2T was possible though.

 

[HoneyBadger84]

sorry didnt' realize this isn't a general 9900K thread, please deleto lol

 

[{EAC} Shoot em UP]

No, besides the fact that no P0 steppings were ever binned for 9900KS, which is why R0 steppings tend to overclock worse, but can you do what I mentioned earlier please?

In the Gigabyte Bios, set AC Loadline and DC Loadline to the lowest non-zero value (1), set your CPU to 5 ghz, and boot into windows and check the CPU VID shown in HWInfo at idle.
On Asus Bios, set "SVID Behavior" to "best case scenario", and set AC and DC Loadline to Auto. (this does the same thing as the GB Bios--it sets the AC and DC Loadlines to 0.01 mOhms).

This must be done at 5 ghz.
And check the VID (not vcore). This helps tell the silicon quality.

Hey Falkentyne! Question for ya, a buddy is selling me his 10700k for a steal, I can basically flip my z390/9900k for a z490 Vision G/10700k and make 10 bucks. I figure why not, give me something to do and give me an upgrade path.
That said, my 9900k always seemed to have good voltage characteristics, but man oh man it has horrible temp issues.

Anyways, once I get the 10700k system up and running, does this sort of thing still tell silicon quality? Or is there a new way to test? Basically, I’m back to the “overclocking world” starting in a few weeks on the z490 platform and am looking where to start. Also, any idea if RAM VCCIO and SA voltages would be “similar” from 9900k to 10700k? My ram isn’t very fast. 3000Mhz 15-17-17-35 (iirc?) that I run at 3600 18-19-19-36? It’s been a while, I forget and am not home to confirm. Just wondering what I should expect I suppose.

Thanks for the help with the 9900k, been at 5ghz (even though it runs HOT on custom water) for almost a year now.

 

[Falkentyne]

Hey Falkentyne! Question for ya, a buddy is selling me his 10700k for a steal, I can basically flip my z390/9900k for a z490 Vision G/10700k and make 10 bucks. I figure why not, give me something to do and give me an upgrade path.
That said, my 9900k always seemed to have good voltage characteristics, but man oh man it has horrible temp issues.

Anyways, once I get the 10700k system up and running, does this sort of thing still tell silicon quality? Or is there a new way to test? Basically, I’m back to the “overclocking world” starting in a few weeks on the z490 platform and am looking where to start. Also, any idea if RAM VCCIO and SA voltages would be “similar” from 9900k to 10700k? My ram isn’t very fast. 3000Mhz 15-17-17-35 (iirc?) that I run at 3600 18-19-19-36? It’s been a while, I forget and am not home to confirm. Just wondering what I should expect I suppose.

Thanks for the help with the 9900k, been at 5ghz (even though it runs HOT on custom water) for almost a year now.

Sorry I can't really help with this. And I can't help at all with SA/IO.

But the stock VID, when AC Loadline and DC Loadline are both set to 0.01 mOhms (0.01 in Asus, "1" in Gigabyte VR Settings), with c-states/power saving DISABLED, at idle in windows, are a good APPROXIMATION of quality--the lower the VID, the better the chip in general, but not every chip performs the same. There are some "SP 63" chips that overclock as well as SP 90 chips, and some SP 80 chips that are worse than SP63 (Even though the VID may be lower). It's an average approximation. You still have to do the work to test it manually.

 

[Arctucas]

Sorry I can't really help with this. And I can't help at all with SA/IO.

But the stock VID, when AC Loadline and DC Loadline are both set to 0.01 mOhms (0.01 in Asus, "1" in Gigabyte VR Settings), with c-states/power saving DISABLED, at idle in windows, are a good APPROXIMATION of quality--the lower the VID, the better the chip in general, but not every chip performs the same. There are some "SP 63" chips that overclock as well as SP 90 chips, and some SP 80 chips that are worse than SP63 (Even though the VID may be lower). It's an average approximation. You still have to do the work to test it manually.

Do you have any advice for settings on the eVGA Z390 Dark regarding the 'quality' testing?

 

[Falkentyne]

Do you have any advice for settings on the eVGA Z390 Dark regarding the 'quality' testing?

No. I don't own that board.

 

[{EAC} Shoot em UP]

Sorry I can't really help with this. And I can't help at all with SA/IO.

But the stock VID, when AC Loadline and DC Loadline are both set to 0.01 mOhms (0.01 in Asus, "1" in Gigabyte VR Settings), with c-states/power saving DISABLED, at idle in windows, are a good APPROXIMATION of quality--the lower the VID, the better the chip in general, but not every chip performs the same. There are some "SP 63" chips that overclock as well as SP 90 chips, and some SP 80 chips that are worse than SP63 (Even though the VID may be lower). It's an average approximation. You still have to do the work to test it manually.

