Processors
Intel® Processors, Tools, and Utilities
14403 Discussions

i7 4790K temperatures

idata
Employee
27,131 Views

Hi!

I have a question about the temperatures that I'm getting for my 4790K processor. Like some other people I'm getting extremely hot temperatures when testing the processor at full load.

I am using a Gigabyte GA-Z97X-UD3H motherboard in combination with a Thermalright True Spirit 140 cooler in order to get lower temperatures than the stock Intel cooler.

The temperatures are measured with HWMonitor, RealTemp and the Gigabyte System Information Viewer app. The last one shows slightly lower temperatures then HWMonitor and RealTemp.

The room temperature is around 19°C.

Idle I'm getting temperatures between 20°C and 26°C (depending which core). The Gigabyte app shows even a temperature of 18°C. So far so good.

In what follows I'll continue with the Gigabyte app temperatures, as the graphs are made with the Gigabyte app. Remind however that HWMonitor and RealTemp are showing higher temperatures (around 6°C higher).

As soon as I use Prime95 with the small FFTs test (latest version) to put the cpu under load, the cpu temperature increases instantly to temperatures around 92°C (HWMonitor is even showing 100°C from time to time).

In the graph below you can notice 2 temperature peaks. The first one (lower one) is achieved by running the Prime95 Blend test (max temperature after 3 minutes around 57°C), the second one is achieved by running the Prime95 small FFTs test as described above, with temperatures above 90°C.

The test is done immediately after a cold boot.

The cpu-cooler is at startup running at 550rpm, during load it increases to 1250 rpm.

Another thing I've noticed is that the temperature increases instantly from 18°C to 90+°C, within 2 seconds. As soon as I stop the test the temperature decreases instantly as well: within 1 second the temperature is decreased till 30°C. Next it takes 2 minutes to further decrease to 19°C (idle). This behavior looks very strange to me.

What is going wrong? Why do I get such a high temperatures during load? 90+°C without overclocking is dangerously high. I don't dare to overclock with such a temperatures.

Many thanks for your help!

Edit:

I've done the test with the default BIOS (F4) and the newest BIOS (F7) from Gigabyte.

Edit 2:

In the meantime I was able to do a stress test with the Intel Processor Diagnostic Tool and the Intel Extreme Tuning Tool (10 minutes). Temperatures increased until 71°C, core voltage increased until 1,269V (measured with RealTemp and HWMonitor). These temperatures are better than with Prime95, but still very high.

Edit 3:

After running Intel Extreme Tuning Tool for 1 hour, the max temperature was 73°C. In average the temperature was below 70°C. The idle temperature is around 25°C regarding the Intel tool. Room temperature should be around 19°C.

1 Solution
idata
Employee
20,099 Views

Hi all,

Since yesterday I have a new CPU (RMA). I've done some (short) tests with Prime95 small FFT's and I can confirm that based on these tests, the max temperature seems to be 10°C - 15°C lower. Without any BIOS adjustments I'm getting temperatures beginning 80°C. I have to admit that besides the new CPU I'm also using other thermal paste (Arctic Silver 5), but this shouldn't change the temperatures a lot: maybe 1°C or 2°C.

Next I've applied the recommended Intel settings from Ken (Intel support), which can be found back in the topic mentioned before (/message/260210# 260210 https://communities.intel.com/message/260210# 260210). You can find a summary of the proposed settings in this official Intel document:

2 changes I've made on my Gigabyte UD3H mobo are:

  • I've put the VCORE voltage on 'normal' instead of manually adjusting it (because otherwise the voltage remains all the time the same instead of dropping when idle). Instead I've used a negative offset of -0,060V. This seems to run stable, however I didn't do long duration tests yet. But I didn't have any system crashes during testing yesterday evening, nor in idle, nor in stress. It might be even possible to decrease the negative offset even further (-0,070V or maybe even -0,080V).
  • I've activated the XMP1 profile of my RAM. This could increase the power usage a bit (and hence also the temperatures), but since I am using Crucial Ballistix DDR3L memory running @ 1,35V (even when activating the XMP1 profile!) I don't expect any (significant) negative effect.

