How best proceed with overheating i7-4790K?

[REnso1]

I have an i7-4790K in a GA-Z97MX Gaming 5 mobo latest F4 BIOS.

I used a Noctua NH-L12 cooler rated at 95W for cooling the processor rated at 88W in a Lian Li PC V354 with 4 fans, 2 in 2 out, case closed and case open, ambient is 27-30°C.

Temperatures in BIOS and memtest86+ were high so I decided to try stress testing and in Prime95 small FFT cores 1&2 overheated to 100°C using Core Temp.

I tried reseating the heatsink and renewing the NT-H1 TIM and opening the case but it made no difference. I have a photo of the contact pattern here.

When I tested using the OCCT benchmark I was unable to complete a test due to the processor overheating so I underclocked the processor to 3.6 GHz, disabled turbo and manually set vCore to 1.1v.

With an underclocked processor I was able to get a heating and cooling curve using the OCCT auto capture, to enable me to study the problem.

Even when underclocked the processor was reaching high temperatures, rapid fluctuations in temperature with work load suggest a bottleneck in the thermal pathway. When I tested with the intel retail cooler which came with the CPU the cooling was much less effective than the NH-L12 (even when underclocked taking just over a minute of OCCT to reach the 85°C cut off point see below) indicating the NH-L12 was doing a good job of removing heat, which meant the processor was making the heat or the source of the bottleneck.

I have discussed it http://forums.hexus.net/cpus/327593-4790k-overheating-nh-l12.html elsewhere. Advice was to contact Intel due to an absence of information relating to my retailer's testing procedures. I have asked about these but am still waiting for a reply.

So my question is how should I proceed from here? Does this qualify for an RMA? If so is it possible to negotiate this with Intel direct or do I have to go through my retailer?

I have done my best to make sure I am not doing anything wrong and I would be grateful for any pointers to any mistakes I may be making.

[SPark48]

Still can't get it to show TDP, but I did lower my current limit to 255A and here's what I got:

No throttling, not even 1%. This is on ASUS Maximus VII Formula with ken-settings + overclock to x48, BCLK 102.1, and VCore 1.375.

And that's what I would have thought about TDP. It really shouldn't matter that you're not getting max TDP as long as you have normal temperatures and your CPU is not throttling. In fact, it should be a good thing that your processor can reach it's max frequency without your TDP getting to max, since once it gets to the max, if it still needs more power, it will throttle.

Did I just get really lucky with a CPU/MB combo that isn't having too many of these problems? Or is there something I am doing that I can help you guys with as far as settings? This is a (expensive) hobby for me, but I am by no means a professional with all this, so I wish I could tell you guys exactly what I am doing to make it work. Let me know if you want me to test anything else. In the mean time I'm going to try to figure out why the TDP is not showing up in XTU.

[RChiz]

Woolphy, I had the same thing happen. 255A, current limit throttle, temps reasonable. 256A, no current throttle but thermal throttle as temps shot up to upper 90s.

[TVere1]

Hi Guys!

Woolphy írta:

Hey sampark1980,

I don't know why it doesn't show up in graph because it should, but try to click the little wrench on the right side, maybe you can find the parameter there.

it would be interesting to know your TDP while running 48x multiplier on 4cores.

Did you set the 48x in XTU or in Bios?

I thought to try it out too, but with the XTU, and when I applied the changes my computer crashed. I didn't set anything else ( current TDP ect.) everything was on auto mod.

And how did you manage to lower the temperature on 4,9 GHz? What kind of cooler do you use?

I did a sterss test on 4.6 GHz ( CPU TDP : 95 W , CPU Core Voltage 1,2190 V , CPU Temp : 67 Celsius ) and a have a Dark Rock Pro 3 cooler. What can be my problem? Can it be that I have a faulty cpu?

[SPark48]

The 48x was set in BIOS. All of ken's setting's were also set in BIOS as well. Strangely, even if I put TDP limits in BIOS at 88W boost power and 110W short power, in Intel XTU, it shows up as 704W and 880W respectively. I changed these manually in XTU as you can see in the pictures I am posting.

I did some testing with my current limit and found something interesting:

The above is with current limit set at 166A. As you can see, there is current limit throttling with max of 61% throttle. If I set it at 167A, I get max of 15% throttle.

Now, same test above with current limit of 168A:

As you can see on the screen that I am pointing at, I had one instance of 1 throttle, but otherwise pretty clean. Average temperatures were also 2C lower than with current limit set at 256A. Apparently my magic number is 169A to get through the XTU stress test without any throttling. Everyone else seems to be at 256A. Why the difference?

Also, if you can find my previous posts, at stock speeds (44x x 4 with ASUS MB) my current limit was around 161A before seeing throttling.