SP 63? Is this the stepping? Batch? Sorry don’t even have the 10700k in hand yet and have done no real research behind reaching out here. And no prob, I’ll hit up the ddr4 OC thread for info there for SA/IO. Thanks!

 

[tty]

Hi guys,

I have a 9900k and a gb z390 gaming sli.
I oc the cpu to 5GHz all core no avx offset using:

--------------------------------------------------------------------------------------------

1.295v , LLC High, IA LLC auto, IA Loadline 0, DC Loadline 0
SVID reading 1.4v , 1.308v vcore hwinfo

Blender 2.9 render:
185W load, 1.265v vcore hwinfo

-------------------------------------------------------------------
Second test with same settings just changed IA Loadline 1 DC loadline 1
SVID reading 1.21v, 1.308v vcore hwinfo

Blender 2.9 render:
158W load, 1.265v vcore hwinfo

Which one is the realistic setting since with auto (0) IA/DC Loadline cpu package shows 25W more on load also a huge difference in SVID ?

Edit: I accidentally left DC Loadline to 0 which was messing up the readings.
I set that to 1 now and the CPU package Watts is close to what it was before.

Was weird because I had almost the same temps when load was displayed 158W vs 185W.

 

[anonymousse37]

Hello,

Trying to OC my 9900KF.
I've made a short linpack test on OCCT 5.5.6 (10 min).
This is an OC for 5Ghz @ 1.26v.

• Maximus Formula XI
• AIO ASUS ROG STRIX LC 360
• 4x8GB Corsair Vengeance RGB 3200
• PSU Corsair RM1000
• Corsair C700P

I noticed that Vcore range was

• Min 1.22v
• Idle 1.26v
• Max 1.27v

So, on a load, it drops from 1.26v to 1.22v which is a 0.04v vdrop.
Does it tell something specific ?

I have to set the VCore in the BIOS @ 1.28v to get 1.26v in CPU-Z.
I dont really get why there is such a gap.

What do you think about the curves ? Good, bad ?

Thank you. Kind regards.

 

[Hawkjoss]

Hello,

Trying to OC my 9900KF.
I've made a short linpack test on OCCT 5.5.6 (10 min).
This is an OC for 5Ghz @ 1.26v.

• Maximus Formula XI
• AIO ASUS ROG STRIX LC 360
• 4x8GB Corsair Vengeance RGB 3200
• PSU Corsair RM1000
• Corsair C700P

I noticed that Vcore range was

• Min 1.22v
• Idle 1.26v
• Max 1.27v

So, on a load, it drops from 1.26v to 1.22v which is a 0.04v vdrop.
Does it tell something specific ?

I have to set the VCore in the BIOS @ 1.28v to get 1.26v in CPU-Z.
I dont really get why there is such a gap.

View attachment 2484301 View attachment 2484302 View attachment 2484303

What do you think about the curves ? Good, bad ?

View attachment 2484300

Thank you. Kind regards.

Hard to tell based on only the info you provided.

However, you got an error during your linpack so it is unstable. And if I read it correctly, you are hitting 100C at peak tests, which is pretty "warm" . Use HWIfo for vitals monitoring - it does the job the best.

 

[Arctucas]

Hello,

Trying to OC my 9900KF.
I've made a short linpack test on OCCT 5.5.6 (10 min).
This is an OC for 5Ghz @ 1.26v.

• Maximus Formula XI
• AIO ASUS ROG STRIX LC 360
• 4x8GB Corsair Vengeance RGB 3200
• PSU Corsair RM1000
• Corsair C700P

I noticed that Vcore range was

• Min 1.22v
• Idle 1.26v
• Max 1.27v

So, on a load, it drops from 1.26v to 1.22v which is a 0.04v vdrop.
Does it tell something specific ?

I have to set the VCore in the BIOS @ 1.28v to get 1.26v in CPU-Z.
I dont really get why there is such a gap.

View attachment 2484301 View attachment 2484302 View attachment 2484303

What do you think about the curves ? Good, bad ?

View attachment 2484300

Thank you. Kind regards.

I know the newer versions of OCCT use HWiNFO as monitoring software, not sure about v5.

Anyway, the difference in VCore under load is 'VDroop' or voltage droop. Completely normal. LLC is used to offset it.

All in all, not bad, but better cooling would help with temps.

