4770k weird temperature jump

[trylo]

Hi guys!

 

I'm working on my 4770k overclock.

4.5GHz @ 1,35V is not stable, but 4,4GHz manages to start IntelBurnTest, temps are good for 2-3 minutes (around 68-72*C), and then all the sudden jump to 100*C and CPU starts to throttle itself down. I tried max speed on the pump and performance settings on my H100i - no difference. first few minutes are fine and then temps go crazy. My load line calibration is set on Medium.

 

Please help.

[tp95112]

Hi guys!

 

I'm working on my 4770k overclock.

4.5GHz @ 1,35V is not stable, but 4,4GHz manages to start IntelBurnTest, temps are good for 2-3 minutes (around 68-72*C), and then all the sudden jump to 100*C and CPU starts to throttle itself down. I tried max speed on the pump and performance settings on my H100i - no difference. first few minutes are fine and then temps go crazy. My load line calibration is set on Medium.

 

Please help.

1.35V seems high 

[mvitkun]

@trylo

 

2-3 minutes seems like a long period of time but according to OC3D's Tom after a small period of time,when a large potion of the system memory is being used,haswell CPU's temps skyrocket by 20-30C.

has something to do with the memory controller.

 

 

 

also 1.35v seems very high as said above

[ProKoN]

Use aida64 or intel tuning utility to stress haswell.

The point of stability testing is to test system stability

Not to see how much heat you can generate under artificial

load.

[MegaDave91]

FPU only will make your CPU incredibly hot. Just a word of caution.

[Facety]

Use Aida64 for stability, it will stress your CPU without overwhelming your cooler, I found out the temps stay a bit lower than using prime or occt with avx and linpak by around 15 - 20C

[RDilux]

temps jumps are normal in the haswell cpu's

but i see you have a H100i and there is a big issue with the mounting of the H100i

i suggest you to contact corsair with it because this is a know issue on the H100i

or sell the cooler and buy a H110 or kraken X60 instead

[trylo]

OK, so I used IntelBurnTest as this is what NCIX recommended.

 

 

temps jumps are normal in the haswell cpu's

but i see you have a H100i and there is a big issue with the mounting of the H100i

i suggest you to contact corsair with it because this is a know issue on the H100i

or sell the cooler and buy a H110 or kraken X60 instead

 

What???
I've been reading forums and tracking news and everybody said that the mounting is exactly the same, the only difference is that it is 0,04 mm difference which is compensated by simply tighting the screws.

 

I had so much hassle mounting it in 350D case, I hope I don't have to unmount it and sell it (BTW H100i costs the same as a radiator to my car!!!).

So can someone elaborate do I really need to change the cooler. All the hardware gurus in internet used H100i when testing Haswell...

 

 

Back to the topic:

 

So I set my 4770k to 4.4GHz @ 1.3V, turned on AIDA64 and left my house for 5 hours. After I came back I found out that it crushed... Is it possible to get some info from AIDA64?

 

I will try now 4.3GHz @ 1.3V. As soon as I get stable I will start lowering voltage.

 

Side question: so if Haswell becomes hot when RAM gets full won't I have problems in real-life video rendering situation? My previous computer had 8GB (this one has 32GB) of RAM and it was full all the time.

Thanks in advance for all the help!

[Scia]

OK, so I used IntelBurnTest as this is what NCIX recommended.

 

 

 

What???

I've been reading forums and tracking news and everybody said that the mounting is exactly the same, the only difference is that it is 0,04 mm difference which is compensated by simply tighting the screws.

 

I had so much hassle mounting it in 350D case, I hope I don't have to unmount it and sell it (BTW H100i costs the same as a radiator to my car!!!).

So can someone elaborate do I really need to change the cooler. All the hardware gurus in internet used H100i when testing Haswell...

 

 

Back to the topic:

 

So I set my 4770k to 4.4GHz @ 1.3V, turned on AIDA64 and left my house for 5 hours. After I came back I found out that it crushed... Is it possible to get some info from AIDA64?

 

I will try now 4.3GHz @ 1.3V. As soon as I get stable I will start lowering voltage.

 

Side question: so if Haswell becomes hot when RAM gets full won't I have problems in real-life video rendering situation? My previous computer had 8GB (this one has 32GB) of RAM and it was full all the time.