[SPark48]

Sorry, forgot to answer the rest of your question:

My original processor was horrible. I was getting 100C+ with stock settings and stock cooler while running Intel XTU as well as other stress tests. I got an RMA and this second chip seems to be doing much better. Here are the steps I took to get my temperatures in low 80C's with 4.9Ghz:

1) Got RMA for new CPU

2) Clear CMOS

3) Set Ken's settings through BIOS except current limit to 256A (although as seen above, I've just lowered it down to 169A and still got 4.9Ghz with no throttling)

4) Ran tests at stock speeds to find temperatures in 50C's at max load

5) Gradually increased multiplier to 48x

6) Gradually increased VCore until stability was reached at 1.375V

7) Increased BCLK frequency to 102.1 (to get to 4.9 GHz)

8) Set CPU Input Voltage at 1.9V

9) Set Cache Voltage to 1.25V

10) Set RAM Voltage to 1.65V (as noted on my memory)

After that, I ran stress tests to ensure stability and that temperatures never went above 83C. In fact, 90% of the time temperatures are no higher than 77C, but during short bursts throughout the stress test, it will go up a little higher, but max was 83C. Average temperature was 78C. This is with XTU, AIDA64, and OCCT. For now, I'm staying (FAR) away from Prime95.

Cooler is a Corsair H60 with 2 fans in push-pull configuration. Surprisingly not a very great cooler by any means... I am using IC Diamond as my thermal paste. Case is a Corsair Air 540 with 5 additional case fans other than the 2 on the H60.

Hope that helps.

[RChiz]

Wow, there is a lot of weird stuff going on. As for TDP, the low values some are getting doesn't seem to make sense. I mean power is a function of voltage and current. Unless something is way off with voltage or current, you should be getting close to 88W or more under full load. You would have to be working at sub zero temps for the impedance to affect the current to lower power that much. Maybe using the benchmark would be a better tool to ensure the CPU really is working correctly.

[SPark48]

slaveonepunx,

I think you are correct, but the difference between what you're saying and I'm saying is:

You: "Something is wrong with processor X unless it reaches max TDP of 88W"

Me: "Something is wrong with processor X if it goes OVER the max TDP of 88W"

If my processor is getting to it's max frequency without throttling and I am not going OVER the 88W max TDP, then I should be OK, right? That's how I've been understanding it. I may be wrong.

Here is my benchmark. Unfortunately, there is no way for me to know what my TDP was while running it because my XTU for some reason doesn't show it... which defeats the whole purpose of why you asked me to run the benchmark... Still trying to figure it out.

[TVere1]

Thanks sampark1980 for your answers !

But I'd like to share some more facts and question.

First is the case of Intel's XTU. I use the latest version of the program, and when I click on the Settings menu in the top right corner the program stops working ( quite funny heh ), and some older version offer much more setting in the manual tunning section with my mobo ( btw it is a GB z97x gaming 5 ). So it's a strange issue for me. I don't know wheter anyone else experiences this.

Secondly I still have some issues with the settings given us by ken-intel. I did the changes in the Bios, then I ran XTU stress test with turbo boost enabled ( I didn't try with disabled option ), and I noticed that the cpu utilization fluctuates between 98-100 percent. And my system was also unstable dispite that I raised the cpu vcore voltage to 1.15 V ( maybe it wolud be stable without turbo boost, I dont know ), so I changed the Vcore setting back to auto. And now I don't have stability problems. For the 4,6 GHz I needed to raise the turbo boost power max because of the power limit throttling, and the cpu utilization is a straight 100% ( on all turbo frequencies, the fluctuation disappeared ).

About the RMA. I bought my cpu in late September, and encountered the overheating a few days later ( of course while testing a game ), so the chip ran on 100 C without any warnings from the system. I used the stock cooler that time. Now my new one keeps the temperature on moderate levels, but I don't know how I could make sure, that my CPU is not faulty. Any suggestions to that?

I'm not an expert on this field, so if I said something stupid, I am sorry for that.

And I am really thankful for all your help!

[SPark48]

Your CPU utilization fluctuating between 98-100% is normal. This just means how much the XTU program itself is using from your CPU. If you had absolutely no other programs running in the background, this may stay at 100%, but because you have other programs (operating system, maybe chrome/iexplorer, other system apps) running in the background, you will be in the 98-99% range most of the time. When you say your system is not stable, what do you mean? Do you mean you have a hard time opening up programs or doing other things on the computer? Because if this is what you're talking about, this is expected also. Since XTU is taking up 98-99% of your CPU during the stress test, that means only 1-2% of the CPU is available for you to run other programs, so it will slow down your computer significantly. This is the case for all stress tests, not just XTU and is completely normal. If you mean unstable as in your computer gives you BSOD or shuts off, then there is a problem. So all of those are OK.