After applying these changes I've notice that the CPU speed is limited to something between 4,10 and 4,20 Ghz instead of the default 4,4Ghz (with default BIOS settings). These new values are actually correct, as 4,4 Ghz should, regarding Intel specs, only be achieved when only 1 or 2 cores ares stressed, not when all 4 cores are stressed as with Prime95 Small FFT's. (I didn't do any single core stress test yet.)

The temperatures I'm getting with these settings are even better. After running Prime95 with small FFT's for a couple of minutes, I'm getting stable temperatures around 64°C!

Notice this is achieved with an aftermarket air cooler, not with the stock cooler. I am using a Thermalright True Spirit 140 BW cooler.

I will do some more testing and if everything seems to be stable, I'll forward my settings to Ken (Intel support) so they can add them to the official documentation mentioned here above. I'll also mark the link to this thread as 'correct answer', but keep in mind that it's a combination of a new CPU with this settings which solved my problems. Without the new CPU I would expect that the temperatures would be 10°C - 20°C higher.

View solution in original post

0 Kudos
48 Replies
Allan_J_Intel1
Employee
16,170 Views

Thanks for joining the processor community.

I am currently researching on this issue. As soon as I can, I will send you a message with my findings.

Thank you for your patience and understanding.

Allan.

0 Kudos
idata
Employee
16,170 Views

Thank you for your reply!

In the meantime I did some additional testing after playing around in my BIOS. I will put my findings here, maybe they can help you.

I have put the VCORE setting in the BIOS to 'normal' and next I could change the VCORE OFFSET setting. I have changed it into a negative offset value of -0,080V. By doing this I could reduce the voltage during load (and also idle of course) and hence reduce the temperatures with around 10°C. That means that the VID voltage, measured by HWMonitor during the Intel Extreme Tuning Tool stress test, was reduced from 1,269V to 1,189V, and the temperatures went from max 73°C to max 63°C. Also the temperatures during the Prime95 small FFTs test went down a bit, from around 100°C to beginning 90's°C. The voltage during this test was 1,244V instead of 1,3+V.

The idle temperatures remained more or less the same, but the idle voltage decreased as well till 0,647V.

I'm actually not sure if my pc is running stable enough right now with these changes: therefor I should do a long (stress) test. I'm also not sure if this 'undervolting' is good for my system. And I don't think that this is default behavior that one must do in order to keep his/her cpu cold enough.

 

Besides this is 90+°C during prime95 still a very high temperature I think.

Many thanks for your help!

Edit:

There are 2 more things I've noticed during testing and monitoring the voltages and temperatures:

  • I have never seen the CPU running @ 4,0 Ghz. During load it's constantly running @ 4,4 Ghz and idle the clock speed is going down till 0,9 Ghz. It's varying between these 2 upper limits but honestly I didn't see the default clock speed of 4,0 Ghz yet.
  • I have never seen the CPU throttling, even not when the temperatures almost reached 100°C during Prime95. Maybe when they exceed 100°C, but I don't dare to test this to avoid cpu damage.
0 Kudos
Allan_J_Intel1
Employee
16,170 Views

Thanks for the update. I am sure this information is helpful in order to isolate this matter.

Our engineering team is requesting some screenshots of the programs you are using to monitor processor temperature. Would you send some pictures of them? This would be really important.

Thank you

Allan.

0 Kudos
idata
Employee
16,170 Views

Dear Allan,

Many thanks for your time!

I have created 3 screenshots with some of the programs that I'm using to monitor temperatures, voltages, cpu fraquency, ...

I have reloaded the optimized defaults from my BIOS, so all voltage regulation is handled by my motherboard and the cpu.

The first screenshot is taken when the cpu is idle (ca. 5 - 10 minutes after a cold boot). Watch the low temperatures (room temperature is ca. 18-19°C). I don't know if the measured temperatures are even possible (maybe the sensor is not working correctly?).

The second screenshot is taken when running the Intel Extreme Tuning Tool stress test for ca 4 minutes.

The last one is taken when running Prime95's small FFTs test. Watch the extreme temperatures (when the voltage regulation is done by the motherboard, you can see the cpu throttling in the screenshot) and the voltage which is used (1,3+V). This is after only 20 sec of running the test.

