These i7 4770K Temperatures Are Normal?

[feelingtheblank]

Hello everyone,

I recently upgraded my good old Core2Duo rig and bought a new cpu-motherboard-ram trio. My new specs is as follows:

- i7 4770K @ stock speed + Coolermaster 212Evo cpu cooler
- MSI z87 g43 motherboard
- 2x4 GB G-Skill Ripjaws X 1866Mhz
- Case: Corsair Carbide 400R

The problem is I wasn't aware of that new Haswell cpus are running slightly hotter and now I'm little bit worried about my temperatures. Since I was also planning to overclock my cpu a bit and trying to find a point that doesn't need a vast voltage increase, I'm losing my sleeps over this situation at the moment.

Anyways. I'm using softwares like HWmonitor, Coretemp, Realtemp and Intel Extreme Tuning Utility. Turbo Boost is also active. So the cpu is going up until 3.9 GHz. Other than that I'm on stock speeds, and motherboard's default values. Here are my temperatures:

Ambient room temp: Varies between 23-25 °C

Idle: 28-30 °C

While playing demanding games like Battlefield 4: Max 58-65°C

With Intel Extreme Tuning's stress test for 15 mins: max 65-70 °C

With Prime 95 Blend and OCCT burn tests for 15 mins: max 78-82 °C

I also run realtemp's sensor test and the values are identical since it's using Prime95 too.

I also noticed that Prime 95 and OCCT is increasing my cpu voltage value from 1.156 to 1.22 while Intel Extreme Tuning's stress test is using the default voltage. Couldn't understand why. Will I encounter these kinds of random voltage increases during normal tasks? Like playing games, rendering some stuff etc..?

On the other hand I tried motherboard's OC Genie future to see what happens. It overclocked the cpu automatically to 4.1 GHz @1.15v. With this setting I've seen max of 70-72 °C for a second and mostly 65-68°C under OCCT stress test. And also my voltage didn't increase at all. I'm a bit confused about these values. Since with default settings I'm getting hotter values.

So I'm wondering if my temperature values are normal with the cpu cooler i have (Coolermaster 212evo)?

I also could buy some extra fans for my case (1 exhaust to top & 1 intake to side) and maybe a second fan for the cpu cooler if you guys think that these would help a bit.

Sorry for my English by the way. I'm not a native speaker.

Thanks for all your comments and suggestions already.

Best

 

[RaistlinZ]

Your temps sound normal to me for an air cooler.

 

[BonzaiDuck]

Only as I recall (and I'm getting old . . ) there may have been more variability in stock temperatures for the Ivy Bridge and Haswell processors. This is because Intel changed its traditional method of mating the integrated heat spreader or processor cap to the CPU silicon chip.

Instead of using an indium solder, Intel used a thermal paste (TIM) between IHS and the die. This would then mean that any slight variability in the distance between the processor cap and die, or any difference in the way the TIM were applied -- would cause different processors to show different temperatures.

Even so, your temperatures are not extraordinarily hotter than they should be, based on my readings here. Some with firsthand IB or Haswell experience would have more accurate remarks for you, but there are other factors.

Such as: The Hyper 212 EVO is a great cooler, and costs less than some others, although the others might slightly outperform it. So it's not the very best cooler you could get.

Another thing to understand is that the throttling temperature -- more or less a maximum temperature the chip can have before throttling to "save itself" -- is close to 100C. Your maximum burn-test temperatures are almost 20C below that. In normal use, they won't get that high.

There are a couple things you can do.

You can get a better cooler.

You can attempt to improve throughput of air in your case -- to lower overall case interior temperatures.

You can find a better TIM for mating the surfaces of your cooler and the IHS. You could try IC Diamond nano-diamond paste, or something like Coollaboratory Ultra or Liquid Ultra.

These things won't void any of your warranties. There are other things you can do.

You can lap the IHS to remove all the nickel plating and make it perfectly flat. [voids your warranty, but little risk].

You can de-lid the processor after studying de-lidding guides carefully. [This will void your warranty, and there's more risk.] But if you are careful, you can achieve as much as a 20C improvement over your current temperatures. But you will need to use a TIM between IHS and processor die that has a low thermal resistance. Recommended again is Coollaboratory's Liquid Ultra. The guides should give you cautions about how to avoid the risk of electrical damage, since the compound is metallic.

If you do the last two things, you should have a "pretty cool CPU."