Now, if your temperatures are going over 85C (some people want lower, but for overclocked i7, I think 85C is OK), then you are having overheating issues. To check if your CPU is faulty, run the Intel Processor Diagnostic Tool (PDT) and if it passes, run XTU or other stress tests to ensure your temperatures are under 85C. At stock with stock cooler, I would say under 78C is where you want to be.

Hope that helps you.

[BSant8]

slaveonepunx wrote:

Wow, there is a lot of weird stuff going on. As for TDP, the low values some are getting doesn't seem to make sense. I mean power is a function of voltage and current. Unless something is way off with voltage or current, you should be getting close to 88W or more under full load. You would have to be working at sub zero temps for the impedance to affect the current to lower power that much. Maybe using the benchmark would be a better tool to ensure the CPU really is working correctly.

Slaveonepunx,

I might be wrong but 88w TDP is not the actual power that the CPU consumes (like a 2000w vacuum-cleaner which at full speed will consume 2000w / hour). It is the nominal max value of the dissipated heat from the CPU. So, while it is true that the higher the voltage on the CPU, the more actual power it consumes and the more heat it generates, it's given more as a cooling parameter. From what i read in the last couple of hours, the better you cooling solution is, the lower your TDP will be. So, not reaching your TDP is actually a good thing.

[TVere1]

Hi!

By unstability I meant BSOD

What can I do in that case?

[SPark48]

Exactly. I agree with what Woolphy said. slaveonepunx, what you are saying is that something may be wrong with our processors because it isn't going anywhere near the max TDP of 88W, especially with your ASUS Maximus VII MB. I think as long as your temps were OK and your CPU frequency was not throttling, then your TDP was better with the Maximus than with the MSI board you are using now. If you were throttling, you may have to adjust your voltages and current limits until you are no longer throttling. Doing this COULD increase your TDP, so with your Maximus, you have some wiggle room to adjust until your TDP hits a max of 88W. With MSI, you really have no wiggle room, since you are showing a max of 88W already, in which case, you would need to get a better cooler and/or lower voltages.

[SPark48]

Do you remember what the BSOD code was? If you were at stock settings, resetting CMOS and then putting everything to auto SHOULD take care of everything, depending on your board. In your case, if you're overclocked at 4.6GHz, that means you need to increase CPU Voltage. Set it to manual in BIOS and increase it little by little until you don't get BSOD. I started at 1.10V and increased it by 0.05V each time until I could run a full 15 minutes of AIDA64 stress test without BSOD. While doing this, KEEP AN EYE ON TEMPERATURES. If it goes into high 80's, you need to stop increasing voltage and instead decrease multiplier to 45x or if possible, get a better CPU cooler to keep temperatures within normal limits.

[RChiz]

Sampark1980, while I do think the lower the TDP the CPU shows under load the better, it has to be within reason. With same settings, my Asus board was showing CPU total TPD of 45W or less with XTU and temps of 85C with my sweet Swiftech H220-X water cooling. Just ran XTU on MSI board and reached TDP of between 70-78W and peak temp of 68C with my very old Hyper 212+ (not Evo) and cheap terrible TIM.

I'm pretty sure the 88W Intel rating is based upon thermal requirements for real world situations and not synthetic benchmarks and stress tests and that during stress tests it is expected to meet or exceed this number. It is a numeric value to help determine requirements for CPU cooling as well as enclosure size, airflow, etc.

So what I'm saying is that a CPU that shows 75W under the same load is better than one that shows 90W. However, if a CPU shows 45W, something is wrong unless using phase change cooling.

where C is capacitance, f is frequency, and V is voltage.

[SPark48]

Right, but as you said, you're getting 45W with the Swiftech H220X water cooler while getting 78W with Hyper 212. I may be completely wrong about this, but that means your Swiftech and ASUS Motherboard combo is cooling your CPU by 43W less than the maximum TDP and the Hyper 212 with MSI is cooling CPU by 10W less than maximum TDP. Is there any way to test this by switching the coolers around? Put Swiftech with your MSI board and Hyper 212 with ASUS board and see what you get.

Read this article: http://arstechnica.com/gadgets/2013/01/the-technical-details-behind-intels-7-watt-ivy-bridge-cpus/ The technical details behind Intel's 7 watt Ivy Bridge CPUs | Ars Technica. Especially the 4th paragraph. It's about Ivy Bridges, but the idea behind TDP is the same.

It has nothing to do with performance or how your processor is functioning. I honestly don't think there is anything wrong with your CPU running at 45W as long as you're getting full 4.4GHz with turbo on your ASUS Maximus.

