10850k won’t overclock. Temps too high D15

LLC on auto. I am down to 1.20ghz with a 15 minute realbench run stressing 16gb ram out of 32gb.

What I am finding is that whatever vcore I set in bios it’s is always more under load. Eg. Vcore 1.220 becomes 1.230 under load. Is that normal or even right?

I thought that if I could run Realbench at 1.200vcore for 6 hours then I would be fine. What I have been reading is that the system may crash when gaming even if stable at 1.2vcore under load with Realbench.

The main reason for the under volt was to lower heat and power draw. 1.310 stock is just dumb and Realbench and Prime95 (won’t use it again now) just hammered the system at stock. 100c on some cores in under 10 seconds even with an aggressive fan curve and -2 avx offset.

m
 
Last edited:
Lael Traylor said:
What I am finding is that whatever vcore I set in bios it’s is always more under load. Eg. Vcore 1.220 becomes 1.230 under load. Is that normal or even right?
Click to expand...
No mine is opposite, if I set 1.30v, its using 1.293v under 100% load according to cpu-z and checked in Dragon Centre.
Are you running latest MSi bios?
 
Yeah cpuz and Hwmonitor both read 0.01 higher that vcore and 0.02 in some areas. 1.200vc and readout under load is 1.220 or 1.221....


Good try hardwareinfo. Is there a monitoring system I can turn on?
 
Lael Traylor said:
LLC on auto. I am down to 1.20ghz with a 15 minute realbench run stressing 16gb ram out of 32gb.

What I am finding is that whatever vcore I set in bios it’s is always more under load. Eg. Vcore 1.220 becomes 1.230 under load. Is that normal or even right?

I thought that if I could run Realbench at 1.200vcore for 6 hours then I would be fine. What I have been reading is that the system may crash when gaming even if stable at 1.2vcore under load with Realbench.

The main reason for the under volt was to lower heat and power draw. 1.310 stock is just dumb and Realbench and Prime95 (won’t use it again now) just hammered the system at stock. 100c on some cores in under 10 seconds even with an aggressive fan curve and -2 avx offset.

m
Click to expand...

Don't run Auto.

Buildzoid did prob testing recently. You want mode 3 for llc on msi:

unknown.png
 
Was just driving to the store and I figure it out... MSI default under auto to either LLC mode 1 or mode 2. That was pushing the voltage over the vcore. I know the papers say LLC does increase voltage over set vcore but CPUZ, HWMONITOR and HWINFO all showed the same amount within 0.001 of each other. I changed LLC to 4 and low and behold the voltage is now 1.176 under load and vcore 1.200.

I will NEVER buy MSI again the stock settings on this board are awful.

Fail at LLC 4 and 1.2 volts.
 
Lael Traylor said:
Was just driving to the store and I figure it out... MSI default under auto to either LLC mode 1 or mode 2. That was pushing the voltage over the vcore. I know the papers say LLC does increase voltage over set vcore but CPUZ, HWMONITOR and HWINFO all showed the same amount within 0.001 of each other. I changed LLC to 4 and low and behold the voltage is now 1.176 under load and vcore 1.200.

I will NEVER buy MSI again the stock settings on this board are awful.

Fail at LLC 4 and 1.2 volts.
Click to expand...

My temps at 5.2ghz and 1.373v using cpu-z stress test - LLC Auto
J0yYsuY.jpg
 
Hmm, I have an MSI Z490 MEG Unify and 10700k and I've been having temp issues under a full custom loop. My CPU can easily hit 30 degrees above my GPU which is in the same loop when under a full stress test. I may check out the voltages as up until now I've been blaming the CPU block.
 
ChrisD. said:
Hmm, I have an MSI Z490 MEG Unify and 10700k and I've been having temp issues under a full custom loop. My CPU can easily hit 30 degrees above my GPU which is in the same loop when under a full stress test. I may check out the voltages as up until now I've been blaming the CPU block.
Click to expand...
If you’re running that stock you can certainly afford to bump down the voltage. I haven’t done a full six hour test, I’m going to wait until I get the CPU reseated just in case there is an issue with the initial paste that I used. Once the CPU is reseated I will run a six hour benchmark on realbench.

if it’s stable then at 1.225vcore I’ll be happy. Temps have dropped 10c under full load. My CPU is running super hot for that voltage so I may play with SA and IO just to see if AI can get a little improvement.
 
Is that just dialling in that voltage manually and everything stock? I don't need it overclocked, the PC is only ever use for games and I'm 3440x1440 so I'm GPU bound..
 