Also, if your Hyper 212 has nickel-plating on the base, you can lap it to remove the nickel and make it flat. If it doesn't have nickel (I don't remember myself . . ) lap it anyway to make it flat and smooth. This might void the warranty on the cooler, but for the price -- who cares?! There's no risk of damage if you follow guides written by the many experienced people on this and other forums.

 

[GreenChile]

I second that. Those temps look fairly typical for that cooler.

 

[ChronoReverse]

Sounds perfectly normal. Prime95 AVX will generate tremendous heat.

 

[JDG1980]

The problem is I wasn't aware of that new Haswell cpus are running slightly hotter and now I'm little bit worried about my temperatures. Since I was also planning to overclock my cpu a bit and trying to find a point that doesn't need a vast voltage increase, I'm losing my sleeps over this situation at the moment.

Haswell runs hotter than the fires of Hell. With the stock cooler, at stock speeds, it tops out at 86 degrees. This is partially because of the process node, but also in large part because of a badly designed integrated heat spreader that has a gap (poorly filled with paste) between the IHS and CPU die. Haswell-E (the enthusiast version, coming in Q4 2013) should fix this latter problem. At the moment, though, there's nothing you can do about this short of de-lidding.

 

[feelingtheblank]

Thanks a lot for the quick replies everyone. These informations are quite helpful and it's good to know that I'm in safe zone for now.

And BonzaiDuck, thank you very much for your great suggestions. My priority now is adding 2 more fans to the case (and maybe slowing down the rpms a bit to reduce the noise) and maybe a 2nd fan to the heatsink. I will also try to reapply the thermal compound. De-lidding looks beyond my skills however.

I also found out my Bios version is old and MSI has released few more bios versions lately. But the whole process of flashing bios scares me for some reason. I'm currently not planning to dive in a deep overclocking ocean (maybe few months later). My first priority now is to get the coolest point possible as things stand. But on the other hand I'm wondering if the voltage increasing during Prime95/OCCT tests is a bios related issue.

With OC Genie feature enabled @ 4.1 GHz, cpu voltage seems to sit at 1.15 and temperatures around 70C under Prime95&OCCT burn tests. But with stock values, during these stress tests voltage is increasing to 1.21 from 1.15 for some reason. And this leads some high temperatures up to 80-82. I have Turbo boost enabled but I only boosts me up to only 3.9 GHz. What troubles me is that if I encounter these voltage increases while doing daily tasks- like gaming, rendering stuff etc.. Or maybe i'm just being paranoid since this upgrade is a bit hard earned one after 6 years

I think I will just flash my bios to the newest one to see if there's a change. But still the whole process gives me chills...

 

[JDG1980]

Also, if your Hyper 212 has nickel-plating on the base, you can lap it to remove the nickel and make it flat. If it doesn't have nickel (I don't remember myself . . ) lap it anyway to make it flat and smooth.

My Hyper 212 Evo has "direct-touch" heat pipes - that is, the bottom of the heat pipe surfaces are machined so that they touch the IHS directly. Personally, I wouldn't recommend further sanding on a heatsink like this, because there's a risk of going through the copper (who knows how thick it is?) and ruining the pipes.

 

[BonzaiDuck]

My Hyper 212 Evo has "direct-touch" heat pipes - that is, the bottom of the heat pipe surfaces are machined so that they touch the IHS directly. Personally, I wouldn't recommend further sanding on a heatsink like this, because there's a risk of going through the copper (who knows how thick it is?) and ruining the pipes.

I would agree on that issue, although you could attempt to get rid of the high spots -- if any -- on the HSF base. These direct-touch designs do offer a slight advantage. If someone were to ask, I'd recommend IC Diamond paste more than ever. You could pack it into the crevices -- maybe to some unnoticeable but positive effect.

Maybe Liquid Ultra would do the same thing, though.

thinking about it some more -- the thickness of the heatpipes would be considerably greater than any nickel plate on the base. And also -- this is really dumb -- I actually have one of those, new in the box, sitting in a corner of this desk! Nope, just the box storing a bunch of computer manuals, but from the picture, it looks as though the base itself is nickel plate.

Thinking about it again, with the direct-touch feature, grinding off the base's nickel plate is not likely to have much effect, anyway.

 

[GLeeM]

You could pack it into the crevices

Lapping is not so much removing the plating as making the two surfaces perfectly flat so that no air space or as little TIM as possible is between.

 

[coffeejunkee]

I'd say those temperatures are completely normal with that cooler and ambient.

