Why is my i5-4670k so hot at only 4ghz?

[DigitalWolf]

Only thing I'm deducing from the context of this thread is that haswell might not be OC limited in the same way Ivey was/is.

My personal experience is more like a lot of the posts I read on another OC site. I get to a certain point and no amount of voltage manipulation will result in a stable OC. While my thermals are perfectly fine... oh and yes for reference I do use a water loop.

This isn't a case of I cannot give the chip more voltage due to heat.. it just won't be stable past a certain point.. period. (I see a lot of similar posts).


Now if you want to see a thermal show (lol) then they should run the Intel optimized Linpack. Just leave the settings on auto (especially the vcore) and see what the temps do... (you won't see the heat extreme if you have figured out a good fixed vcore to use).

 

[Jodiuh]

Ok, so w/ everything set to auto, the Maximus VI Hero ran 1.264V through my 4670K and had a turbo of 3.8Ghz. It seemed to load @ 76C in AIDA64's stability test.

Ok, so I followed the ROG guide and toned everything down manually. CPU-Z now reads 1.7V, lol. But AIDA64 reads 1.149V, exactly what I set it to...so screw CPU-Z for now. Core temp's temps didn't match AIDA64 either, so I threw that out.

Right now I've been through 11 minutes of this test and my max temps 69C on 3 cores and 66C on the last. This is @ 4Ghz w/ a Thermalright Venomous X and a Scythe SFLEX E 1200 RPM fan. Case is a the 550D w/ 2 SFLEX E's for intake, and a single exhaust. All fans run @ 1200RPM and the intakes have filters.

In the beginning, I was really hoping for 4.5Ghz, but due to all the crazy options on this board and none of the normal tools working well, I would be happy w/ 4Ghz.

Does this seem in line w/ typical 4670K's?

 

[CakeMonster]

*edit* I should add that I am assuming from what I read you are using a fixed/override vcore setting. If you are using adaptive then my above humor wouldn't be accurate as it will jack the vcore in those tests.

I was puzzled by the fixed/override vcore options on my Gigabyte Z87-D3HP with Cooler Master 412S HSF. I basically had to turn it to fixed because it would jack up voltage (and thereby temperature) by crazy amounts if I didn't. Even below 4GHz. But running it on fixed also means it won't turn down the voltage when its not needed.

I might have missed something but seems I'm stuck with one fixed voltage. I resent the fact that I can't set a ceiling for it. I ended up at 4GHz and 1.16v. If I increase the voltage it gets unstable and 90C+, and if I increase the speed it freezes.

 

[Jodiuh]

I noticed this too when setting the voltage to offset. It would skyrocket to 1.35 when I had it set to 1.15.

 

[Kenmitch]

I call it Haswell hell. The added .1v or so during heavy loads when using offset or auto voltage is a killer. Combine it with the thermal lotto and a lot of chips are SOL.

Running fixed vcore doesn't look to effect the electric bill too much tho as far as I can tell at idle. Hasell will suck down the watts at higher clocks tho.

Non scientific results with my 4670k using fixed vcore values.

Open Hardware Monitor looking at the Powers results Idle readings all energy savings settings still enabled in uEFI.

2.4ghz @0.739v

CPU Package 9.8w
CPU Cores 0.7w

VS

4GHz @ 1.046v

CPU Package 9.9w
CPU Cores 1.7w

 

[Jodiuh]

I also noticed that. It doesn't seem like I'm being penalized much for having manual voltage set constant. The speed is still throttling, though, that doesn't seem to matter from your numbers. My i5 760 was 50 watts hotter wo/ all the EIST/C1E business. Haswell doesn't seem to use much electricity up there, so that's a plus.

 

[Concillian]

My 4670k was hitting mid 80s during Intel Burn Test at 4.25GHz with a Noctua at full voltage. I haven't checked the AVX2 version, I bet that would put it to near throttling.

I'm not sure what the value of that is, though. AVX isn't really used by anything else.

Finally de-lidded last night and same setup with Noctua fans at low voltage / near silent I hit 70C for a moment... so gained quite a lot of headroom. Didn't try full voltage on the fans, but that'd probably be over a ~20C drop. That's with Liquid Pro (silicon to heatspreader) and Noctua paste (heatspreader to heatsink). Overall, this was more than I got from a de-lidded Ivy.

I haven't investigated adding voltage yet to see how this will affect ultimate overclock I end up at. Previously I had tuned 4.25 GHz & 2133 10-12-12 to edge of stability and turned it down to 4.1 GHz & 2000 at same voltage as my 24/7. Last night it passed 4.3 / 2000 at same voltage, but failed 4.4. I don't really need mass CPU performance so may just stick with my previous 4.1 and just chalk all this up as an interesting experiment.

Power consumption was 190W at the kill-a-watt during AVX linpack at 4.3 with an old Seasonic 80+ (no bronze, gold, etc... just 80+) PSU that I have in my bench test case. Actually is only a few watts different at 4.1 Really doesn't change much unless you adjust voltage.