Thanks in advance for all the help!

 

I think it was Logan who mentioned that there was a slight gap, and a gap means air, which means bad heat conduction which means bad temps on your CPU.

 

and 1.3V seems REALLY high for a Haswell CPU.

But there also is the possibility that your CPU simply won't go that high. Intel claims 4.8ghz, most retailers and testers say that the normal midrange silicon-lottery chip will top out at 4.2-4.4

If you need 1.3V for 4.3 GHZ, don't do it. A H100 cannot cool the amount a Haswell puts out when running at this kind of voltage.

 

And yes, once the memory gets loaded, the CPU will heat up more, since the memory controller is stressed aswell.

[trylo]

@ Scia

 

I've been running AIDA64 for 18 minutes now (4,3GHz @ 1.3V). and the temps are belowe 78*C (once every 30-60 seconds there is a jump for a fraction of second up to 85*C). So I am assuming that cooling works alright.

 

Does this make sense: as soon as I get a stable 5-6h run on AIDA64 I will start lowering voltage, when I drop below stability I will move 0,01 up?

 

[Scia]

Sounds about right. But really don't overdo it on the voltage. Most OCs I have seen are somewhere around 1.25 for about 4.4. Of course every chip is different.

If you get it somewhat stable have it running for 24h on Prime 95 small fft. If it passes it's as stable as it gets for an overcloked CPU.

[RDilux]

I think it was Logan who mentioned that there was a slight gap, and a gap means air, which means bad heat conduction which means bad temps on your CPU.

 

and 1.3V seems REALLY high for a Haswell CPU.

But there also is the possibility that your CPU simply won't go that high. Intel claims 4.8ghz, most retailers and testers say that the normal midrange silicon-lottery chip will top out at 4.2-4.4

If you need 1.3V for 4.3 GHZ, don't do it. A H100 cannot cool the amount a Haswell puts out when running at this kind of voltage.

 

And yes, once the memory gets loaded, the CPU will heat up more, since the memory controller is stressed aswell.

you should stop saying 1.3v is high for haswell because once again that is balony...

i got my on 1.43vcore with the H110 on 4.7Ghz and temps won't go higher then 81c....

what brings me back to the issue of the H100i the mounting is bad

[Scia]

you should stop saying 1.3v is high for haswell because once again that is balony...

i got my on 1.43vcore with the H110 on 4.7Ghz and temps won't go higher then 81c....

what brings me back to the issue of the H100i the mounting is bad

 

1.43V is a very good voltage. Perhaps already think about which one you are going to replace it with.

1.43V for 4.7ghz, GJ.

[RDilux]

1.43V is a very good voltage. Perhaps already think about which one you are going to replace it with.

1.43V for 4.7ghz, GJ.

lol stop talking crap dude 1.45 is the most maxed used vcore of every 1155 cpu and won't hurt the cpu on 1150 either

i do overclocking for 10 years now and i know that to do and dont to do

as long you keep the temps nice and below 70-80c (and 80c is in a not realistic senario) then you are good to go

overclockers done this for years and my i7 2600k is stil alive after 3 years with a voltage of 1.420 4.9Ghz 24/7

and don't start on that was sandy bridge because as long you keep everything nice and cool there won't be a issue at all

[Scia]

lol stop talking crap dude 1.45 is the most maxed used vcore of every 1155 cpu and won't hurt the cpu on 1150 either

i do overclocking for 10 years now and i know that to do and dont to do

as long you keep the temps nice and below 70-80c (and 80c is in a not realistic senario) then you are good to go

overclockers done this for years and my i7 2600k is stil alive after 3 years with a voltage of 1.420 4.9Ghz 24/7

and don't start on that was sandy bridge because as long you keep everything nice and cool there won't be a issue at all

 

You might want to read up on this Electromigration

Because

 

 

as long you keep the temps nice and below 70-80c (and 80c is in a not realistic senario) then you are good to go

is a very uneducated statement.

[RDilux]

 

 

You might want to read up on this dear Mr. Pro. Electromigration

Because

 

 

is a very uneducated statement.

