It's that Time Again - 9900k OC Help... Please.....

[Whiskey11]

Lets just say, my transition from the trusty 4790k at 4.8GHz all core to the 9900k has been... well... less than pleasant. Lets go over the setup:

Case: NZXT Switch 810
CPU: 9900k
MoBo: MSI MEG Z390 ACE
Mem: G.Skill DDR 3466 2x16GB (32GB Total, Model #F4-3466C16D-32GTZR)
GPU: MSI Geforce 2080 Seahawk EK X
Storage: 1x 970 EVO 512gb NVME M.2 (OS Install), a mix of older Samsung 850's and a 1TB HDD
PSU: Zalman 1000W (Have a Corsair 850 around here too)
Cooling setup:
2x240mm rads, Swiftech Apogee CPU Block, EK 2080 waterblock (preinstalled). Front 240mm (Intake) Rad is push/pull, top (exhaust) 240mm rad/res/pump combo is push/pull/pull
MISC: CPU has CoolLabs LiquidMetal Ultra between IHS and waterblock

Goal... well the goal was always simple... try and replicate the 5.0GHz all core that most people are getting, evaluate temperatures and OC as high as the 9900k will go and be stable all core. This machine is a gaming machine, but as I do YouTube videos, the extra cores of the 9900k were what drew me to it over the 9700k which can clock higher, supposedly.

Results:
Not good... Needless to say, all the extra crap in the BIOS of these newer motherboards make me want to go back to my delidded 4790k. I also believe that I lost the silicon lottery hard...

Try #1: 5.0GHz all core. When I first installed everything in the computer and started pushing OC's, I was pushing something like 1.5V Vcore to get it to not BSOD instantly... after growing extremely frustrated, I started looking through my voltages and realized the XMP profile was undervolting the memory from it's spec. I upped the voltage on the DRAM back to 1.350v where the memory is supposed to be and rebooted the computer. LLC was set to setting 2 which on this BIOS, according to the graph, is supposed to be no v-droop (it's actually over volting some, so yay?) All other voltages were set to AUTO, all other settings left at AUTO except the power limits (min and max, as well as current) were maxed out.

Try #2: 5.0GHz all core: 1.350v Vcore Commanded, 1.372 observed - passes Prime95 30 minute test of 1344k, but BSOD after about 5 hours of the 8kmin 4096k max 15min FFT size run. Also wouldn't game either, BSOD frequently.

Try #3: 5.0GHz all core: 1.360v Vcore Commanded, 1.381 observed - BSOD in full P95 at about the 8hr mark. Would still BSOD in games

Try #4: 5.0GHZ all core: 1.375v Vcore Commanded, 1.390 observed - BSOD in full P95 after 15hrs and 16min. BSOD was a Registry_Filter_Driver_Exception Thinking that we must be getting close, I upped the voltage for the next attempt:

Try #5: 5.0GHZ all core: 1.390v Vcore Commanded, 1.411 observed - BSOD in Full P95 after 22hrs and 45min. Seriously, how close to 24hrs of P95 do we need to get? This run did have a WHEA error that was corrected (indicating instability) at around the 17hr mark during a 1440k or 72k test in P95

Try #6: After getting extremely frustrated and knowing how close it is to 24hr P95 stability, I opted to just set the commanded VCore to 1.400v, was seeing 1.432v in HWiNFO64 (which matches BIOS readings). After gaming for an hour and change, BSOD.......................

At no point in any of these tests did the CPU throttle. Distance to TJmax never got closer than 7ºC and that was a spike that lasted for two seconds or so. That spike only happened on two cores. All other cores are about 10-12C from TJmax at 1.400V commanded. Water temps are at or under 10ºC to ambient under full load, highest it got to was just 31ºC with ambient being 22ºC

The reason why I'm here is because I'm thinking I've missed something in the 8 billion new BIOS options since the Z97 MoBo's that is causing the instability and hoping that the knowledge here can help. I have yet to see anyone struggle this hard to get 5.0GHz all core stable without missing a setting and I'm hoping that's the same for me. It'd be nice to see higher than 5.0GHz since a few of the games I play are CPU bound. I've also noticed that even at default settings, the 9900k is not boosting to 5.0GHz, only 4.8GHz. I've also noticed that in the BIOS, the BClock is set to 100.0, yet when I boot into windows, BClock is showing 100.8. I'm guessing due to the XMP profile. Clock speeds listed above are actually about 40-50Mhz low, I just don't want to type it out every time!

