Overclocking a 3570k with odd thermals

[GamingDaemon]

Hi Fellow Ananders,

About a week ago, I upgraded my gaming rig to an 1155 Gigabyte motherboard and the 3570k CPU. I didn't overclock it to start; instead, I just let it be for a few days to see how much it improved my machine over my old first gen Nehalem 860 proc.

Now I am ready for some overclocking, but I noticed with HWMonitor that Cores #1 and #3 are typically 5 to 7 degrees Celsius cooler than Cores #0 and #2. What that means is that when Cores #0 and #2 are at 34C, Cores #1 and #3 are flipping around between 24C and 29C. Should I be worried about overclocking?

I think the cause is that when I put my thermal paste on the 3570k, I tried something different. Instead of putting a drop of it on, and then putting on the heatsink, I wrapped my finger in a plastic baggy and spread the paste out over the proc. Is that a bad thing?

So, should I be worried about the thermals? Should I re-do the who heatsink?

Thanks in advance...

 

[Kenmitch]

Typical on Intel chips. How's temps under load?

 

[pwoz]

If I remember right from reading I did a while ago, the temperature sensors are not very accurate at low temps. Load the processor and see if you get a delta. You likely still will as many peolpe have found with IB CPUs, but this should not be an issue.

You may have gotten some air bubbles spreading the paste with a plastic bag. The best way is a line down the heatspreader oriented the same direction as the core. The application of the heatsink will spread the paste.

 

[GamingDaemon]

So just a quick test with OCCT, I saw temps of between 49C and 53C for Cores #0, #1, and #2. Core #3 was consistently cooler by about 3 to 5 degrees. This is with no overclocking, but turbo enabled.

Looks a little more consistent when under load.

So are those good temps? They seem a bit high to me without overclocking.

 

[Phynaz]

Temps are fine.

 

[Ferzerp]

Idle temp readings are useless.

 

[coffeejunkee]

So just a quick test with OCCT, I saw temps of between 49C and 53C for Cores #0, #1, and #2. Core #3 was consistently cooler by about 3 to 5 degrees. This is with no overclocking, but turbo enabled.

Looks a little more consistent when under load.

So are those good temps? They seem a bit high to me without overclocking.

Not high at all for a stresstest. But with IVB, core 0 should always be the coolest because it's next to the igp which even if enabled barely gets hot.

 

[GamingDaemon]

So, I tried a simple overclock up from 34x to 38x, giving me 3.8GHz. Easy enough.

But here are my temps from HWMonitor with Cinebench 11.5:


Current Min Max
--------------------------------------
Core #0: 38C 31C 57C
Core #1: 27C 24C 55C
Core #2: 31C 27C 55C
Core #3: 35C 21C 48C
--------------------------------------
Pkg: 38C 31C 57C

Should I remove the heatsink, wipe off the thermal grease, and start over? Or am I OK?

 

[LoveMachine]

For piece of mind, wouldn't hurt to ensure you have good contact by re-installing the HSF. Temperature-wise, anything under 60C is absolutely nothing to worry about. My 3570K at 4.3 runs in the low 60s at full load and I'm not concerned in the least. The processor itself won't throttle until 105C, so temps in the 80s aren't unheard of. The real killer is voltage, and temps go up to dangerous levels because of voltage increases. So if you are at stock voltage, you could run your current OC 24/7 for years at full load with no problems (statistically speaking). Even a little bump higher is fine. At 4.3, I think I'm running something like 1.16V, just a tiny bump up from stock.

 

[GamingDaemon]

For piece of mind, wouldn't hurt to ensure you have good contact by re-installing the HSF. Temperature-wise, anything under 60C is absolutely nothing to worry about. My 3570K at 4.3 runs in the low 60s at full load and I'm not concerned in the least. The processor itself won't throttle until 105C, so temps in the 80s aren't unheard of. The real killer is voltage, and temps go up to dangerous levels because of voltage increases. So if you are at stock voltage, you could run your current OC 24/7 for years at full load with no problems (statistically speaking). Even a little bump higher is fine. At 4.3, I think I'm running something like 1.16V, just a tiny bump up from stock.

Thank you for the quick reply. And yes, I am at stock voltage. I just bumped it up to 40x for 4.0GHz, and all the temps are under 60C (highest one was 58C) with the stock voltage.

