Should my CPU be heating up this much? Even with a brand new, powerful CPU cooler?

Something isn't right. Check your caseflow. Maybe we could see some images? With dual 680s my GPUs only reach 69*C on air. You need to verify that your fans are flowing air smoothly, preferably as much in as out, otherwise you end up with hot spots in the case.

First thing you should try is resitting the cooler with some fresh thermal compound, if you havent already. Take the graphics card and ram out, so its a bit easier to place the cooler down on top of the cpu without them getting in your way.

Sometime, APUs do give a false read-out with certain temperature monitoring programs.

Install AMD Overdrive, or Coretemp. Those have given me the most accurate core (not socket) temperatures the most amount of times, in my experience. (As a caveat, AMD Overdrive reads Thermal *Margin* - the distance you have until you reach your thermal threshold.)

If you can post some screenies, we can help you more.

Even if you did check the thermal paste, you would have to reapply it, as once the contact is removed it loses even more efficiency due to air pockets, etc.

While not 100% accurate, most software solutions for measuring CPU temp can be useful. However, if you're getting large offsets like 38*C vs 90*C, I would recommend comparing the idle temperatures of the motherboard CPU to the idle temperature of the HWMonitor or CoreTemp or whatever software you use. The easiest method is to shut the computer down after you've written down the temperature and jump directly to BIOS from your monitoring software to get the most accurate temperature. While it wont be 100% accurate, it will probably give you an idea of which temperature monitoring solution can be trusted.

Software like CoreTemp will have a temperature offset function and that will let you put in the difference. For the sake of longevity it's recommended you don't run a CPU higher than 20*C below it's maximum.

My coretemp temperature was 72*C in overclocking + Prime95 stress testing but it was shutting down randomly at that temperature. The thermal shut-off limit was 90*C or so for the FX-8350 I was testing. I've since put the thermal offset at +18*C for it and haven't had an issue when stress testing since. When overclocking you usually try to push a CPU to the edge of it's limits, and then pull back a little, until it's 100% stable. If you don't know the actual temps it can be dangerous to try to push hardware. Some may tell you otherwise. In this same sense, it's perfectly reasonable to have a negative offset. I'm using a Corsair H110 cooler right now and haven't had any temps while gaming above 65*C @ 4.8GHZ.

I have had issues however where the original paste I put down was poorly applied and I was getting excessive temperatures anyways. If you are worried about thermal paste and think you may want to reapply it toms featured an article regarding types of pastes tested. One of the higher scoring ones was Noctua NT-H1, and that's the one I've been using. In contrast, surprisingly Artic Silver performed relatively low.

Coolaboratory liquid is the highest scoring one but it corrodes aluminum and is electrically conductive, and without a successful burn-in period of 140*F which is nearly impossible on most AMD chips, the compound performs worse than the cheapest pastes.

http://www.tomshardware.com/reviews/thermal-paste-performance-benchmark,3616-17.html

As far as the GPU is concerned you would have "Safe" Temperatures up to 90*C. I don't recommend letting them run above 80*C. The only reason for this is while the chips can run stable at 100*C, the longevity of them greatly diminishes. The solder on the chips softens, and during the heating and cooling process, solder between contact points on the chipset can become a complete seperation or a hotspot, resulting in the death of the card. That said, you're fine at 76*C. The only reason I get cool temps of 66-68*C under load for my GPUs is because I have a 240MM fan blowing right on top of them from the side panel.

Either you did not read my caveat, or you did not understand.

AMD Overdrive does not tell you the temperature of your CPU, it tells you the Thermal Margin. Thermal Margin is a backwards way of telling temperatures - it tells you the difference you have until you reach your thermal threshold.

This means that your 38.1C reading means you have 38.1 degrees until you reach your maximum operating CPU temperature. If it is telling you 67C, that means you have 67 degress until you reach your maximum operating CPU temperature - so your CPU is actually *colder* when it tells you 67.

Install and run Coretemp. When you run Dying Light again; I wouldn't be surprised if your Core Temperature is actually around 40C.

Good call on the Thermal Margin. I didn't even notice it said that. Hopefully he gets back soon with his findings.

It's all good; the wonderful world of CPU thermals and the arcane mysteries of reading them is always scary rabbit hole. It sucks that a program decided to wig out on you and provide you with an inconvenience. But, on the bright side, you learned a couple new things - including that AMD Overdrive is properly calibrated to your CPU. Also, your cooler is working great, around 40C is excellent.