Any help would be amazing!

 

[Kenrou]

Registry_Filter_Driver_Exception - WdFilter.sys

My google-fu claims Windows Defender corrupted your GPU drivers somehow - disable WD completely, re-install drivers, and it SHOULD go away. Or re-install Windows.

 

[wingman99]

How does the rig do when the only change from default BIOS is XMP? If it does fine then memory is not the problem it is the 5.0GHz overclock.

 

[Whiskey11]

Registry_Filter_Driver_Exception - WdFilter.sys

My google-fu claims Windows Defender corrupted your GPU drivers somehow - disable WD completely, re-install drivers, and it SHOULD go away. Or re-install Windows.

I can give it a shot but this is the only time I've seen this error.

How does the rig do when the only change from default BIOS is XMP? If it does fine then memory is not the problem it is the 5.0GHz overclock.

It runs just fine gaming. Admittedly, I was lazy and didn't run a P95 test with just XMP. I'm guessing it's the overclock too, just confused as to why 1.400v+ is what it takes when others are nearly a tenth of a volt lower and half a tenth lower on higher clocks. Hence my comment about losing the silicon lottery hard. Legitimately, the number of settings to tweak in the BIOS to get stable OC's was pretty sane a few years ago. Today it's just eye watering how much control they give you. Kinda gives you decision paralysis!

 

[wingman99]

I can give it a shot but this is the only time I've seen this error.



It runs just fine gaming. Admittedly, I was lazy and didn't run a P95 test with just XMP. I'm guessing it's the overclock too, just confused as to why 1.400v+ is what it takes when others are nearly a tenth of a volt lower and half a tenth lower on higher clocks. Hence my comment about losing the silicon lottery hard. Legitimately, the number of settings to tweak in the BIOS to get stable OC's was pretty sane a few years ago. Today it's just eye watering how much control they give you. Kinda gives you decision paralysis!

Lots of options do little for 5.0GHz stability. I would say you lost the voltage lottery. From Siliconlottery As of 12/07/18, the top 41% of tested 9900Ks were able to hit 5.0GHz or greater at 1.300v. Link: https://siliconlottery.com/collections/all/products/9900k50g

 

[Whiskey11]

I hate mobile sometimes...

Tested the RAM using MEMTEST and stock CPU+XMP failed at 1.35v (rated) and 1.4v. Runs fine with default non XMP timings and voltage.

I have down clocked the memory to 3400mhz and the MEMTEST results are showing memory stability. I'm guessing something about the necessary XMP profile that was giving a 100.8 BCLK was causing the issues. Might try the next highest speed eventually but maybe, just maybe, I'll be able to run lower voltages if the stability issues I had were a result of memory issues. Then I'll have to decide if returning this memory is worth the headache or not.

 

[wingman99]

I hate mobile sometimes...

Tested the RAM using MEMTEST and stock CPU+XMP failed at 1.35v (rated) and 1.4v. Runs fine with default non XMP timings and voltage.

I have down clocked the memory to 3400mhz and the MEMTEST results are showing memory stability. I'm guessing something about the necessary XMP profile that was giving a 100.8 BCLK was causing the issues. Might try the next highest speed eventually but maybe, just maybe, I'll be able to run lower voltages if the stability issues I had were a result of memory issues. Then I'll have to decide if returning this memory is worth the headache or not.

Most folks try different memory voltage settings without success. Vccio and Vccsa try increasing Vccio Stock Intel 0.950v Auto to 1.165v Vccsa (System Agent) stock Intel 1.050v Auto to 1.250v. If that does not work I would return for 3200 speed.

