Overclocking a 5820K

[Cereal5]

Hi all,

 

This is my first time overclocking anything (except a monitor) so I wanted to make sure everything looks alright. I have my 5820K set to 4300 MHz at 1.25 volts, nothing else was changed. I'm using ROG RealBench as a stress test and MSI Afterburner to monitor core temps because I don't really know what else to use for that. on to temps:

 

Cores 1, 3, and 4 are hovering around 69-70 degrees, core 2 is about 67, core 5 is around 72, and core 6 is 64. Overall temp reads anywhere from 68-73.

 

Any reason for these variances in temperature? Is this normal, especially with 6+ cores? would the CPU be safe if one core ran at 83 while the rest were 80 and under? (if i turn the voltage up)

 

[edit] should I expect lower tempteratures when i overclock the RAM also or no?

[Enderman]

yes its normal for cores to be different temps

and yes 83C is still safe

[vanished]

Some variance is totally normal.  I noticed it on a Athlon 64 X2 3800+ back in the day, and I've noticed it on every other multi-core CPU I've seen since then.

[Cereal5]

[edit]

[Cereal5]

5 minutes ago, Ryan_Vickers said:

Some variance is totally normal.  I noticed it on a Athlon 64 X2 3800+ back in the day, and I've noticed it on every other multi-core CPU I've seen since then.

 

6 minutes ago, Enderman said:

yes its normal for cores to be different temps

and yes 83C is still safe

30 minute stress test just ended in a blue screen in the last minute. got the message: VIDEO_SCHEDULER_INTERNAL_ERROR. crashed literally at the very end of the stress test

[vanished]

well that's not so bad.  At 1.25 v I can get my 4770k stable at 4 GHz, but it won't do 4.1.  In fact, right now I've just got it set to full turbo but on all cores (3.9 GHz).  It manages that on stock voltage, but if I drop it by another 0.003 v it crashes  

[Energycore]

3 minutes ago, Cereal5 said:

30 minute stress test just ended in a blue screen in the last minute. got the message: VIDEO_SCHEDULER_INTERNAL_ERROR. crashed literally at the very end of the stress test

A crash like that suggests not enough voltage. Try 1.3v and see if you get any crashes after an hour.

 

Make sure to test stability for at least 8 hours once you find a voltage that completes the 1 hour stress test.

 

Keep in mind increasing voltage will increase temps, more so than increasing clockspeed. Make sure you don't go over 85.

[Cereal5]

Just now, Ryan_Vickers said:

well that's not so bad.  At 1.25 v I can get my 4770k stable at 4 GHz, but it won't do 4.1.  In fact, right now I've just got it set to full turbo but on all cores (3.9 GHz).  It manages that on stock voltage, but if I drop it by another 0.003 v it crashes  

so should i try it again at something like 1.26v? or skip to like 4.4(or 4.5) at ~1.3

[Cereal5]

4 minutes ago, Energycore said:

A crash like that suggests not enough voltage. Try 1.3v and see if you get any crashes after an hour.

 

Make sure to test stability for at least 8 hours once you find a voltage that completes the 1 hour stress test.

Will do, thanks

[Cheddle]

2 minutes ago, Cereal5 said:

Will do, thanks

No need to go so high. stability isnt very far for you. try 1.265v. try creeping the voltage in to the chip (cant recall what asus call it, vrin or something) up to your core + 0.6v i.e if you run 1.3v set the vin to 1.9v.

 

try adding 0.025v to system agent your still unstable. 

 

if still then add another 0.01v to core. 

 

30 mins stresss test is fine. leave it at that and then play some games do some things - if its still unstable for daily use then add more volts as temps allow or drop the mhz down.

 

FYI my daily stable clocks are 4.5ghz @ 1.27v but to pass a LINX stress test (you shouldnt LINX on haswell) I need 1.31v - yet I never get a BSOD during normal use or with cinebench or XTU...

 

doing crazy 8 hours stress tests are a total waste of time. if its stable daily - its stable.

[Cereal5]

2 minutes ago, Cheddle said:

No need to go so high. stability isnt very far for you. try 1.265v. try creeping the voltage in to the chip (cant recall what asus call it, vrin or something) up to your core + 0.6v i.e if you run 1.3v set the vin to 1.9v.

 

try adding 0.025v to system agent your still unstable. 

 

if still then add another 0.01v to core. 

 

30 mins stresss test is fine. leave it at that and then play some games do some things - if its still unstable for daily use then add more volts as temps allow or drop the mhz down.

 

FYI my daily stable clocks are 4.5ghz @ 1.27v but to pass a LINX stress test (you shouldnt LINX on haswell) I need 1.31v - yet I never get a BSOD during normal use or with cinebench or XTU...

