Intel 8086k OC Stress Test & Core Voltage question

[Shadowciaw]

hey guys, i've been trying to pull a 5Ghz oc on my 8086k recently but i have some questions about stress tests and core voltages. 

 

so ive been following this tutorial: https://overclocking.guide/gigabyte-z370-overclocking-coffee-lake/ on overclocking the cpu and using prime95 to stress test it. I followed their instructions and started out at 1.27V, then as I found that the core utilization wasnt all hitting 100%, i'd stop the test and increment the core voltage by .01. i'm sitting at 1.31V right now, and I keep seeing people getting by just fine with 1.3V 5GHz. I was talking to my friend earlier and he recommends me using Asus realbench to stress test instead of p95. I havent downloaded it yet but I realized i can try another stress test that i currently have on my pc

 

i got intel xtu a few weeks ago when i was too scared to do bios level ocing and i set the clock ratio to 5 every time i booted up my pc, and thats all I did. i stress tested briefly and it passed just fine. now that i have the cpu set to 5ghz and 1.31v in the bios, i stress tested with intel xtu and it passed no problem with all cores at 100% (hitting about 85 degrees occasionally tho) i only stressed with xtu for 5 minutes, but i stressed with p95 for even shorter bc when i saw that not all cores were at 100% i stopped the test and assumed it wasnt stable. 

 

what should i do guys? can i assume it's fine and leave it be? should i bring the voltage up to the maximum of 1.35V and test again?? idk if my cooler (cryorig h7) would be good enough for it (i was building on a budget, so i couldnt go for a d15 or a decent aio :c) what stress test is more preferred? should i trust p95 over xtu stress test? what do you guys recommend for stability testing?

 

all my bios settings are the same as in the tutorial, i have the loadline calibration set to turbo because i would blue screen at 5ghz 1.27v. 

 

i just want to keep a safe and stable 5ghz oc :c should i just settle for something lower instead or change the avx settings? thanks guys

[xg32]

prime blend is a very basic cpu/memory stress test that a build should be able to pass for 1 hour (don't need to run fft), however a very small % of the time, you can pass blender/realbench 1hour and still fail prime. (this is very unlikely to happen)

 

if you blue screen at 1.27v then 1.31v can possibly be not stable.

 

I recommend keeping ur existing oc, and use an avx offset of 1, and see if you can run prime at 4.9 instead while most games will run at 5ghz

[Shadowciaw]

Just now, xg32 said:

prime blend is a very basic cpu/memory stress test that a build should be able to pass for 1 hour (don't need to run fft), however a very small % of the time, you can pass blender/realbench 1hour and still fail prime. (this is very unlikely to happen)

 

if you blue screen at 1.27v then 1.31v can possibly be not stable.

 

I recommend keeping ur existing oc, and use an avx offset of 1, and see if you can run prime at 4.9 instead while most games will run at 5ghz

firstly, thank you for the response. (also, my mobo is gigabyte z370 aorus gaming 5 if that affects anything)

 

i havent stress tested in p32 for an minute, much less an hour. From the wording of the tutorial, I thought that if CPU utilization (im watching coretemps) isnt all at 100% immediately and for the next few seconds, it means that the overclock is unstable. that's why i shut off the stress test a few seconds in because i saw that only some cores were hitting 100%.

 

and to clarify, I bluescreened because loadline calibration was set to auto. once i set it to turbo, i booted just fine with 1.27v. 

 

I ran the "small FFTs" test on p95 as per the tutorial and didnt have all cores at 100%. however, if i selected the blend preset, the stress test seemed more normal (with all cores on 100% and temps of 60-70 instead of some cores 100% and those getting to 90+). what's going on? whats the difference between the two presets that causes this discrepancy? 

 

sorry for so many questions, but should i stick to the p95 blend preset instead of "small ffts" preset to stress test? why did the author of the article choose to use the latter preset and why does it cause such a weird result while monitoring compared to the blend preset?

 

here's a picture of the blend test core monitoring 

 

and here's a picture of the small ffts test

[xg32]

7 minutes ago, Shadowciaw said:

 

don't use small fft, it is a burn test that generate loads that are impossible to replicate in real world applications, and it's blue screening due to hitting 100C, i'd suggest an hour of p95 blend test (i failed this at multiple cpu oc and narrowed it to a memory instability), if that's stable you are likely good at 1.31v, i wouldn't suggest pushing further due to motherboard vrm and cpu cooler limitations.

 

To clarity use p95 blend (not fft) for cpu/mem stability, and blender/realbench for real world temps (u won't hit these temps while gaming, but it's a good gauge, try to keep the blender/realbench temps under 90C at all times.

[Shadowciaw]

1 minute ago, xg32 said:

don't use small fft, it is a burn test that generate loads that are impossible to replicate in real world applications, and it's blue screening due to hitting 100C, i'd suggest an hour of p95 blend test (i failed this at multiple cpu oc and narrowed it to a memory instability), if that's stable you are likely good at 1.31v, i wouldn't suggest pushing further due to motherboard vrm and cpu cooler limitations.

 

To clarity use p95 blend (not fft) for cpu/mem stability, and blender/realbench for real world temps (u won't hit these temps while gaming, but it's a good gauge, try to keep the blender/realbench temps under 90C at all times.

ah i see! thank you so much. have a great night