ChrisD. said:
Hmm, I have an MSI Z490 MEG Unify and 10700k and I've been having temp issues under a full custom loop. My CPU can easily hit 30 degrees above my GPU which is in the same loop when under a full stress test. I may check out the voltages as up until now I've been blaming the CPU block.
Click to expand...

This is what I have done but I am new at this. If you problem is high core temps at stock you can down volt. I would suggest you use auto to start and just bring the core voltage down to 1.22 and manually set the cpu ratio to 48 or leave it as auto. Also turn off enhanced turbo under CPU settings. Leave turbo technology 3.0 and turbo boost on. ENHANCED TURBO is the MSI version of MCE or multi core enhancement and you don’t need it at stock. Disable it.

If you leave everything at auto and just do cpu core voltage at 1.220v you can post in windows. Then test with Realbench with half your ram on a 15 minute stress test. WATCH TEMPS. If they hit 91c+ stop the test. At that voltage they shouldn’t but you cooling could be the problem.

if you post and tempt are stable anywhere up to 85c you are good. That high BUT you will never get there under normal workloads really. Once it completes the 15minute stress and you are happy with temps run for an hour but baby sit it.

you should test it on stress test over night or for 6 hours but I would advise leaving it to a day time to make sure it doesn’t BSOD. This isn’t gonna damage the CPU because all you are doing is decreasing voltage make sure you are using the (.) and always DOUBLE CHECK YOUR MANUAL VOLTAGE before saving and rebooting.

As a guide. I have run a 15 minutes stress with 16gb with xmp active. Core ratio is 48. LLC is mode 3 and everything else is stock. I may play with IO and SA voltages to see if I can get it cooler but with all that and. D15 I still sit at 86c under load. That is with no avx offset. I could bring temps down more by setting an offset but I want to find a good base voltage without it.

will try 4.9 again if the reseating of the cpu gives me the temp drop I am hoping but I am not optimistic. I this this card just doesn’t have the guts to overclock.
 
I did a bit of playing around, set all core to 4.8 GHz, VCPU to 1.175 and LLC to 3 and running the OCCT PSU test (which is high stress) and CPU temps topped out mid 80's, previously on default settings it was high 90's. In gaming my CPU seems to sit around 5 degrees higher than my GPU, whereas before it could be up to 30 degrees higher. No noticeable difference in gaming performance.
 
Lael Traylor said:
Was just driving to the store and I figure it out... MSI default under auto to either LLC mode 1 or mode 2. That was pushing the voltage over the vcore. I know the papers say LLC does increase voltage over set vcore but CPUZ, HWMONITOR and HWINFO all showed the same amount within 0.001 of each other. I changed LLC to 4 and low and behold the voltage is now 1.176 under load and vcore 1.200.

I will NEVER buy MSI again the stock settings on this board are awful.

Fail at LLC 4 and 1.2 volts.
Click to expand...

Stock settings are awful across the board. Even my Apex left on Auto is not good at all and that's meant to be a high end oc board.

Always set you own voltages for vcore/io/sa/vdimm
 
Lael Traylor said:
Any advice for io and sa? Should I set dimm to 1.5v the system always reads just under that anyway and if it’s the spec of the ram. Never played with ram beyond xmp...
Click to expand...
Corsair recommend 1.4v max (not surprising?). Assume heat can be an issue at 1.5v and 4 sticks, so I dialled mine back to 1.45v and 3866. Its the first time XMP hasn't worked, but both pairs were bought about a year apart.
 
Lael Traylor said:
Any advice for io and sa? Should I set dimm to 1.5v the system always reads just under that anyway and if it’s the spec of the ram. Never played with ram beyond xmp...
Click to expand...

1.325 for io and just at or right under for SA but unless you're pushing super hard timings like 16/4600, you shouldn't need that much. That's for CML. about 50mv lower SA for CFL.
 
Tinkers2004 said:
have you definitely removed the clear sticker that comes on the base of the heatsink when installing, on my sons he has the dark rock pro 4 pro at 5Ghz all core never exceeds 80 degrees on his 10850K
Click to expand...

In passing, I have done that before, and it really blows the temperatures. Lol. Plastic is not generally a good conductor of heat, haha.


But that temperature seems just way too high. With the case and cooler you have it should not be getting that high at stock. I mean it's immediately leaving you absolutely no room for overclocking at all. I would be checking the cooler to really see if the plastic is still in place, lol, or it is not seated correctly, or there is a paste issue. I mean I know these have been mentioned before, but I find it really difficult to understand why it is so high. The Noctua is as good if not better than most AIO so I am really not sure that's the problem.
 
You must log in or register to reply here.
Back
Top Bottom