Haswell cpu's automatically add 0.1V when running AVX loads like Prime, Linx, OCCT or Aida fpu, at least when using auto/offset/adaptive voltage. This causes them to get super hot during these already intensive stresstests.

The reason your cpu stays cooler when using oc genie is because it uses fixed voltage.

Prime/Linx loads are unrealistic and don't even do a good job of catching instability. Using a combination of real life apps works much better.

Adding casefans won't do much. Haswells high temps are caused by a combination of transistor density, non-soldered heatsink, integrated voltage controller and most importantly the 0.1V AVX bump. The problem is not getting heat out of the case, it's transfering the heat from the cpu to the cooler.

 

[nwo]

Adding casefans won't do much. Haswells high temps are caused by a combination of transistor density, non-soldered heatsink, integrated voltage controller and most importantly the 0.1V AVX bump. The problem is not getting heat out of the case, it's transfering the heat from the cpu to the cooler.

This!!!

My CM Hyper 212+ barely gets warm to the touch when my slightly OCed i7 4770k is under full load (CPU mining) @ 70*C core temp. The heat transfer between CPU and HSF is atrocious which is why lapping/de-lidding is so popular but I don't think Haswells can be de-lidded.

 

[feelingtheblank]

I'd say those temperatures are completely normal with that cooler and ambient.

Haswell cpu's automatically add 0.1V when running AVX loads like Prime, Linx, OCCT or Aida fpu, at least when using auto/offset/adaptive voltage. This causes them to get super hot during these already intensive stresstests.

The reason your cpu stays cooler when using oc genie is because it uses fixed voltage.

Prime/Linx loads are unrealistic and don't even do a good job of catching instability. Using a combination of real life apps works much better.

Adding casefans won't do much. Haswells high temps are caused by a combination of transistor density, non-soldered heatsink, integrated voltage controller and most importantly the 0.1V AVX bump. The problem is not getting heat out of the case, it's transfering the heat from the cpu to the cooler.

Thanks for the suggestions.

So is there way to keep that voltage under control with Turbo Boost? I checked BIOS but there are lots of options there and don't really know which works for what.

 

[feelingtheblank]

This!!!

My CM Hyper 212+ barely gets warm to the touch when my slightly OCed i7 4770k is under full load (CPU mining) @ 70*C core temp. The heat transfer between CPU and HSF is atrocious which is why lapping/de-lidding is so popular but I don't think Haswells can be de-lidded.

What are your temperature values nwo?

 

[bo6a]

Sounds perfectly normal. Prime95 AVX will generate tremendous heat.

 

[BonzaiDuck]

Lapping is not so much removing the plating as making the two surfaces perfectly flat so that no air space or as little TIM as possible is between.

I have to politely disagree there. Back in '07, I'd gone "hot-dawg" on refined air-cooling, ducting and TIM choices. It was about that time that Joe Citarella -- one of the founders of the Overclockers website -- started his Innovative Cooling company and introduced IC Diamond paste.

Let me elaborate. Despite various people here touting thermal greases, some containing Aluminum Oxide and other materials, IC Diamond had proven itself as a TIM made of nano-diamond particles, with diamond always known to have a lower thermal resistance than just about any other material. For practical purposes, electronic components have used aluminum (poor), copper (better), or silver (expensive but better yet). Diamond has a thermal conductivity many times that of silver, or conversely -- a thermal resistance that is noticeably lower. Generally, this also extrapolates to Carbon in other forms, but for various reasons, Citarella and others had seized on diamond early in the game. Synthetic diamond is relatively cheap in micro-particulate form. Even if more expensive than other formulations, it is economically feasible for a thermal paste.

Now, of course, diamond particles aren't the best. Products like Indigo Extreme and Coolaboratory Liquid Ultra use a formulation of Gallium or Indium, and perform better than diamond by some several degrees.

Applying this same idea of relative thermal conductivity or (conversely) thermal resistance, Nickel performs much worse than Copper. But the chip-makers and heatsink-makers chose Nickel to plate their otherwise Copper heatsinks and heat-spreaders, because its place in the "galvanic series" of metals makes it much more resistant to corrosion than copper.

I found that -- between being just "flat" with much of the nickel-plating left intact -- IHS or heatsink bases were about 2 or 3C more effective with all the Nickel stripped off. This improvement essentially doubled when the Nickel was removed from two mated surfaces -- for instance, both the IHS and heatsink base.

 

[nwo]

What are your temperature values nwo?

