Overclock.net banner

1 - 7 of 7 Posts

·
Registered
Joined
·
3 Posts
Discussion Starter #1 (Edited)
After a few days of iterative testing I believe I've finally achieved a 5.0 all core overclock on my i9 10850k. I haven't seen any well documented stable 10850k overclocking, as many people claim they can run 5.2ghz only running Cinebench and not perform any real stress testing. I originally used Prime 95 small FFT AVX disabled to test but it would take 30-40 minutes to find errors that OCCT SSE would find in 10-15.

Hardware:
i9 10850k (integrated graphics for OC testing)
Z490 Gigabyte Vision G
Seasonic PX-850
garbage Corsair 3600mhz 18-22-22-42 (while my G. Skill Samsung B-die is in shipping, XMP not enabled)
Arctic Freezer II 280
Thermal Grizzly Conductonaut
Bequiet! 500DX (front intake, top/rear exhaust)

BIOS Settings;
MCE: Disabled
Ring Ratio: 43x
Vcore Voltage Mode: Adaptive
VF Point Offset Mode: Legacy
Internal CPU Vcore Offset: -.185v
BCLK Adaptive Voltage: Disabled
VAGX: 1.200v
DRAM Voltage 1.350v
VCCIO: 1.310v
VCCSA: 1.310v
LLC: Extreme
Vcore Current Protection: Turbo
Hyper Threading, Speed Shift, Thermal Monitor, EIST, RTH, Turbo Boost, Turbo Boost 3.0. Flex Ratio: Enabled
Down Bin: Disabled
Energy Efficient Turbo, Voltage Optimization: Disabled
Flex Ratio: 36x
Frequency Clipping, Voltage reduction initiated TVB: Disabled
Active Turbo Ratios: Enabled
All Core Turbo Ratio: 50x
C-States Control: Auto
Turbo Power Limits, Turbo Per Core Limit Control: Disabled
Any settings not specifically listed: Auto

Results;
0 Errors OCCT 7.3.2 Small SSE (non AVX) 1hr
VCore: 1.33v
VROut: 1.26v
Package Power: 212w
Current: 184A
Package Temp: 94*c
VRM Temp: 80c

In my finding increasing VCCIO/VCCSA was required in order to achieve 0 errors otherwise L0 Cache errors occur. Setting VCCIO/VCCSA @ auto drew much more power than needed or cause fluctuations I believe would create errors. Tinkering with Fixed Vcore voltage mode drew nearly 80w more package draw than Adaptive regardless of Vcore voltage and LLC.

It seems 5.1ghz is pretty unrealistic for me because of thermal headroom. I'm not sure how well my chip places in the silicon lottery and my motherboard/cooler is limiting any more performance gains. It took forever to find what values I needed to change because of outdated forum threads filled with outdated (pre 8th gen) or completely incorrect information.

I used Hardware Lens' and Actually Hardcore Overclocking's videos for guides as they provide (somewhat) explanations to which settings are responsible for what on the motherboard.

The goal of my OC was to be efficient as possible. My cooler was less capable than I thought even after reseating as slight vcore offsets would increase temperatures into thermal throttling. Suggestions are very welcome because 5.1ghz all core would be great and I'm not super experienced in overclocking. thanks for reading
 

·
Premium Member
Joined
·
9,827 Posts
@ 1.3v and Prime with AVX off my cores still throttle. I have same chip and cooler, but not using conductonaut. Used MX 4 that came with cooler. Which test did you run?
 

·
Registered
Joined
·
3 Posts
Discussion Starter #3
I tested using OCCT 7.3.2, small data pack, SSE. I just ran Prime 95 small FFT non AVX and it never actually throttled
2479248
 

·
Just hangin' out
Joined
·
1,309 Posts
Custom loop.
 

·
Registered
Joined
·
887 Posts
If you look at Silicon Lottery's binning statistics for Comet Lake (bottom of the page) you'll see only the top 6% of the 10850k chips they tested could do 5GHz all core non-AVX stable, at least with 1.17v true load vcore (Die sense). I mean if you think about it, these essentially were supposed to be 10900Ks but couldn't meet the standards.

Historical Binning Statistics – Silicon Lottery
 

·
Registered
Joined
·
3 Posts
Discussion Starter #6
If you look at Silicon Lottery's binning statistics for Comet Lake (bottom of the page) you'll see only the top 6% of the 10850k chips they tested could do 5GHz all core non-AVX stable, at least with 1.17v true load vcore (Die sense). I mean if you think about it, these essentially were supposed to be 10900Ks but couldn't meet the standards.

Historical Binning Statistics – Silicon Lottery
Makes sense to release a slightly lower spec for rejects since it's still a strong chip. Neat I was able to get a good one, I'm gonna see how I can go to cheat out a Cinebench score. Hindsight I should've tracked the prices a little closer because a 10900k is only $50 more, but it was fun getting to 5.0ghz
 

·
Registered
Joined
·
187 Posts
I must have gotten lucky then. I'm at 5.1 at 1.35v load. I ran it for an hour on OCCT small, SSE and had no errors. I'm using a MF360 AIO cooler and it can stay below 80C during stress testing when my AC is on. My 8700K only does 4.8 so I guess my luck evened out.
 
1 - 7 of 7 Posts
Top