What is a problem and more on topic to this thread is the fact that you're getting 85C on your ASUS MB at stock settings and a nice cooling system, while your MSI is giving acceptable temperatures with less cooling. In either case, I would not be happy with any of those temperatures if I was running stock settings. Of course it is different case by case, but with that Swiftech cooler, I would expect stock load temperatures in the low to mid 50's. Even with Hyper 212, if you're getting 78C at stock settings, then that doesn't give you much room to play with it. For me, it would be the temperatures that I would worry about, not the TDP.

[RChiz]

I am going to be installing the Swiftech tonight on the MSI and will test temps. I can't test the Asus because I no longer have it.

Took it back to Microcenter last night and asked to speak to a manager. I told him that I had a defective motherboard but I had bent some pins putting the socket cover on incorrectly and was hoping he could give me the same $40 bundle discount if I bought a replacement board. He said he would allow and to pick out a replacement. So I brought the MSI Gaming 7 back to customer service and the manager starts doing the override. I catch him giving me a full refund and stop him saying I screwed up and bent pins and was just hoping for $40 bundle discount. He told me he understood but was going to give me full credit because I was honest and admitted to the bent pins. He says no one ever admits and tries to say they were like that. He was rewarding me for doing the right thing. Wow! I thanked him profusely and walked out the store.

I'm sure I can have lower temps with the Hyper 212 if I lowered the voltage as well as using better TIM since I ran out of the good stuff and am using some cheap white stuff while saving the last of the swiftech TIM for the reinstall. This is the old Hyper 212 with the separate heat tubes on the block, only two of the four heat tubes come close to covering the die.

[SPark48]

Awesome! That's what I love about Micro Center and will never go anywhere else go buy CPUs and MBs. Of course the $40 bundle discount helps, too... =P It looks like your positive karma worked out for you!

How is that MSI board? Looks pretty nifty. How does it hold up to the quality of Maximus VII Hero?

Anyways, check your TDP again after installing the Swiftech cooler onto the MSI board with 4790K. My wager is that it will give you TDP near what you got with the Maximus VII or at least in the 50s.

I hope you have better luck keeping the 4790K cool with the MSI board and Swiftech cooling!

[RChiz]

This is with stock settings. Looks like the MB changes some power limits but everything seems to be within Intel specs. Compared to what I was getting, I'm content for now. I can even run IBT now and stay in the 80's.

As for how the MSI G7 stacks up to the ASUS M7H, I think its a toss up. The UEFI on the ASUS is more mature and I like navigating it more but I do find some values easier to get to on the MSI. I like that I have 8 x USB3.0 and 2 x USB2.0 on IO plus additional headers as I feel the Hero was lacking. The USB3.0 header on MSI is also 90 degree which is nice for cable tidiness. I also like the placement of the onboard power and reset switches because the H220-X reservoir hid them on the M7H. The debug LED also shows cpu temps by default if no error codes which is nice. I think I like the sound better on the MSI also except that the EMI cover for the chip glows more pink than red. Haven't tried the onboard voltage check points yet which may be useful in the future. My biggest gripe on the MSI is the poor fan control via UEFI or using the provided software and only having 3 system fan headers plus two for CPU. Second gripe is lighting color. Audio trace glows red, audio EMI cover glows red but looks pink, power and reset switch glow green, MSI logo on PCH glows white and various onboard surface mounted LEDs flow bright blue. As someone with a windowed case, the ASUS looks better with its uniform color scheme.

Both are great boards. If I had to do it over, I would probably have just gotten an AsRock Extreme6 but they were out at Microcenter. My second choice would be the MSI board. If I knew I could get a Maximus VII Hero with no problems, I would go that route first but reading various forums, it just has too many issues.

[dadi]

I think there is a bug in the maximus vii hero firmware, it multiplies by 8 the values for Turbo Boost Power Max and Turbo Boost Short Power Max; i used in bios 11 and 14 (can;t use 13.75) and the xtu reads as 88 and 112 W. I posted a photo here : /docs/DOC-23517# comment-16217 https://communities.intel.com/docs/DOC-23517# comment-16217

[SPark48]

Ummm... why would you put Turbo Boost Power Max and Turbo Boost Short Power Max at 11 and 14, respectively? Those have to be at 88 and 110, respectively, according to Intel's recommendations. I'm not sure how low those can go, but I doubt the computer could even turn on with 11 and 14. Can I ask where you got 11 and 14 from?

[dadi]

The values (704 and 880) looked a bit to uniform to me, and divided by 8 you get 88 and 110. I put 11 and 14 in bios and the pc is running just fine, and the xtu reads 88 and 112 W, see the photo, i guess you did not read the post linked above

You can test yourself, worst case it won't boot and you can change it back to what you want. But trust me, I'm using the pc now and it's fine