 

doing crazy 8 hours stress tests are a total waste of time. if its stable daily - its stable.

trying 4.3 at 1.28, thanks for the info as well

[Kaiu]

I have the same cpu (i7 5820k) as you running at 4.2ghz overclock with 1.21vcore. I'm getting 70-71 degrees celsius on a H100i when I use asus realbench at 100% load. I dont get it, shouldnt a system with 2 gpus, a slighty higher overclock and higher voltage clock be alot more hotter than mine? I feel like mine should be slightly lower like around 65-70

 

[Cereal5]

1 minute ago, Kaiu said:

I have the same cpu (i7 5820k) as you running at 4.2ghz overclock with 1.21vcore. I'm getting 70-71 degrees celsius on a H100i when I use asus realbench at 100% load. I dont get it, shouldnt a system with 2 gpus, a slighty higher overclock and higher voltage clock be alot more hotter than mine? I feel like mine should be slightly lower like around 65-70

 

I have an H100i v2 if it makes any difference. but one would think it would be similar yeah. at 1.28125v only my 5th core is consistently above 70 degrees. do you have your memory overclocked as well? i do not, could definitely contribute to more heat

[Cereal5]

13 minutes ago, Kaiu said:

I have the same cpu (i7 5820k) as you running at 4.2ghz overclock with 1.21vcore. I'm getting 70-71 degrees celsius on a H100i when I use asus realbench at 100% load. I dont get it, shouldnt a system with 2 gpus, a slighty higher overclock and higher voltage clock be alot more hotter than mine? I feel like mine should be slightly lower like around 65-70

 

 

42 minutes ago, Energycore said:

A crash like that suggests not enough voltage. Try 1.3v and see if you get any crashes after an hour.

 

Make sure to test stability for at least 8 hours once you find a voltage that completes the 1 hour stress test.

 

Keep in mind increasing voltage will increase temps, more so than increasing clockspeed. Make sure you don't go over 85.

Passed a 30 minute stress test, 4.3GHz at 1.28125

[Cereal5]

28 minutes ago, Kaiu said:

I have the same cpu (i7 5820k) as you running at 4.2ghz overclock with 1.21vcore. I'm getting 70-71 degrees celsius on a H100i when I use asus realbench at 100% load. I dont get it, shouldnt a system with 2 gpus, a slighty higher overclock and higher voltage clock be alot more hotter than mine? I feel like mine should be slightly lower like around 65-70

 

It might be Asus' "OC socket" supposedly it makes overclocking easier... idk how real it is lol

[Energycore]

1 minute ago, Cereal5 said:

 

Passed a 30 minute stress test, 4.3GHz at 1.28125

Awesome! How were temps?

If those were good you can proceed to long stress test. CPUs usually take longer to make sure they're stable than GPUs because you can't exactly look for artifacts while doing numbercrunching, haha~!

 

30 minutes ago, Kaiu said:

I have the same cpu (i7 5820k) as you running at 4.2ghz overclock with 1.21vcore. I'm getting 70-71 degrees celsius on a H100i when I use asus realbench at 100% load. I dont get it, shouldnt a system with 2 gpus, a slighty higher overclock and higher voltage clock be alot more hotter than mine? I feel like mine should be slightly lower like around 65-70

 

Remember that ambient temperatures play a big role in the matter. Perhaps you're living in an area hotter than OP

[Cereal5]

Just now, Energycore said:

Awesome! How were temps?

Temps were basically the same. If i had to objectively tell you, I'd say 1 degree higher, maybe 2 at most. Will do a long

[Energycore]

1 minute ago, Cereal5 said:

Temps were basically the same. If i had to objectively tell you, I'd say 1 degree higher, maybe 2 at most. Will do a long

Awesome looks like you're good

[Cereal5]

3 hours ago, Energycore said:

Awesome looks like you're good

Awesome, thank you for all the help!

[Kaiu]

my ambient is 77 degrees F. so thats probably why. I thought about redoing my thermal paste because i think its crap.

[Energycore]

2 minutes ago, Kaiu said:

my ambient is 77 degrees F. so thats probably why. I thought about redoing my thermal paste because i think its crap.

You can probably get 4-5 degrees out of good thermal paste. My preference is Arctic Silver 5 but that's just cuz it's what I can get locally. You might be able to do better than that paste.

[vanished]

56 minutes ago, Energycore said:

Remember that ambient temperatures play a big role in the matter. Perhaps you're living in an area hotter than OP

Yes, absolutely they do, and it's no mystery how it works either.  If your ambient temp goes up 5 degrees, so will the component temps.  It's that easy

[Cereal5]

3 hours ago, Kaiu said:

my ambient is 77 degrees F. so thats probably why. I thought about redoing my thermal paste because i think its crap.

yeah my ambient is about 70, and as for thermal paste i use arctic MX-4 which according to linuses video (2 years ago) it was better buy about 3-4 degrees

 

[edit] probably closer to 72 after bench marking lol

[Energycore]

Just now, Ryan_Vickers said:

Yes, absolutely they do, and it's no mystery how it works either.  If your ambient temp goes up 5 degrees, so will the component temps.  It's that easy

I know that's consensus and I've seen that myself also, but I still don't know why that's how that is. It feels like it shouldn't be that simple, you know?

[vanished]

14 minutes ago, Energycore said:

I know that's consensus and I've seen that myself also, but I still don't know why that's how that is. It feels like it shouldn't be that simple, you know?

Unless you want to get into the minor effects like radiation heat transfer, specific heat capacity as a function of temperature and stuff like that which will all make less than a fraction of a degree of difference, it really is that simple, trust me

 

Just as a basic explanation, remember that the rate at which energy (heat) is transferred from the heatsink to the air is dictated by the product of three variables: the surface area of the heatsink, a parameter that varies with how the air moves (in very simple terms - this is actually quite an in-depth topic in and of itself) and the difference in temperature between the heatsink and the air.  Since neither of those other terms are functions of temperature, we can say with mathematical certainty that rate of heat transfer is proportional to temperature difference, and thus that for a given setup, the component temperature will be a certain amount of degrees above ambient, and that this amount is characteristic of that setup (that chip at that load level with that cooler, etc.)