I'm going to run OCCT for 8 hours to make sure it is stable, and then think about 4.3. Perhaps, when I have some time, I will re-seat the HSF if it is still bothering me.

 

[coffeejunkee]

Ambient temp? Mobo and hsf you are using? Case? Like I said, it's strange to see your core 0 being the hottest, everyone else reports this being the coolest which makes sense because it's next to the igp so it can dump its heat there.

Also, auto does not mean stock voltage. Your mobo is probably overvolting which explains fairly high Cinebench temps. I got 45C at 4.1GHz.

 

[GamingDaemon]

Ok, here's all the info you were asking for, coffeejunkee:

Case: Fractal Design Define R4 Black Pearl
Motherboard: GIGABYTE GA-Z77X-UD4H
Heatsink: COOLER MASTER Hyper 212 Plus

Ambient Temp: Not sure. We like to keep it comfy inside. Probably more than 72 degrees Fahrenheit because my media server is about a foot away in another fractal Design case...so maybe 74?

I pushed it up to 4.2GHz and ran OCCT for 8 hours last night. According to OCCT, here's what the temps were:

Current Min Max
--------------------------------------
Core #0: 33C 30C 62C
Core #1: 21C 20C 66C
Core #2: 26C 24C 64C
Core #3: 20C 14C 59C
--------------------------------------

Those are slightly different (by around 1C) from HWMonitor, which I had running at the same time, but close enough.

According to OCCT, VCore maxed during the 8 hours out at 1.2v, but in the pictures it created, it looked steady at 1.16v.

Finally, OCCT said there were no errors, but just after 3 hours in last night, all 4 cores dipped in usage from 100% to about 70% then down to 0% for a split second, back up to 70% and then back up to 100%. Not sure if that means anything.

So, coffeejunkee, do I have an issue with the HSF?

 

[coffeejunkee]

Well, assuming you mean normal occt test (not linpack), I ran it for 5 minutes at 4.2GHz and got 51-57-59-57 across cores. We can't really compare because I use a different case and cooler and also less vcore but at least my core 0 is coolest and core 2 hottest, like it should be.

I'm not sure if those temps are normal or not for Hyper 212 but I do think the thermal paste isn't spread out optimally. With a DHT cooler it can make a difference. I think it's actually best to apply tim to the cooler itself instead of the cpu, also filling the gaps between heatpipes a bit.

 

[GamingDaemon]

So I decided to remove the HSF, remove the thermal paste, re-apply it, putting it on the bottom of the HSF evenly along with a little bead on the processor, and then put the HSF back on.

Once that was done, I ran OCCT for about 5 minutes. Here's what I saw:

Current Min Max
--------------------------------------
Core #0: 33C 31C 60C
Core #1: 28C 21C 64C
Core #2: 31C 24C 62C
Core #3: 21C 18C 58C
--------------------------------------

This appears to be pretty consistent to my last run. It just seems to behave this way.

My only other idea is that I have the processor's onboard video switched on when I am just displaying the OS video. It might not be using my GeForce 680 GTX at that time. I checked the Gigabyte manual, and this is turned on by default. So, that might be the cause of Core #0's higher temps.

UPDATE: After disabling the onboard video (I believe) in the BIOS, not much changed:

Current Min Max
--------------------------------------
Core #0: 33C 31C 61C
Core #1: 28C 21C 64C
Core #2: 31C 26C 63C
Core #3: 21C 19C 56C
--------------------------------------

Hmmmm....

 

[coffeejunkee]

Ok, I'm sorry if I caused you extra work but I think it's safe to conclude your 3570K is a bit different from the rest. Just to show I'm not making this up, please look at the die shot in post 5 of this thread: http://forums.anandtech.com/showthread.php?t=2306804&highlight=3770k+core+temps. A possible explanation could be the distribution of tim under your cpu's own integrated heatsink. Or more likely, a not completely flat cpu ihs or cooler base.

My numbers were actually with the igp active and there still was a big difference, that's because it uses only like 8W max. And when you use a discrete card the igp is disbabled automatically. Anyway, apart from the fact your core 0 isn't the coolest those temps are nowhere near worrying levels and I wouldn't bother with remounting again.

 

[GamingDaemon]

No worries. I needed to do it to appease my OCD tendencies anyway. As you say, the temps are not worth worrying about, so I will leave it be. It could be the proc or HSF are not flat surfaces, or some other quantum imperfection is causing the temp disparity.

Oh well.

Thank you.