 

well yea idc :)

it's you not owning the cpu vs mine oc experience.

user should stop reading reviews and then think the know stuff because that not how this work at all.

own in and them come talk

[Willson94]

lol stop talking crap dude 1.45 is the most maxed used vcore of every 1155 cpu and won't hurt the cpu on 1150 either

i do overclocking for 10 years now and i know that to do and dont to do

as long you keep the temps nice and below 70-80c (and 80c is in a not realistic senario) then you are good to go

overclockers done this for years and my i7 2600k is stil alive after 3 years with a voltage of 1.420 4.9Ghz 24/7

and don't start on that was sandy bridge because as long you keep everything nice and cool there won't be a issue at all

Mate 1.45 is high. Don't really want to be going over 1.4 on even 1155 unless you want your chip to deteriorate quickly. 

[Scia]

well yea idc :)

it's you not owning the cpu vs mine oc experience.

user should stop reading reviews and then think the know stuff because that not how this work at all.

own in and them come talk

 

That's the scientific approach I have been looking for.

Ty for stating my point, and in the same post showing off your "10yrs of experience" LOL

Have a good say Sir kkthxbye.

[Scia]

Mate 1.45 is high. Don't really want to be going over 1.4 on even 1155 unless you want your chip to deteriorate quickly. 

 

Nono, this doesn't happen with his CPUs. He's the king of prenatal overclocking with his 120yrs of experience. L O L ^^

[RDilux]

That's the scientific approach I have been looking for.

Ty for stating my point, and in the same post showing off your "10yrs of experience" LOL

Have a good say Sir kkthxbye.

learn overclocking get a haswell and come back when you have one

you don't know anything

[Leonard]

Hi guys!

 

I'm working on my 4770k overclock.

4.5GHz @ 1,35V is not stable, but 4,4GHz manages to start IntelBurnTest, temps are good for 2-3 minutes (around 68-72*C), and then all the sudden jump to 100*C and CPU starts to throttle itself down. I tried max speed on the pump and performance settings on my H100i - no difference. first few minutes are fine and then temps go crazy. My load line calibration is set on Medium.

 

Please help.

Max speed on your cooler won't work. All the videos i have seen on Haswell CPUs all say not to go higher than 1.2 - 1.25v and that is with custom water loops and beefy air coolers. You have gone to 1.35v, your CPU just can't handle it that amount of voltage. You better drop it like its too hot!

[RDilux]

Max speed on your cooler won't work. All the videos i have seen on Haswell CPUs all say not to go higher than 1.2 - 1.25v and that is with custom water loops and beefy air coolers. You have gone to 1.35v, your CPU just can't handle it that amount of voltage. You better drop it like its too hot!

ppl say if you are a extreme oc you don't really care about what ppl say

[Facety]

But is't that the point off a stress test to overwhelm it ?.......................

 

To overwhelm the CPU, not your cooling solution. Aida64 is recommended for haswell because it tests the new instruction sets

http://www.aida64.com/whatsnew/memory-benchmark-amd-kabini-temash-intel-haswell-l4-cache

 

You can run the other tests but I think those tests that let the CPU pump out a whole lot of heat are geared more towards testing heatsink/cpu cooler solutions rather than the cpu itself.

[Facety]

Your cpu is mean't to be 100% stable if you gonna test with different software just to get more stable your doing it wrong.

 

I agree with you and I do use different software to test my setup. I tested with OCCT with avx and linpak  and 90% memory used at a certain OC and it was stable for hours with temps hovering at around 90C, I ran that same OC on prime and it was stable for hours as well....then I tried it on Aida64 and it BSOD in less than 10 minutes with temps around 78C, I couldn't believe it so I ran it on Aida64 5 more times and it BSOD around the same time period.

 

So I got it stable with Aida64 first and guess what, the others were stable as well.

 

You may say my testing methodology is wrong but this is what I use for stability

Aida64 - CPU

OCCT with linpak, avx and 90% memory - Heatsink and RAM

Prime - reassurance

[Facety]

I see, my RAM is stable, and that OC was just to see how far I could push it on air, I dailed it back to 4.2 at 1.15v where the temps dont exceed 70C in the stability test, in real-world scenarios it doesn't pass 60C so its all good until I decide to setup a water loop to push it further.