Sanity check for 5820k overclock

[ClockBock]

Hi,

I just performed my first cpu overclock and I was hoping I could get a sanity check from some more experienced overclockers on this board.

Setup:
Cpu: 5820k
Mobo: x99-a
PSU: EVGA SuperNOVA 850 G2
Cooler: phanteks tc14pe
Gpu: Gtx 980 ti
Memory: 16 Gb Crucial DDR4 CT-17000 (not sure that is the name though)
Case: Phanteks enthoo primo

I ended up with 4.6 GHz with 1.23 vcore, 1.9 vcinn and idle temps at 37-40 and peaking att 77 c during XTU benchmark (around 70 during stress test). I'm not 100% sure it is fully stable as it seems a fair bit higher than other numbers I have seen, but what a rush! Almost makes we wanna upgrade to watercooling and push on

Questions:

Does this sound reasonable given that I would like the cpu to live a few years?

Is there any point in running the cpu at 4.6 if i don't overclock any other component?

I found the XTU benchmark to be much more reliable in terms of triggering bsods compared to the stress test. I often bsoded immediately when running bench after passing a 20 min stress test. When I began benchmarking before stress I never had a failed stress test after succesful bench. Is this something others also experience or did I run too short stress tests?

I was sort of surprised to not see temps rise alot when i started increasing vcore, given that there are alot of places when people say that watercooling is required to oc haswell. I made a shot for 4.7 and it wasn't until 1.28 i saw a peak above 80 c during benchmark right before bsod. Any chance my temp monitoring is unreliable? I was using HWmonitor.

Is there any good software which can help alerting if something runs too hot? I will mostly be streaming games from the basement to the livingroom so I might not notice if fan speeds good up or if the fan just dies.

One of my ram sticks was broken so I only run with a single 8 Gb stick. I got the replacements today (they even let me keep the two first ones). Any chance I need to retune my settings when I insert the new ones?

End questions.

Long winded story on what I did in case someone would have the kindness to bear with it and see that I haven't screwed up anything too badly:

I basically digested a bunch of guides and forum posts and decided to probe the scaling and vcore stepwise until I found a setting I was happy with, running relatively short stress tests (~10 min) with XTU to check stability in between.

I have the computer in my basement and I don't expect higher ambient temps compared to what i have now (~22 c).

I started by setting vcore to 1.05 and started to increase ratio in steps of 2, running 10 min stress tests between.

At 4.2 I was confident my temps were ok and i rised vcore to 1.1. I think i got to 4.4 before i saw the first bsod after which I began increasing vcore in steps of 0.01.

To get 4.5, i had to go to all the way up to 1.17 and I ended up with 4.6 GHz at 1.22 vcore at ~70 degrees in stress test as the last seemingly stable setting before I called it a night. Somewherr around here did i change vcinn to 1.9 as well.

I tried for 4.7 but gave up when seeing peak temp of 80c right before a bsod.

I then rised the vcore to 1.23 just to be on the safe side in terms of stability. My plan is just to start using the computer as normal and see it is is stable enough and temps stay within my comfort zone (which Im sort of trying to find with this tread).

If not stable or if temps are too high I guess 4.4 seems like a sweet spot since I think it ran stable at 1.1 vcore (might remember wrong though).

Thanks for the attention.

 

[EarthDog]

Voltage for the clockspeed seems pretty darn good. Nice chip you have there... enjoy!

Does this sound reasonable given that I would like the cpu to live a few years?

Is there any point in running the cpu at 4.6 if i don't overclock any other component?

I found the XTU benchmark to be much more reliable in terms of triggering bsods compared to the stress test. I often bsoded immediately when running bench after passing a 20 min stress test. When I began benchmarking before stress I never had a failed stress test after succesful bench. Is this something others also experience or did I run too short stress tests?

I was sort of surprised to not see temps rise alot when i started increasing vcore, given that there are alot of places when people say that watercooling is required to oc haswell. I made a shot for 4.7 and it wasn't until 1.28 i saw a peak above 80 c during benchmark right before bsod. Any chance my temp monitoring is unreliable? I was using HWmonitor.

Is there any good software which can help alerting if something runs too hot? I will mostly be streaming games from the basement to the livingroom so I might not notice if fan speeds good up or if the fan just dies.

One of my ram sticks was broken so I only run with a single 8 Gb stick. I got the replacements today (they even let me keep the two first ones). Any chance I need to retune my settings when I insert the new ones?

