Sign in with
Sign up | Sign in
Your question

3930k one core too hot for 4.8ghz OC - can't set per core multiplier

Last response: in Overclocking
October 13, 2012 7:25:29 PM

Hi Tom's! First off, thank you all for participating. Without these forums this first time system builder, OCer, and WCer would have been lost. :love: 

Here's my primary problem; I've got a 3930k where core 6 ranges from 15 - 20 degrees C hotter than my coolest core. This core is, I believe, the only thing standing between me and a stable 4.8ghz OC. Getting my temps from CPUID HWmonitor and realtemp. Stability testing with Prime95 small ffts. At 4.8 my setup passes Intel Burn Test at standard but not but not at high setting. Will test today to get BSOD codes and take screen shots of BIOS settings if necessary.

At load cores 1 - 5 bounce nicely around 67 - 77 degrees and stay there. Core 6 starts out higher 81 - 84 (depending on voltages right now I'm at 1.410) then at about 8 or 9 minutes into the small ffts test core 6 creeps up to 86, 87, 88 even higher then we BSOD, lock up, or I stop the test.

I've tried lower vcores seems I BSOD before I even get too hot, I assume because of under-volting. Tried but can't find the sweet spot that keeps core 6 down.

Same temp difference on air, fyi, (where I topped out at 4.2ghz with a hyper 212 evo push/pull) or now under water at 4.8ghz - multiple remounts and applications of thermal paste with both cooling solutions. Tried rotating my block. Temps got better in general by a degree or two but did not fix the problem. So, fairly certain it's my 3930k and a problem child of a core 6. Right? Intel support (surprise, surprise) was no help.

OK so, I thought a work around in my BIOS would be to set a lower per core multiplier for that hot core 6. A good idea? I don't know because of my secondary problem;

When I set a per core multiplier and boot into windows the CPU max is stuck at x36 or 3.6ghz. This happens if I set all cores to the same multiplier and if I start from optimized defaults and try a very low multiplier. I found one other guy online with the same problem - different Asus mobo - same cpu. He doesn't seem to have found a fix, if so he did not update on the forums.

I'm open to any and all suggestions and I will upload screenshots or provide more info if it will help. Thanks for reading.
a b à CPUs
October 13, 2012 8:40:34 PM

You can not set different multipliers for individual cores. In a Core i CPU, all active cores always use the exact same multiplier. Lightly loaded cores can go to sleep and enter various sleep states. This will reduce the average multiplier but as soon as a core wakes up, it will be using the same multiplier as all of the other active cores. The setting in the bios lets you adjust what multiplier your CPU will use based on how many cores are active.

Intel's core temperature sensors are not 100% accurate temperature monitoring devices. They were only intended to be used for thermal throttling and thermal shutdown control so there was no need for Intel to invest extra money in making them perfect. There can be variations of +/- 5C from core to core. If some cores read high and some cores read low, the combined reported difference between cores can approach 10C even when all cores are at the exact same temperature.

There is nothing you can do about this. You can remount your heatsink 101 times but you probably won't see any significant difference.
October 13, 2012 9:45:25 PM

Thanks for responding I do appreciate it. I'm having trouble though grasping why the per core multiplier option exists in my BIOS if it isn't to "overcome the thermal limitations of individual CPUs" (paraphrasing) as in this video at 22:50 :

He (JJ from Asus) is using my mobo with a 3960x as opposed to a 3930k but the same principle should apply, or am I missing something?

As an update with lower ambients today I was able to pass IBT at the "high" setting and Prime95 small tfts ran for 17 mins with core 6 hitting TJmax then BSOD code 0x00000101 which a quick googling of seems to support that my temps on core 6 are the problem.

Also, temp variation is not as great as I estimated. My other cores maxed between lowest 76 and highest 84. This is leading me to believe if improve my cooling I might be good to go. I have several ideas how to do that (better rad fans as a first step) but I'm over-budget and don't have much room left in my case without tearing it all down and taking it to my buddy's metal shop.
Not trying to turn this into a hardware thread haha but FYI I have two 2000rpm Blademasters on my Swiftech mcr220xp rad - I've tried it out of the case which got me down 1 or 2 degrees. Push pull outside got me down 2 to 4 degrees but once back inside did not have enough room to pull and push air.
Gentle Typhoons 1850 or 3000rpm with pwm mod is what I'm thinking of trying next if better cooling is my only possible solution. Honestly I've seen mixed reports about the blademasters as rad fans - I like their price and sound profile and happen to have four on hand so, yeah.

Also FYI Asus phone tech was able to set per core multiplier as I propose to on one of their test systems and thought my problem could be a CPU, mobo, or BIOS problem. He was a nice guy but that much I had guessed. :)