Strange acting 4770k

TurboJ · Nov 19, 2013

I have been getting BSOD 124 occasionally when doing longer stress tests.

I'm running my 4770k with a True Spirit 120 mm cooler - it's not very big but it's supposed to be good for its size.

Running 1.18 Vcore and all cores synced to 42 (now 41 to avoid that BSOD)

I'm very puzzled about all these stress test programs though.

*Running Aida64 CPU stress or all the related stress tests together gets a max temp of 58 C
*Running RealBench stress test gets a max temp of 65 C
*Running various games gets a max tem of 58 C
*Running Prime95 small FFTs gets a max temp of 98 C

How is that possible? How can P95 get 34 degrees hotter than I'm able to achieve using any other stress method?

What can be the problem here? I don't get it, I really don't.
If it's just the newest version of Prime95 that doesn't want to play nice, what would be a second reliable stress test that I could use?

Y0GI · Nov 19, 2013

TurboJ :

I have been getting BSOD 124 occasionally when doing longer stress tests.

I'm running my 4770k with a True Spirit 120 mm cooler - it's not very big but it's supposed to be good for its size.

Running 1.18 Vcore and all cores synced to 42 (now 41 to avoid that BSOD)

I'm very puzzled about all these stress test programs though.

*Running Aida64 CPU stress or all the related stress tests together gets a max temp of 58 C
*Running RealBench stress test gets a max temp of 65 C
*Running various games gets a max tem of 58 C
*Running Prime95 small FFTs gets a max temp of 98 C

How is that possible? How can P95 get 34 degrees hotter than I'm able to achieve using any other stress method?

What can be the problem here? I don't get it, I really don't.
If it's just the newest version of Prime95 that doesn't want to play nice, what would be a second reliable stress test that I could use?

You could also try Intel Burn Test but it is likely to be just as hot. If it is, shut down the test to avoid damaging your CPU.

Yogi

arges86 · Nov 19, 2013

I've always trusted Prime95 to consistently get the hottest.
With a cooler designed for mini ATX your not going to achieve that hight of an OC without baking your chip.
You may want to try re-applying the thermal paste, do a good 24hr. burn in and see what kind of temps you get.

What are you using to measure your temperatures?

Som3one · Nov 20, 2013

Are you using a fixed voltage or adaptive/offset/auto?

TurboJ · Nov 20, 2013

I have re-installed the heat sink, applied new paste and the temps did not change.

I have monitored the temps with HwMonitor, CoreTemp and Aida64. All show the same temps.
Now I have been testing with 1.08V and 3.9 GHz - I still get 84C temps on P95 small FFT.
This is very strange. Haswell is supposed to be hot, but not this hot!
I believe if I changed back the stock cooler, it would go over 90C with stock settings.
Also, even when the core0 was at 98C, it did not throttle. It should at this point, right?
Which temp, BTW, is dangerous for even small durations? 100+ ?

Fixed voltage now, for testing. Normally I use adaptive, but that doesn't work with AVX tests so now I have a fixed 1.08V.

The True Spirit cooler SHOULD be able to handle 1.2V from what I hear. But now it's struggling to do 1.08V.

Now, if this is a case of very bad IHS contact, would it be possible that all cores get very close temps? I have heard that usually when the IHS contact is bad, you get one or two cores exceptionally hot.
I'm just wondering if deliding the chip would be a possibility to fix this issue.

Som3one · Nov 20, 2013

I thought Haswell throttles at 105°C.
And since the throttling is supposed to prevent dangerous temperatures, this also answers your next question.
However, I wouldn´t advice on getting temperatures that high several times.

Anyway, have you tried the stock cooler at stock settings? Maybe there really is something wrong with your cooler.

TurboJ · Nov 20, 2013

OK... I was just wondering since CoreTemp shows max temp of 100 C and it was already so very close to that and did not throttle..

I did run P95 on the stock cooler & stock settings and I remember It was reaching maybe 82 C degrees max. I didn't do the same small FFT test though, only blend.

I have today made some new discoveries: All my fans were controlled by Asus' Thermal Radar 2, and it seems the temp measurements on it are off. I saw a 13 degree difference between TR2 and CoreTemp! So my fans were not at 100% earlier, not even the CPU fan.

I also remounted the CPU cooler again and this time used some Arctic MX2 thermal paste. After this I got a full 10 degrees lower max temp!

Setting my fan profiles so that they take into account the inaccurate temp readings of the Thermal Radar system, I have achieved some more cooling performance.

Right now my CPU seems to be stable, up and running P95 small FFT now for 30 minutes with no errors at 4.1 Ghz and 1.12V. That doesn't sound so bad anymore, but max temp is still 80 C. Avg is around 74C running this test.

So volts vs. clock seems quite all right this time, but temps are a bit high still.
Of course, all other stress tests get lower temps, but still.

So how do these new numbers look to you? Cause for concern or OK?

Som3one · Nov 20, 2013

I think those temperatures are ok now.

As for the temperature readings: AISuite uses a different temperature sensor than the normal monitoring tools.
The sensor AISuite uses is positioned only "near" the CPU while the other monitoring tools use the on-chip sensor. Which is why AISuite temperature readings are usually 10°C-15°C (or even more) lower.

Search

Strange acting 4770k

TurboJ

Honorable

Y0GI

Splendid

arges86

Distinguished

Som3one

Honorable

TurboJ

Honorable

Som3one

Honorable

TurboJ

Honorable

Som3one

Honorable

TRENDING THREADS

Latest posts

Moderators online

Share this page