1. Absolutely. Memory is not a bottleneck in the first place, nor is cache.
2. Absolutely. Memory is not a bottleneck in the first place, nor is cache.
3. I run the stress test in AIDA64, or I run the stress test in XTU. They both work fine for me.
4. Maybe... try a different application like RealTemp
5. The CPU will throttle if it gets too hot... eventually it will shutdown. I don't keep an eye on temps that much to know of any alerting software outside of what comes from your motherboard. Not sure if that can email you though...(what good is an alarm if the PC is in the basement? )

 

[JrClocker]

If you got 4.6 GHz at 1.23V on the core, you definitely won the silicon lottery!

Water cooling won't necessarily help with the overclock (that's dependent mainly on the chip), but it may help with the temperatures.

Temperature of the CPU is related to how well it cools, and how much power it is putting out. The power dissipation of the processor scales linearly with frequency, and squared with voltage. Said differently:

1. If you increase the frequency by 10%, your power will go up by 1.1 (10%)
2. If you increase the voltage by 10%, your power will go up by 1.1*1.1 = 1.21 (21%)
3. if you do both at the same time, your power will go up by 1.1*(1.1*1.1) = 1.33 (33%)

I don't know of any alerting software either. If you need your PC to be "hands off" stable, I would recommend a bit more testing before you consider yourself good.

At the frequency and voltage of your overclock, your processor alone will be putting out 150W or more...and that's a lot of power.

For stress testing, I want to make sure that my system is 24/7 reliable at the overclock I choose. Others do/feel differently. For reference, here is my test:

1. Set CPU multiplier and voltage (I used fixed...others don't)...if previous setting was bad either decrease CPU multiplier or increase Vcore
2. Run Intel Extreme Tuning Utility processor stress test for 10 minues...CPU good, go to step 3...bad, go to step 1
3. Run Intel Extreme Tuning Utility memory stress test for 10 minutes...CPU good, go to step 4...bad, go to step 1
4. Run HyperPi at 32 M...CPU good, go to step 5...bad, go to step 1
5. Run Prime95 (version 26.6...28.5 or higher will push CPU power over 200 W) for 10 minutes in small FFT mode...CPU good, go to step 6...bad, go to step 1
6. Run Prime95 (version 26.6) for 30 minutes on blended mode...CPU good go to step 7...bad, go to step 1
7. Run SETI@Home at 60% processor and Heaven 4.0 graphics benchmark for 30 minutes...CPU good, go to step 8...bad, go to step 1
8. Hurray!!!

If I make it to step 8, then I have a stable overclock.

Watch your CPU temperatures...I like to have my CPU temperatures under 80 C. (Others accept higher, others like lower). With my current rig and overclock, temps stay under 70 C.

Hope this helps! And welcome to the fun!

 

[tachi1247]

I'd say you found a great chip. I have same processor, cooler and the usb 3.1 version of the same board and I only got to 4.1 @ 1.23v. Couldn't get 4.2 stable at 1.25 and any higher voltage is probably going to result in temps that are too high when ambient increases come July so I stopped trying. Pretty sure my chip is below average.

I also found the xtu benchmark to be a much better tool. I can run the stress test for 3 hours and be fine and then fail in the first 5 seconds under the benchmark.

My temps peak between 70-73 for reference...

 

[ClockBock]

Thanks alot for the replies. Seems I didn't mess up anything too bad then.


It would be great if I won the lottery (I never win anything), but om the other hand I haven't done so any rigorous stresstesting yet. I tried gaming for about one and a half hours yesterday and didn't see any temps above 70 (alt tabbing out every now and then to check temps) so I'm thinking of just using the computer as is for now.*


I hope I don't come out as some arrogant noob who wants to impress the big boys by saying I found the perfect chip or something


I don't need it to be hands off stable, just stable enough to be convenient. The basement thing is a combination of that being my man-cave and that it is off limits for my 2-year old son who has little respect for daddys nice things. With that said I will retune if I ever encounter a bsod during normal operation. Or could there be long term damage if the processor runs at a too low voltage even if it doesn't crash?


About temps, I guess temps just don't suddenly jump up som 10-15 degrees unless ambient temps increase or there is some kind of hw failure. Im thinking of just having realtemp or hwmonitor up an running and check it every now and then, at least initially.


Thanks for providing your temps as a reference. I guess noone knows exactly how the lifetime varies with temps, so it becomes a soft judgement on how important it is to have long life with little possibility to make a quantative assessment right?*


JrClocker, your description of voltage to powet matches my view as well (P = U^2 / R) which is why I was a bit surprised to not see significant *jumps in temps when increasing the voltage. I didn't collect any staistics for comparison though, and temps definitely went up with increasing power. Thanks for providing your method as well. Seems pretty rigid. I might try it to figure out at which setting its is really stable.


Also good to hear that others find the XTU benchmark to be more stressful than the actual stresstest. I was thinking that something might be off in my setup.

 

[wingman99]

I like prime95 test it really helps when you have a long up time and working time with your PC.