Between 69-71*C under full load while CPU mining @ 4.1GHz with Asus auto OC. Ambient room temps are in the low 20s. I have pretty good case cooling but my reference 7970 and Sapphire 7950 dual X definitely benefit more from it than my CPU. Running only one fan on the CM 212+. Seems like adding another would be a waste since as I already mentioned the HS doesn't even get warm to the touch. I actually have tried adding another fan onto the HS and it barely made my temps drop by 1-2 degrees. Wasn't worth it imo.

 

[BonzaiDuck]

Between 69-71*C under full load while CPU mining @ 4.1GHz with Asus auto OC. Ambient room temps are in the low 20s. I have pretty good case cooling but my reference 7970 and Sapphire 7950 dual X definitely benefit more from it than my CPU. Running only one fan on the CM 212+. Seems like adding another would be a waste since as I already mentioned the HS doesn't even get warm to the touch. I actually have tried adding another fan onto the HS and it barely made my temps drop by 1-2 degrees. Wasn't worth it imo.

No firsthand IB or Haswell experience, but there was a consensus that "auto-OC" imparted excess voltage to a CPU with stock-turbo of 3.8Ghz trying to reach 4.4 or 4.5Ghz. In your case, you're wringing a bare 200 Mhz extra speed out of your Haswell K chip. The "excess" voltage from auto-OC is likely less (if any) than for a more ambitious OC.

I would like to see how those temperatures drop under load after de-lidding and TIM replaced with -- say -- Liquid Ultra.

But the point of it: If you don't want to try wringing another 300Mhz plus out of it, it should be plenty fast and stable as is. And if you DO want to be more ambitious, then you have to do the work and submit yourself to the risk.

 

[nwo]

Yep, overvolting is definitely a big part (and con) of the auto OC feature.

I bricked my 4670k when I tried to take it above 4.5, so I'm happy with 4.1 (for now)

 

[BonzaiDuck]

Yep, overvolting is definitely a big part (and con) of the auto OC feature.

I bricked my 4670k when I tried to take it above 4.5, so I'm happy with 4.1 (for now)

I'm still tweaking my 2600K-with-Z68-chipset. I've been "watching" now through IB and into Haswell. The only way to "do" these newer CPUs is to de-lid them. It all looks manageable, and you might do as much work lapping a heatsink or IHS (which you would also do). There are enough online "guides."

These days, I just keep asking " . . . but do I NEED another computer?" I might let the "WANT" side of my brain get the better of it -- after I find a tenant for a property I own . . . and start collecting the rent again . . .

 

[tweakboy]

Those HSF are noisy. Water cooling you can adjust and take the rpm low, very low noiseless almost.

 

[feelingtheblank]

I can't really understand why Intel decided to use cheap TIM in their newer processors. It's really weird. Can we just send our CPUs to RMA and make them de-lid for us?

 

[coffeejunkee]

Thanks for the suggestions.

So is there way to keep that voltage under control with Turbo Boost? I checked BIOS but there are lots of options there and don't really know which works for what.

Two options:
1. use fixed voltage (but you give up voltage drop during idle/eist)
2. don't run AVX loads.

Another thing, check if multicore enhancement isn't active. This option runs all cores at single core turbo boost freq. and is often active by default. The speed boost is actually nice, but the board will increase vcore as well. Most likely completely unnecessary but board manufacturers like to play it safe.

 

[feelingtheblank]

Two options:
1. use fixed voltage (but you give up voltage drop during idle/eist)
2. don't run AVX loads.

Another thing, check if multicore enhancement isn't active. This option runs all cores at single core turbo boost freq. and is often active by default. The speed boost is actually nice, but the board will increase vcore as well. Most likely completely unnecessary but board manufacturers like to play it safe.

Thanks for the suggestions.

I haven't checked if multicore enh. is active but probably it is since under full load, the freq. of all cores were going up to 3.9 GHz.

I will probably play with the voltages a bit and find a stable frequency around 4.1-4.2 Ghz. But need to update the bios first i guess...

 

[coffeejunkee]

Thanks for the suggestions.

I haven't checked if multicore enh. is active but probably it is since under full load, the freq. of all cores were going up to 3.9 GHz.

I will probably play with the voltages a bit and find a stable frequency around 4.1-4.2 Ghz. But need to update the bios first i guess...

In that case mct is active and your board is most likely overvolting.

Stock vcore should be somewhere around 1.050 i think, at least it is for my Ivy. Haswell may be slightly less. With mct it's probably around 1.090-1.100 or so which will add some 5-10 degrees.