 

[anonymousse37]

Thanks for the replies, so i tried something else :

I tried at fixed Vcore to modify LLC then AVX to see what happens ( See attached image )
The Higher LLC will always provide a higher Vcore while idle and the Vdroop will be smaller than using a lower LLC
The Lower LLC will always provide a lower Vcore while idle and the Vdroop will be bigger

I'm not really sure what VID means in Core TEMP and if it is reliable. I added this row just to experiment the values. Doesnt know if it helps or not understanding something.

There is something i dont understand.
We always try to minimize the Vcore for a given frequency in order to minimize heat.
So, the lower the vcore is, the better will be the temperatures on idle.
So, why during a load, the VCore is decreasing and the températures are increasing. Why is the VCore not increasing instead ?
Why when we lower the VCore in the BIOS, the temperatures are devreasing and why under load, the VCore is deacreasing as well and temperature are getting higher.
Intel says Vcore max is 1.45v for this chip right ?
Does it means that 1.45v is the limit no to get overshoot

I've heard that SVID on Asus motherboard were bad to report Vcore.
I mean, is it normal to set a Vcore = 1.275v in the bios and reading IDLE Vcore = 1.243v on CPU-Z or HWINFO ( report the same values ) with LLC5 - ( MAX is LLC8, never tried )
Its a 32mV difference. With LLC=7, the difference is still 15mv.
Usually, on tutorial, i can relate a 0.01mv with medium LLC. So i'm a bit confused.

Reminder :

Motherboard : Asus Formula XI BIOS 1502 ( I read this one suxs, i might want to update, can seomeone relate ? )
CPU : 9900KF
AIO : ROG STRIX LC 360

Kind regards

 

[Hawkjoss]

Thanks for the replies, so i tried something else :

I tried at fixed Vcore to modify LLC then AVX to see what happens ( See attached image )
The Higher LLC will always provide a higher Vcore while idle and the Vdroop will be smaller than using a lower LLC
The Lower LLC will always provide a lower Vcore while idle and the Vdroop will be bigger

I'm not really sure what VID means in Core TEMP and if it is reliable. I added this row just to experiment the values. Doesnt know if it helps or not understanding something.

There is something i dont understand.
We always try to minimize the Vcore for a given frequency in order to minimize heat.
So, the lower the vcore is, the better will be the temperatures on idle.
So, why during a load, the VCore is decreasing and the températures are increasing. Why is the VCore increasing ?
Why when we lower the VCore in the BIOS, the temperatures are devreasing and why under load, the VCore is deacreasing as well and temperature are getting higher.
Intel says Vcore max is 1.45v for this chip right ?
Does it means that 1.45v is the limit no to get overshoot

I've heard that SVID on Asus motherboard were bad to report Vcore.
I mean, is it normal to set a Vcore = 1.275v in the bios and reading IDLE Vcore = 1.243v on CPU-Z or HWINFO ( report the same values ) with LLC5 - ( MAX is LLC8, never tried )
Its a 32mV difference. With LLC=7, the difference is still 15mv.
Usually, on tutorial, i can relate a 0.01mv with medium LLC. So i'm a bit confused.

Reminder :

Motherboard : Asus Formula XI BIOS 1502 ( I read this one suxs, i might want to update, can seomeone relate ? )
CPU : 9900KF
AIO : ROG STRIX LC 360

Kind regards

RE:your question about voltage
Voltage is only a “fuel” for cpu. At idle it can be even 1.52 and do no damage to the cpu and temos will be relatively low - because there is no LOAD. As soon as you throw the cpu under stress, current and power draw (respectively) increases, but Voltage decreases.
ohm’s laws will help you with understanding the processes

 

[ThinbinJim]

There is something i dont understand.
We always try to minimize the Vcore for a given frequency in order to minimize heat.
So, the lower the vcore is, the better will be the temperatures on idle.
So, why during a load, the VCore is decreasing and the températures are increasing. Why is the VCore not increasing instead ?
Why when we lower the VCore in the BIOS, the temperatures are devreasing and why under load, the VCore is deacreasing as well and temperature are getting higher.
Intel says Vcore max is 1.45v for this chip right ?
Does it means that 1.45v is the limit no to get overshoot

I've heard that SVID on Asus motherboard were bad to report Vcore.
I mean, is it normal to set a Vcore = 1.275v in the bios and reading IDLE Vcore = 1.243v on CPU-Z or HWINFO ( report the same values ) with LLC5 - ( MAX is LLC8, never tried )
Its a 32mV difference. With LLC=7, the difference is still 15mv.
Usually, on tutorial, i can relate a 0.01mv with medium LLC. So i'm a bit confused.

