Is this temperature safe?

[biocentrism]

My setup is a 5960X on an Asus X-99 Deluxe.

I have overclocked to 1.3v and a 44 multiplier.

Running RealBench for 15 minutes here are the max temperature readings for 6 cores (Real Temp does not display readings for the last 2 cores).

Core1 - 91C
Core2 - 87C
Core3 - 96C
Core4 - 98C
Core5 - 92C
Core6 - 95C

I know that the throttle temp is 105C but from what I am reading online it seems that these stress test temps are higher than what most people experience.

My question is are my overclock settings safe? Real world applications will come no where near the stress test levels so maybe this is ok? Could I push it a little further until it gets to 105 C or should I be reducing the multiplier/vcore, and if so, to what temperature range?

Thanks.

 

[-Lone-]

That's pretty high, for me at least, but I've never even been close to that.

 

[biocentrism]

Yeah that's what I figured.

Could it be because there isn't enough voltage? If I increase the voltage would that reduce the temperature?

 

[-Lone-]

Yeah that's what I figured.

Could it be because there isn't enough voltage? If I increase the voltage would that reduce the temperature?

I'm not exactly an OC expert, lol. But I think it's the other way around, the more voltage, the more heat and power. Which cooler do you have by the way, also which case with how many fans, do you have good air flow?

Edit: But I do know 85C is the highest you should reach max before you go up to high to extreme temperatures.

 

[biocentrism]

I would have thought that too but then I read this on another thread:

Jeankyyyy said:
If you want a good performance try putting 1.300v and 4.5ghz on the cpu,because 1.2v is not enough for 4.2ghz and it will get the cpu hot since it needs more volts,also tell me what kind of liquid cooler you have and i'll help in every single thing you have .

I have a 3 fans, water cooled. No idea what the case is. Its rectangle.

I am running at 1.31volts now. Its about 1 degree hotter. I think cored 3 and 4 are bad silicon.
Core1 - 91C
Core2 - 87C
Core3 - 97C
Core4 - 99C
Core5 - 93C
Core6 - 94C

 

[-Lone-]

Well, that's your problem right there, only 3 fans total? My full specs are at bottom right corner, but I use the cooler master Nepton 280L liquid cooler.

 

[biocentrism]

There are 3 external case fans. I believe there are a total of 6-7 fans if you count the internal ones, not exactly sure.

 

[jshoop]

3 fans if configured correctly is all you will need. however, i do believe your temps are high. i like to keep my stress test temps under 85, but thats just me. i think your on the edge a little bit. i would back down the voltage and if you have to drop the multiplier.

as for cores 3 and 4, the could be in the middle of the chip itself, so it heats up more

 

[biocentrism]

as for cores 3 and 4, the could be in the middle of the chip itself, so it heats up more

ah very good!

 

[-Lone-]

There are 3 external case fans. I believe there are a total of 6-7 fans if you count the internal ones, not exactly sure.

Oh, well, that's different then

 

[biocentrism]

lowering the voltage did indeed lower the max temp in this instance.

1.29 lowered the max temp 6 degrees to 93.
1.28 lowered the max temp 2 degree to 92.

given that cores 1 and 2 seem to be running a lot cooler, could I not increase the multipliers on those 2 cores only, say to 46 and leave the others at 44? has anyone tried this and what would be the effect?

 

[jshoop]

all the cores run at the same multiplier, you cant selectively overclock single cores. you can disable some cores in bios but i dont believe you can choose which cores to disable

 

[biocentrism]

my BIOS provide the ability to set different multipliers for each core, but seems they must be in declining. read online that 47-47-46-46-45-45-44-44 would be good. I am trying that now but it is running really hot (97 C). seems to run at the lowest multiplier when at load. if system only is using fewer than all cores then it runs at the lowest multiplier of those it is using.

 

[jshoop]

hmm, thats interesting. can you set the vcore for each core though? as you would need each core having a different vcore in order to lower temperatures all around. even though the core is running at 44x multiplier, it could still be recieving the vcore of the core running at the multiplier of 47

 

[biocentrism]

no the vcore applies to all of them. from what I read the way it works is that if the computer needs to use all cores, then all the cores will run at 44. if the computer only needs 1 or 2 cores, then those will run at 47. in both instances at the vcore. but if you apply adaptive voltage, then you can run the fewer cores at lower voltage and then the higher voltage will be used when at load. not sure about this but it is what I gathered.

 

[jshoop]

the easiest way is to do some testing

download prime 95 and cpu z, on prime 95 start only 2 workers, instead of all of them, and see if cpu shows the 47 for the multiplier, then start two more workers and see what happens

 

[Karadjgne]

If using prime95, make sure it is version 26.6. The newer versions use the avx and fpu instruction sets differently and severely overheat the Haswells unrealistically. Check out the Intel temp guide sticky on this site for better information. And there is a version of Realtemp that'll run 6core cpus, use the ROG version, it can be found with the other branded versions.

 

[jshoop]

that was a very good read, thanks for posting that

 

[biocentrism]

so I have done some testing. I have set the multipliers to 47-47-46-46-45-45-44-44 for my 8 core CPU with the vcore set at adaptive 1.20 +0.18 offset.

from what I have found, unless you disable cores, all programs will run assuming you are running 8 cores and the speed will be 44x.

even when I run single-threaded processes that only use 1 core, the CPU operates assuming 8 cores and hence the 44x multiplier.

but disabling cores does give you that added boost. but its not practical to manually turn on/off cores before running programs.

 

[Karadjgne]

Is the pump tight to the cpu? I'm not talking average tightness like used with an air tower, I'm talking overly tight where you actually broke out a screwdriver, because for some odd reason (maybe pump vibration) when the pump is exceptionally tight it shows lower temps. Also may want to check for an air bubble trapped in the pump if your feeders are downward pointing or close.