I hope these screenshots will help you. Please let me know if you need more details.

0 Kudos
Allan_J_Intel1
Employee
16,170 Views

Great, thanks for the pictures. I will forward them to our engineering team.

Thanks for your patience.

Allan.

0 Kudos
idata
Employee
16,170 Views

FYI:

I have removed my processor cooler this evening 2 times, cleaned the cpu and cooler with denatured alcohol and re-installed the cpu-cooler again. I have used different amounts of cooling paste, but I can't get the temperatures down. In contrary: the first time I had temperatures which were 4°C higher, the second time the temperatures were more or less as before (1 or 2 degrees higher maybe). So a badly installed cooler doesn't sound very likely.

I've also used the tool IntelBurnTest to test the temperatures, and with the stress level set to "standard" I'm getting temperatures mid 90's degrees Celsius. When I put the stress level to maximum I'm getting temperatures around 100°C and the cpu starts to throttle.

0 Kudos
Allan_J_Intel1
Employee
16,170 Views

Thanks for the update.

Allan.

0 Kudos
BSant8
Beginner
16,170 Views

Hello all!

I'm having the same problem with an i7 4790k processor and a asus z97 maximus hero VII motherboard. I'm using a Corsair H80i closed loop cooler. The processor runs at stock speeds with AUTO mode in BIOS. I was wondering if a solution was found yet.

Thank you!

0 Kudos
idata
Employee
16,170 Views

Hi Woolphy,

I am still waiting for a reply from Intel.

@allan_intel:

 

What's the status of the investigation?
0 Kudos
Allan_J_Intel1
Employee
16,170 Views

This problem has been escalated to our engineering department; any information on this topic would be updated at the thread below:

/thread/54032?start=30&tstart=0 https://communities.intel.com/thread/54032?start=30&tstart=0

Allan.

0 Kudos
idata
Employee
16,170 Views

Good morning Allan,

I was following the thread you've suggested to me the last weeks. But since there is still no solution or reaction from the engineering team, I was wondering if you could give me a status update. I am still using my pc as some kind of test system right now, but it actually should be my production system. I however don't dare to start cpu-intensive workloads due to the temperature issues. What do you suggest me to do? I could request an RMA but if there is a structural problem with these cpu's (what multiple sources seem to suggest) this will not be a solution.

Many thanks for your help and support!

Kind regards

0 Kudos
Allan_J_Intel1
Employee
16,170 Views

I still do not have an update for this topic.

Any information on this matter would be updated at the thread below:

/thread/54032?start=30&tstart=0 https://communities.intel.com/thread/54032?start=30&tstart=0

Allan.

0 Kudos
KMcke2
Beginner
16,170 Views

I also have this issue. Processor works fine but I can't keep the load temperatures remotely under control even with an aftermarket cooler and arctic silver. As you say it jumps from room temp to near tjmax within about a second after starting prime.

When running with a few less threads to avoid thermal cutoff, still huge temps on the cores but honestly the heat sink isn't particularly hot, I can touch it, just mildly warm at best, so could it be the temps being incorrectly read / not actually conducting well onto the top plate of the processor? I mean a jump from like mid 20's to 90 in a second is pretty insane even for these sorts of processors.

I've checked the thermal paste its great, redone it a few times assuming it must have been poorly spread, used the same cooler on a 4770 which I upgraded from, no issues at all with temps. CM evo212.

0 Kudos
BKirk
Beginner
16,170 Views

Just adding my name to the list of people with this issue. My cooler however is a Swiftech H240-X AIO water-cooler, which to be honest shouldn't have any issue cooling this processor. Without load, the processor stays at ambient as you'd expect. The In-Place Large FFTs and Blend in Prime95 run about 40 degrees above ambient (70 degrees). Using the Small FFTs results in instant >100 temperatures and the computer will eventually restart. I had read recently that someone removed the heatspreader from their 4790K and found cooling afterwards substantially better, which supports the idea that there's terrible heat conductivity between the top of the die and heat-spreader. If it matters, my batch is L424B559, but from the looks of it, a lot of people are reporting thermal issues with this processor.

Edit: Re-seating my cooler helped a little bit (my bad on that one, but I had changed from stock for the same reason anyway), can run the Blend after the initial extra heating, but after a couple of minutes where it packs on more heat my computer will restart.

Using XTU my CPU temperature stays around 58-60 degrees which of course is fine. I assume this would be simulating normal 100% cpu conditions?

Edit 2: Weird, my CPU isn't thermal throttling, even with it specifically enabled in my BIOS (Gigabyte Z97X-UD5H Version: 1.0/F8), so instead it's just restarting.

Edit 3: Reading more, it sounds like it's Prime95 putting unrealistic load on the processor (AVX in the FPU??).

0 Kudos
idata
Employee
16,170 Views

AVX + FPU + PRIME95 is propagande from Intel side

It's very realistic load. Just imagine you are compiling few androids a day (it tooks 1 hour, and uses all 8 threads), or have DVB streaming server, or other high-duty server, ....

You can test your cpu with this software:

http://www.cyclone-networks.eu/IntelThermalTest.rar http://www.cyclone-networks.eu/IntelThermalTest.rar

This software is written inDelphi 2007, and I'm assure you compiler that times didn't knew what AVX it, so it can't compile floating point to AVX code.

AVX is just comfortable propagande answer by Intel to avoid mass replacment procedure.

Let me know about tet results, but probably, your CPU, as thousands or millions of others should be replaced because of manufacturing fault.

0 Kudos
BSant8
Beginner
16,170 Views

Hey Karwos and welcome to the community!

You raise an interesting point about this being a marketing propaganda on intel's side. But we can't just take your word for it. Could you please back up your affirmations with some sources (links / articles / etc.)? Otherwise it's just some words from an angry customer that people might take as a fact and it might not be the case.

0 Kudos
idata
Employee
16,170 Views

I already written my affirmations.

I've written testing utility and i'm assure you - it doesn't use AVX extensions (because Delphi2007 compiler simply don't support it). It could be verified - SW is packed with upx only and you can disassemble and check it. Same problem.

Secondly, IPDT utility also not using AVX extensions. Same problem.

In case of Intel, there is more interesting things: IPDT software is written that way, it will never fail temperature tests, and i've posted prooves on diffrent thread:

/message/258423?tstart=0# 258423 https://communities.intel.com/message/258423?tstart=0# 258423

I guess you should know now why AVX is just comfortable point of view by Intel.

0 Kudos
BSant8
Beginner
16,170 Views

Hey Karwos!

I have run your test and here is the result:

As you can see, my CPU passes this test but when i run Prime95 (ver. 28.5) the temps go from 35idle to 85-90.

What should i understand from this? That my CPU is fine and there is a problem with the way Prime95 stresses the CPU?

0 Kudos
BKirk
Beginner
11,922 Views

Seems to pass fine. Lower temperature than even the XTU test. According to HWMonitor, the max package power is around 60W for your ITT, while for the XTU it's around 80-90W. When you run the small FFTs in the latest version of Prime95 it's around 150W. The TDP is only 88W, suggesting Intel never considering running 4/8 threads of this instruction at 100% a typical use case. I had a short search around, and couldn't find a maximum die to spreader (or something similar) rating. Is there actually a maximum thermal dissipation (vs the average TDP based on an Intel specified workload) documented somewhere for the package?

0 Kudos
idata
Employee
11,922 Views

To all who use IntelThermalTest and their tests being passed and under Prime95 it's failed

Can you please check, and re-run IntelThermalTest, and check in task manager (Ctrl+Shift+Esc) what CPU Load generates "IntelThermalTest.exe"?

It may be possible it didn't utilize full CPU time because i forgot to set each thread affinity mask to being executed by diffrent thread/core, so MAYBE Kernel OS for some reason executing some of threads sequentially on same core. Also i didn't increased default thread priority, this also may be cause.

So please check and let me know. It should oscilllate around 99-100. If it's 50-60, that's reason of PASSING the thermal test. I will relase 0.2 then.

Best Regards

0 Kudos
Reply