Help - heat problems with new i7 3770k overclocking

zootlewurdle

zootlewurdle

zootlewurdle

Associate
Joined
18 Mar 2013
Posts
8
Hello,

hate to have to create an account just to ask for help, but I'm, well, needing some help!

I'm having to replace a problematic mobo/cpu, so the new build currently looks like this:

Antec 902 case
Asus P8 Z77-V LX
i7 3770K
Corsair 800W PSU
Corsair H60 cooler
Thermal paste used - Asaka 455.
2 x 4GB Corsair Vengeance
1 x 1TB Seagate Barracuda

(once stable, added to that will be:
2 x 4GB Corsair Vengeance
1 x nvidia 680GTX - this will add some extra heat inside the case too
1 x 2TB Seagate Barracuda)

I wanted to try and overclock this. I thought I'd start with what I thought was a modest 4.1GHz. I haven't messed about with this stuff for a good 7+ years so I'm fairly rusty.

The problem: After about 15 to 20 mins of running prime95, my temps repidly spike from around the low 70s to over 90. It's a rush to get things shut down before it spikes over 100c, which it did once. On idle the temps are around 33c. Sorry, don't know the ambient room temp, I'd guess around 22c (home thermostat set to 18c, but the computer room always a fair bit warmer).

I'm not sure if this is a BIOS setting, or some misunderstanding of how to run prime95, or something else I'm not aware of.

The fan on the H60 radiator is pulling air in from the behind the case. Top fan vents out. Another fan on the side of the case also pulls air in to the case (changing airflow direction on that has no noticable effect). Fans in the empty drive bays at the front also pull air in from outside.

If it helps, some bios screenshots:

s1.jpg

s2.jpg

s3.jpg

s4.jpg

s5.jpg

s6.jpg

s7.jpg

s8.jpg

s9.jpg
 
The GPU overvoltage might have a lot to do with it, seriously if you want a faster GPU then buy a proper graphics card. :p

A better thermal paste might knock a few degrees off as well.
 
Last edited:
Thanks for the reply, but as I said at the start, once the system is stable I already have a 680GTX to put in the case. I cannot do so now because the card is in use in the old machine.
 
I haven't touched any of the gpu settings as I didn't see the point seeing as I'd later pop in a proper card, so from my perspective those settings are already 'stock' :S

Forgive me for possibly souding really ignorant here, but does prime95 hammer the gpu part as well?

I may pop the gfx card over and retest too, a bit tiresome but worth it.
 
I'm just looking at the "iGPU offset voltage" which is set to +0.150V and is in pink to warn you that voltage is high. At stock offset should be +0.000 meaning nothing added.

Prime95 doesn't touch the GPU but if voltage of the iGPU is high then it's adding extra heat to the processor die, which won't help given that Ivy Bridge already run hot due to TIM/Heatspreader transfer issue.
 
Last edited:
You have though, the igpu voltage has been increased, also the llc is at ultra high that really isnt required for 41 multiplier you shhould be able to keep that on regular. lastly i always use offset voltage that way the chip lowers the vcore under less load
 
Last edited:
:S iGPU voltage, I must have accidentally set that. I've reset it to auto and retested, and retested with the 680 in place. Both times temps went over 90c once prime95 started the 800000 iteration tests and the machine shut down.

I've tried reducing the llc to regular and the one above that, but got a blue screen almost immediately after starting prime95.

Trying again now on the one just below ultra.

As for the 41 multiplier, I'd like to get that higher but thought 41 was a good modest starting point.

edit: I hadn't tried offset mode for the CPU simply because the guides I found whilst googling seemed to suggest manual. For offset I'm not sure what to try for.
 
with offset just stick it at auto to see what it comes to its likely to be too high but theres a chance your bsods are related to not enough vcore at 1.100. Edit id also disable the igpu too and just run with the 680. Do you have any cpuid and core temp screens?
 
Last edited:
Yeah, with offset set to auto and llc reduced to normal, it's bordering on 90c at the earliest prime95 tests, let alone the later ones that get really hot!

CPU-Z reports core voltage at around 1.192 to 1.2v

No bsod thus far though...

I guess at this point then I'd need to start putting in a negative voltage offset?

I dunno, still sounds too hot too early to me, but I have no factual basis for that! :D
 
thats it start with negative see if that helps. One thing to remember that i made the mistake off thinking positive offset isnt plus your auto voltage its plus your base voltage. 1.2v and 90c still sounds warm im running+ 0.010 offset to get to 4.4 which shows about 1.16 vcore and im not breaking 70c
 
That is, in fact, a very good question because I had not tried that myself!

So, that is now tried.

With resetting to defaults, at the same point during prime95 where previously temps would spike to 90 (or 100 if I let it), they now get to around 80c ish. I quit it at that point simply because it was getting far further into that test cycle than I'd ever seen it get.
 
I believe I am correct in saving that 80c for a stock 3770K with a H60 is not supposed to happen? I would expect that result from the stock Intel cooler.
 
What are your temperatures like inside your case? I as because if case is getting hot so will CPU and cooler. Cooler can only cool if it has cool air going into it.
 
Well yeah, my research suggested the H60 should be able to cope. I'll try a different thermal compound, getting some Gelid GC Extreme. Maybe some clean-up and re-seating of the heat pump is also needed.

Internal case temp; I don't have data available. The airflow out feels cool to me.
 
Best guess is either your H60 is defective or it's not seated properly.

Check your seating and reseat is probably good place to start. It doesn't matter much what TIM you use as all the good ones are about the same. Even no TIM wouldn't be any hotter than your temps are now. :D
 
OK so, cleaned and regreased (also used a different tim) and reseated.
At stock (which for the board seems to be a 37 multiplier turbo): I was getting 60c under the same load that was previously getting me 80+c at stock.

Didn't leave prime95 running much beyond 4 cycles of the 800000 iteration tests as that was far further than it would get last time before I got nervous and killed it.

I guess now I can get back to the overclock tweaking.
 
You must log in or register to reply here.
Back
Top Bottom