 

[Kenrou]

BIOS updated? Mobo specs says it supposedly accepts up to 4500mhz memory.

 

[Whiskey11]

Most folks try different memory voltage settings without success. Vccio and Vccsa try increasing Vccio Stock Intel 0.950v Auto to 1.165v Vccsa (System Agent) stock Intel 1.050v Auto to 1.250v. If that does not work I would return for 3200 speed.

Both VCCIO and VCCSA are at 1.250v in auto..... if the ram is stable at 3400mhz then I'll try going up to one above 3466 just to rule out the BCLK oddity in the XMP profile. I'll try that and if that doesn't work, I'll decide if 3466 or faster is worth the headache over 3400.

- - - Auto-Merged Double Post - - -

BIOS updated? Mobo specs says it supposedly accepts up to 4500mhz memory.

Sorry for the double post.

Yes. I updated the BIOS right after booting to windows the first time when everything was 100% stock. BIOS version date is 11/24/2018 IIRC.

 

[wingman99]

Both VCCIO and VCCSA are at 1.250v in auto..... if the ram is stable at 3400mhz then I'll try going up to one above 3466 just to rule out the BCLK oddity in the XMP profile. I'll try that and if that doesn't work, I'll decide if 3466 or faster is worth the headache over 3400.

BCLK 100.8 is fine for memory. My BCLK goes from 99 to 100.5.

 

[Kenrou]

[emoji23] my FX does 198-202 (100mhz+) either way and it's never really been an issue with stability. Bad hardware/settings on the other hand...

 

[kracht pip]

https://www.gigabyte.com/FileUpload/Global/multimedia/2/file/525/946.pdf

this helped me get to a pretty good 5.1ghz OC. I wont do p95(useless heat generator(degrades chips Id bet)) but i did rog realbench for almost an hour before i got bored of testing did p95 for a while but I do not like the temps it does... we will never hit those workloads to be frank.

 

[larryccf]

i'm only at 4.9 on all cores on my 9900K, with a Vcore of 1.245 and AVX offset of -2, but something that i stumbled on, on my way there, one BSOD, when i ran WhoCrashed (lazy man's way to read dump files) the crash was caused by my virus software conflicting with a windows kernal. I disabled virus software (Emsisoft) for the next P95 test (V26.6 small FFTs) and it ran fine for 3 hours (sorry, i haven't tested for more than 3 hours).

Second suggestions, over on the GeForce forum, there are quite a few threads where the mods indicate quite a few of the Nvidia driver releases are buggy, and they recommend a couple of the known stable drivers - here's one thread, scroll down to the 3rd post. I'm not sure this applies to the 2080 series, but it sure corrected one of my issues (unpredictable black screens) https://forums.geforce.com/default/topic/1035287/nvlddmkm-sys-crashing/?offset=1
One other thing in the GeForce forums i noticed, after un-installing graphics driver using DDU, and installing whichever driver you select, to dis-able Windows 10's abilityu top update drivers

and last, MSI has a pretty decent OC guide for these 9th gen CPUs, and one of the things they suggest is to dis-able all presets (XMP, Turbo boost, Enhanced boost etc)

I'm running the same mobo, by the way,, after having an Asrock Taichi fry my 1st cpu after a bios update, then 2 Gigabyte Z390 boards that BIOS were must not stable - values would, with ever increasing regularity, change settings spontantously

here's a link to the MSI OC guide https://www.msi.com/blog/intel-9th-cpu-overclocking-5ghz-with-z390-motherboards you might find something in there that you missed

Only reason i stopped at 4.9 is because i'm still air cooled and didn't like seeing HWiNFO reporting 92C on a couple of the cores when running P95. When rendering video files, with CPU showing 97-99.7% load, i'm seeing temps in low to mid 70C - i'm setting up a liquid cooling system and hopefully that will give me the temp control to go 5.0+. Before i read that MSI guide, i had trouble getting 4.9 stable at Vcore up to 1.295V.

hope that helps some