Reminder :

Motherboard : Asus Formula XI BIOS 1502 ( I read this one suxs, i might want to update, can seomeone relate ? )
CPU : 9900KF
AIO : ROG STRIX LC 360

Kind regards

Temperature rises due to increased current drawn. Power draw in CPUs come from two sources - dynamic power and leakage power.
The one we're interested in is switching power, which is the equation:
Switching Power = (switching activity) * (frequency) * (effective capacitance) * (voltage)^2
Under load, switching activity increases dramatically which causes increased current drawn and results in more heat. In contrast, idle has nearly zero activity due to the amount of power gating that modern processors have, and multiplying any number by a number close to zero results in a near-zero value (i.e. minimal power drawn from switching power).

Vdroop is a separate phenomenon that is regulated by the VRMs. Read one of shamino, elmor or falkentyne's posts that go in-depth about intel load lines.

Also, Intel doesn't state any specific "safe" voltage as safe voltage varies by current draw and temperature. See ryzen's boosting behavior - it pushes high voltage under low load and steps down voltage/frequency when under a power virus like prime95. The 1.45v is likely a recommendation by someone not employed by Intel.

Lastly, the maximus xi lineup is almost universally hated by reviewers, but I've used the Maximus XI Hero and z390 Aorus Master and strongly prefer using the Hero. User friendliness for memory OCing and the AI OC predictions are pretty neat features. It's not worth it to invest in a new z390 motherboard anyway as the best options are difficult to get new (boards like the XI Apex and Aorus Xtreme), and z390 is a dead platform.

 

[anonymousse37]

I take an example of my benchs

CPU Vcore set in BIOS = 1.275v and LLC6 AVX0
So this specific Vcore is what i understand to be the VID ?

CPU Vcore after booting an OS

• Idle = 1,252v
• Load(MAX) = 1,172v / VID = 1.33v ( Read on HWInfo )

I understand that the load applied is represented by Vdroop ( With LLC compensating ) = 1.252 - 1.172 = 0.08v. = R*I with ( I = Intensity (Amp) representing the strengh of the load )

On HWInfo, i can read VID = 1.33v when it has been set to 1.275 in the bios.
Does it mean i'm getting an overshoot ?

I have no clue in electronics, sorry if what i'm saying is complete none sense :S

 

[Arctucas]

VCore and VID are two different things.

Others can provide a more detailed explanation, but I understand them as;

VCore equals what the Voltage Regulator supplies to the CPU after VDroop, whereas VID is what the CPU is programmed to request from the Voltage Regulator at specific frequency.

Do you have a HWiNFO screenshot you can share showing the voltages and such?

 

[anonymousse37]

VCore and VID are two different things.

Others can provide a more detailed explanation, but I understand them as;

VCore equals what the Voltage Regulator supplies to the CPU after VDroop, whereas VID is what the CPU is programmed to request from the Voltage Regulator at specific frequency.

Do you have a HWiNFO screenshot you can share showing the voltages and such?

Tell me if this is what you ask for and if i should remake with others values ?

 

[Arctucas]

VCore and VID are two different things.

Others can provide a more detailed explanation, but I understand them as;

VCore equals what the Voltage Regulator supplies to the CPU after VDroop, whereas VID is what the CPU is programmed to request from the Voltage Regulator at specific frequency.

Tell me if this is what you ask for and if i should remake with others values ?

Thanks.

The graphs really are not necessary. Just the Current/Min/Max/Avg columns would be sufficient.

Here is what I normally use with HWiNFO. I know there is a lot of information, and a lot of it not relative to the conversation, but I did not feel like saving another HWiNFO profile at the moment. I boxed the VID and VCore to show an example of what I get while running CB R23.

I have the following set in the UEFI:
VCore - 1.210V Adaptive with a -75mV offset
AVX offset - 0
VDroop (LLC on your motherboard) set to Default, which is the Intel spec. I am not sure what number you would use for the same.

 

[Arctucas]

Sorry I can't really help with this. And I can't help at all with SA/IO.

But the stock VID, when AC Loadline and DC Loadline are both set to 0.01 mOhms (0.01 in Asus, "1" in Gigabyte VR Settings), with c-states/power saving DISABLED, at idle in windows, are a good APPROXIMATION of quality--the lower the VID, the better the chip in general, but not every chip performs the same. There are some "SP 63" chips that overclock as well as SP 90 chips, and some SP 80 chips that are worse than SP63 (Even though the VID may be lower). It's an average approximation. You still have to do the work to test it manually.

I cannot set my loadline to .01, but found a way to get it to .001.

Optimized defaults loaded in UEFI, C-states disabled, Windows power plan set to High Performance.

Opinions?

Thanks

EDIT:

With EIST/Turbo disabled: