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.

[idata]

I didn't got any respond from them yet.

However, I've already posted here possible problem with other 4th gen cpus: 4770, 4770k, 4771, .... not sure about i3, i5, xeons ...

If problem is broad (and it may be VERY BROAD) that means really big problems for Intel:

1. Replacement costs - let say JUST 2 Millions CPUs x 350 USD it gives 0.7 billions USD - it's big problem not only for Intel, but also for Intel investors. Just get noticed how much already pumped Intel stock value are (20 usd last year vs 33 usd now). It's not like I want spread any doubts, but it may be highly possible that problem is known by Intel from long time, but releasing any official press infomation may cause dropping stock value by 30-50% very fast, and everyone who ever invested at stock markets should know what it means.

2. Costs of eventual court cases filed by class action (in Europe, America, Asia) because of selling device with incorrect technical specification (not helding frequencies) - it's also possible

3. Costs of penalties from various governments for hiding overheating problem (including possible specially faked IPDT software which ignoring problem and saying, hell yes, everything is ok) - for examle European Union, they aren't giving fuk for such actions

4. Costs of penealties from various goverments for shorting lifecycle of product by letting those products cross TCase_max and work at shorter-lifespan temperature

5. Costs of trustability loss in eyes of endusers, distributors, resellers

6. AMD may gain a lot on this action.. and AMD gain is Intel loss

Anyway I think Intel is too big and smart company to do such child plays. But honestly - if you look at all reports from various users - it may be highly likely that prolem is REALLY broad, and 4 months is enough time to release any official announcment.

That's why as I stated already, i will make app for Haswell based cpu's which will test all avaiable on market 4th Gen Intel processors, regarding to speed and temperature. I hope it will spread fast, and we will get some results fast aswell, so Intel will stop ignoring us. Threating so many cases of overheating as simple RMA is simply inslut, but looking from side of investors charts and reports... it looks much better than mass replacement

This might help getting real vision of scale of this problem. Maybe it was just small batch, maybe like 10k units, but it may be really big problem of million of devices.

PS. I have been reading lot of datasheets for 4th gen Desktop CPUs yesterday, and I was analysing various MSR Registers under load. I saw also some very interesting entry

Under fan failure or other anomalous thermal excursions, processor temperature

(either TCASE or DTS) may exceed the thermal profile for a duration totaling less than

360 hours per year without affecting long term reliability (life) of the processor. For

more typical thermal excursions, Thermal Monitor is expected to control the processor

power level as long as conditions do not allow the processor to exceed the

temperature at which Thermal Control Circuit (TCC) activation initially occurred.

Now, imagine your CPU is heavy load (for example 2-3 android builds every day) every day.... and it's crossing 74.04c barrier....

[idata]

Seems there is very serious manufacturing process issue.

I have 2 units, with same specs:

- CPU - i7 4790K

- Motherboard - Supermicro C7Z97-OCE

- RAM Crucial Ballistix Sport DDR3 8GB 1600MHz CL9

- Case - Chieftec UNC-210HS-B

- OS - Debian 7.7 with kernel 3.16.3 and intel_pstate enabled

Both units are rack mounted in air conditioned room (ambient temperature 19-20c). Cooling is stock (those attached to box units. so stock radiator, stock cooler, heatsink, etc). Chieftec cases have total 5x fans. Air circulation is not disturbed any way (rack is open for tests).

Idle temperature are 25-26C. After putting stress test (stress linux utility, params: stress --cpu 8) temperature goes up to 100C after 30-40 seconds.

After that, BIOS thermal alarm sound is played, until I stop stress test. Of course CPU is underclocked (3.3GHz) to avoid burning.

In same rack, i have also mounted server (in same case) with 3770K CPU (rated 3.9GHz - turbo). After putting stress test, temperature don't goes above 70-72c.

Im clearly doubt that this is related to mounting process, because same problem is replicated with two units, and as I can see here lot of other users have exactly same problem.

I used chat to contact Warranty/RMA guy, however they are employing lazy peoples (after 15 mins of waiting got support article about checking heatsing and fans vents ). So just loosing time.

Because CPUs not working as expected (need underclock them), and are not meeting their technical specifications, i'm asking everyone from Europe to also submit this problem here. Together, we can submit legal class action to European Union court, AND MAYBE this will start Intel thinking about problem and taking care of their customers.

Best Regards

[idata]

Ok, so I did further research regarding this topic...

Small foreword: all this turboboost is a little cheat from Intel side.

Regarding to ARK resources, they not mention that turbo speed regards only to 2 cores under high load.

when 3 cores are under load, total speed for each core drops by 100MHz, when 4 cores are under load total speed for each core drops by 200MHz

Of course all this is to meet thermal design. Surely i know this should be called not cheating, but marketing

for example, factory turbo multipliers for 3770k are: 39/39/38/37 (1 core 100% load - 3.9ghz, 2 cores 3.9ghz, 3 cores 3.8ghz, 4 cores 3.7ghz)

Factory turbo multipliers for 4790k are 44/44/43/42

Nevermind...

As I stated before, right now my rack is employed with total 3 i7 based PCs:

1) 3770k

2) 4790k

3) 4790k

each server have stock radiator, cooling, heatsing. Each PC have same casing, so as well same ambient temperature, etc. Nothing touched.

I did tests for 3770k (i7z + stress tool)

Some numbers regarding 3770K from ARK: Max TDP 77W, Thermal Profile 2011D, that means (for stock cooling):

(source: http://www.intel.com/content/dam/www/public/us/en/documents/design-guides/3rd-gen-core-lga1155-socket-guide.pdf http://www.intel.com/content/dam/www/public/us/en/documents/design-guides/3rd-gen-core-lga1155-socket-guide.pdf , page 43)

Power usage 0Watts = TCase_Max 45.1c

Power usage 77Watts = TCase_Max 67.4C - nominal rating for this cpu

Power usage 95Watts = TCase_Max 72.6C - worst case scenario

In any case, because of thermal profile requirements, stock cooling efficiency SHOULD NEVER exceed given thermal values for rated CPU Watt usage.

Let's put stress --cpu 8 (8 threads; 4 cores on full load) and wait exactly 5 minutes...

Cpu speed from cpuinfo 3510.00Mhz

cpuinfo might be wrong if cpufreq is enabled. To guess correctly try estimating via tsc

Linux's inbuilt cpu_khz code emulated now

True Frequency (without accounting Turbo) 3510 MHz

CPU Multiplier 35x || Bus clock frequency (BCLK) 100.29 MHz

Socket [0] - [physical cores=4, logical cores=8, max online cores ever=4]

TURBO ENABLED on 4 Cores, Hyper Threading ON

Max Frequency without considering Turbo 3610.29 MHz (100.29 x [36])

Max TURBO Multiplier (if Enabled) with 1/2/3/4 Cores is 39x/39x/38x/37x

Real Current Frequency 3710.57 MHz [100.29 x 37.00] (Max of below)

Core [core-id] :Actual Freq (Mult.) C0% Halt(C1)% C3 % C6 % C7 % Temp VCore

Core 1 [0]: 3710.57 (37.00x) 100 0 0 0 0 71 1.1158

Core 2 [1]: 3710.57 (37.00x) 100 0 0 0 0 70 1.1158

Core 3 [2]: 3710.57 (37.00x) 100 0 0 0 0 71 1.1158

Core 4 [3]: 3710.57 (37.00x) 100 0 0 0 0 71 1.1158

Conclusion: during test, temperature never rised above 71-72C. It's very stable. Because of very efficient and thermal-compatible cooling, speeds are also very stable and working as expected.

Now let's move to 4790k. The results are exactly same for both PCs i have (both are exactly same builds; cpu; motherboard; ram; etc) so I will post only one.

Some figures from ARK: TDP 88W, PCG 2013D thermal profile, that means:

(source: http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/4th-gen-core-family-desktop-vol-1-datasheet.pdf http://www.intel.com/content/dam/www/public/us/en/documents/datasheets/4th-gen-core-family-desktop-vol-1-datasheet.pdf , page 68)

Power usage 0Watts = 45C

Power usage 84Watts = 72.72c

Because specification lacks values for 88watts (nominal tdp), and let say worst case tdp (95w), let's calculate it from specified calc method , this is : y=0.33 * W + 45, so:

Power usage 88Watts = 74.04c - nominal rating for this cpu

Power usage 95Watts = 76.35c - worst case scenario

In any case, because of thermal profile requirements, stock cooling efficiency SHOULD NEVER exceed given thermal values for rated CPU Watt usage.

So let's confirm that...

Cpu speed from cpuinfo 3990.00Mhz

cpuinfo might be wrong if cpufreq is enabled. To guess correctly try estimating via tsc

Linux's inbuilt cpu_khz code emulated now

True Frequency (without accounting Turbo) 3991 MHz

CPU Multiplier 40x || Bus clock frequency (BCLK) 99.78 MHz

Socket [0] - [physical cores=4, logical cores=8, max online cores ever=4]

TURBO ENABLED on 4 Cores, Hyper Threading ON

Max Frequency without considering Turbo 4090.78 MHz (99.78 x [41])

Max TURBO Multiplier (if Enabled) with 1/2/3/4 Cores is 44x/44x/43x/42x

Real Current Frequency 3579.25 MHz [99.78 x 35.87] (Max of below)

Core [core-id] :Actual Freq (Mult.) C0% Halt(C1)% C3 % C6 % Temp

Core 1 [0]: 3579.24 (35.87x) 99.9 10.4 0 0 100

Core 2 [1]: 3...

[idata]

I done comparision table of current desktop i7 haswell based...

CPU Name Rated TDP Thermal profile TCase_Max @ Rated TDP

======================================================================

i7 4765T 35W PCG2013A 66.35'c

i7 4770 84W PCG2013D 72.72'c

i7 4770K 84W PCG2013D 72.72'c

i7 4770S 65W PCG2013C 71.35'c

i7 4770T 45W PCG2013B 71.45'c

i7 4770R 65W ???????? ???????

i7 4771 84W PCG2013D 72.72'c

i7 4790 84W PCG2013D 72.72'c

i7 4790K 88W PCG2013D 74.04'c

i7 4790S 65W PCG2013C 71.35'c

i7 4790T 45W PCG2013B 71.45'c

i7 4785T 35W PCG2013A 66.35'c

Seems this problem is more broad than we think. All at least 84W TDP rated CPUs have problem with reaching 100C (and in result: shorter life of CPU AND slower performance because of underclocking).

We can easily find lot of reports of overheating... for example:

4770k - http://www.youtube.com/watch?v=iz8Yi5umuRk http://www.youtube.com/watch?v=iz8Yi5umuRk

http://www.tomshardware.co.uk/answers/id-2006364/4770k-temps-reaching-100-degrees-prime95.html http://www.tomshardware.co.uk/answers/id-2006364/4770k-temps-reaching-100-degrees-prime95.html

http://www.tomshardware.co.uk/answers/id-1943697/100-degrees-celsius-load-4770k-normal.html http://www.tomshardware.co.uk/answers/id-1943697/100-degrees-celsius-load-4770k-normal.html

http://www.tomshardware.co.uk/forum/id-1827200/4770k-running-100c-browsing.html http://www.tomshardware.co.uk/forum/id-1827200/4770k-running-100c-browsing.html

http://www.overclock.net/t/1429257/solved-i7-4770k-not-ocd-at-100-degrees-celsius/10 http://www.overclock.net/t/1429257/solved-i7-4770k-not-ocd-at-100-degrees-celsius/10

4770

http://linustechtips.com/main/topic/95085-my-intel-core-i7-4770-gets-up-to-21200%C2%BAf/ http://linustechtips.com/main/topic/95085-my-intel-core-i7-4770-gets-up-to-21200%C2%BAf/

http://www.mmo-champion.com/threads/1610067-These-CPU-temps-okay-Haswell-I7-4770?p=30053382&viewfull=1 http://www.mmo-champion.com/threads/1610067-These-CPU-temps-okay-Haswell-I7-4770?p=30053382&viewfull=1

http://forums.anandtech.com/showthread.php?t=2328917 Core i7 4770 w/ stock cooler - 100 C during Intel CPU stress test - AnandTech Forums

http://www.tomshardware.co.uk/answers/id-1762407/4770-reaching-100c.html http://www.tomshardware.co.uk/answers/id-1762407/4770-reaching-100c.html

http://www.tomshardware.co.uk/answers/id-2176540/4770-reaching-83c-max.html http://www.tomshardware.co.uk/answers/id-2176540/4770-reaching-83c-max.html

http://www.overclock.net/t/1429257/solved-i7-4770k-not-ocd-at-100-degrees-celsius/10 [SOLVED] i7 4770k not OCd at 100 degrees celsius?! - Page 2

4771

http://forums.pcper.com/showthread.php?481606-i7-4771-incorrect-temperature-reporting i7-4771 incorrect temperature reporting?

http://forums.whirlpool.net.au/archive/2234690 Is 55c too hot for i7-4771 at idle? - Core system - Desktops

http://setiathome.berkeley.edu/forum_thread.php?id=73268 Project down for maintenance

http://www.tomshardware.co.uk/answers/id-2127770/4771-underclocks-load.html http://www.tomshardware.co.uk/answers/id-2127770/4771-underclocks-load.html

The problem may be also underreported (not all users measure their cores performance and temperature). So for lot of users there will be no diffrence expect slower PC and shorter life of CPU....

Looks like problem may be a LOT broaden than we think, but for some reasons Intel keep it very quiet.

What reason is it? Are Intel afraid of dropping their stock price (which is already pumped up, nearly 10usd+ since last year)?

Probably problem is bad heatsink between core and IHS, what can be confirmed here:

You can easily see (same problem here) - there was claims that between die and ihs there is soldering.. what is not true for haswell..

http://www.pugetsystems.com/labs/articles/Delidding-the-4790K-for-a-quick-look-at-the-TIM-573/ http://www.pugetsystems.com/labs/articles/Delidding-the-4790K-for-a-quick-look-at-the-TIM-573/

So what will Intel do now?

What about i3, i5, and mobile CPUs?

Are they affected with bad heatsink aswell?

Topic is very interesting for everyone: Endusers, Intel, Intel investors, .... 

[KenF_Intel]

Hi all,

Intel® Core™ desktop processors support up to the 

published Intel® Turbo Boost Technology frequencies with defaultsettings/cooling at nominal room temperature (72°F/22°C) using non-AVXworkloads. If the default settings are modified (e.g. multipliers,voltages, power/current limits, etc.), the Intel Reference Heat sink receivedwith the processor may not handle the extra heat and cause the Thermal ControlCircuit (TCC) to become active – it will attempt to limit the processortemperature by lowering voltage and frequency until it drops below maximum safetemperature; furthermore, AVX instructions (such as those used by Prime95version 26.6 or later) require higher processor voltage and will generate moreheat/consume more power than the default limits will allow.

If frequency limitation occurs, please use the Intel® Extreme Tuning 

Utility (XTU) to monitor the temperatures being generated and/or the powerbeing consumed by the processor (click the little wrench icon near the lower-rightcorner of the window to enable all monitors). If the frequency isn'treaching its specified values due to power and/or current limitations, tryadjusting "Turbo Boost (Short) Power Max" and/or "Processor Current Limit"higher. If the cooling solution isn't capable of dissipating the applicableheat, ensure all "Voltage" controls are set to "Default," and "Voltage Offset"controls are set to 0.

[SBuck1]

Really? It's been like 2 and a half months of many users reporting in great detail and specificity overheating issues. None of these are addressed by your post. Numerous users have reported air gaps and excessive adhesive between the chip and the lid. Numerous users have solved their heating issues by de-lidding and fixing the manufacturing defects contained therein. Your response is frankly insulting. Intel has had a manufacturing issue, and it's time to acknowledge it and replace the affected chips.

[idata]

I think almost everyone can read datasheets, and everyone knows that things can vary if...

1) Your radiator is covered with dust

2) Ambient temperature is too high

3) you are playing with multipliers, VCore,

4) etc.. etc.. etc

Here, such things don't exists.

Secondly, as you can see in movie posted by me, your tool (not prime95) also failing the tests.

From what I can see at IPDT - CPULoad source code, it's starting various threads for making multiplication to occupy processor:

void FloatMultiply(int iTestTime, int iStartCount, bool bRunTemp, unsigned long lTempTol)

{

tick_count startTime = tick_count::now();

tick_count startCurTime = tick_count::now();

tick_count nowtime = tick_count::now();

tick_count curtime = tick_count::now();

// Call Temperature

TempManager tMan;

tMan.bEraseLastLine = true;

tMan.lTempTol = lTempTol;

while((nowtime-startTime).seconds() < iTestTime)

{

nowtime = tick_count::now();

curtime = tick_count::now();

double z=1;

z *= 0.02;

if(bRunTemp == true && (curtime-startCurTime).seconds() > 5)

{

startCurTime = tick_count::now();

if(iStartCount == 0)

{

tMan.TempMon();

}

else

{

iStartCount++;

}

}

}

}

I don't have time to disassembly and check it now, but i don't think so compiled code would use AVX extensions.

Anyway, so, can you explain me why i7 3770K working perfectly as expected:

putting maximum stress tests, CPU works at 3.7GHz (rate for 4 cores occupied), after really long time, temperature readen from MSR ia32_package_therm_status (0x1b1) never exceeding tcase_max from thermal profile. All this on stock cooling. Magic? No, just everything working as expected, and meeting thermal profile designed for this CPUs.... Here, sorry but thing's simply don't work.

[idata]

Btw,

quickly analyze IPDT source code led me to really very interesting conclusions:

1/ CPU temperature is gathered by ring0 wrapper using MSR registers

2/ WarningPoint (lTempTol) is being set at IPDT startup. It's being readen from XML config file from property "TemperatureWarningTolerance". This value is by default 0. IT SHOULD BE SET AS LET SAY T_CASEMAX + some percent of tolerance + conduction diffrence.

3/ During CPULoad test, temperature is frequently monitored and tested by temperature monitor, whether it doesn't exceed previously set warningpoint.

4/ More interesting is what is happening at TemperatureManager:

if(LSIZE[0] > lTempTol)

{

ostrLastLine << LSIZE[0] << " degrees C below maximum. ";</em>

}

LSize[0] in this context is temperature with distance to TjMax (as readen from MSR), so in case of 92degrees C, it's simply "8".

lTempTol is previously set TemperatureWarningTolerance. Correct value should be for 4790k 26. (100-74).

As long as lTempTol is constantly set to 0, this equation will always be positive, unless it could read signed integer from MSR Register. And it couldn't, as long as readen value is unsigned 6bit adc result.

so 8 degrees, 1 degrees (to 100c) will be always higher than 0 .... so below maximum.

In this case, if CPU would have 60 degrees, value from MSR woudl have 40 (100-60=40), and equation would work well.... 40 > 24 .. even here lsize[0] should be substracted from lTempTol (to check distance from warningpoint and current temperature) and only substraction (40-24) would led to real "16 degrees to C below maximum"...

There is more interesting things in sources, however i have no time now to analyze all this.

One is sure: regardless of thermal results, IPDT is written this way it will never fail temperature test.

Surprise?

Hi-tech engineers? Quality assurance teams? Debugging teams? Testers?

It's a joke ?

[idata]

As I've promised, here is tool which will test your processor according to TCase_Max.

This is just early preview version, however everything shoud works okay.

It supports all currently avaiable 4th gen Haswell Desktop i7 (rest will be added after some positive feedback).

This software will calculate TCase_MAX according your CPU TDP and Thermal profile.

During stress test, it will check, if CPU _package_ (not each core) temperature will not cross TCase_Max+10% (this will need some tuning later, I think, but for begining let say 10% tolerance).

it will also display cores requested speed, and current speed.

when cpu is melting, or there is not enough ampers (i7 have no power limiter, but current limiter. so if your vcore is tuned down to let say 1.09, you will get 200-300mhz less, because 1.09*95amps, and 1.16v*95amps gives you diffrent power values.. more power=more speed!)

it will read MSR register and show why power is throttled back (thermal problem, no enough current, power plane limiter, power package limiter, core limiter, etc).

So if anybody can test this SW and let me know about results, would be cool.

Get it here: http://www.cyclone-networks.eu/IntelThermalTest.rar http://www.cyclone-networks.eu/IntelThermalTest.rar

[lpart]

hello

Unfortunately I could not test your software (the pc that I have had not up to me and I've sent earlier this week).

Anyway thank you for all these explanations and for the time you have to spend to look for or had the worries.

Many users (like me) see that there is a problem but can not push it as far as you research did.

I will continue to follow this thread to see the progress.

[idata]

Thanks you for that.

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

Here is example output from software (run at 4790K):

IntelThermalTest v0.1 by karwos - startup

Calling opcode 0xa20f, eax=0...

This is GenuineIntel, good...

Calling opcode 0xa20f, eax=1...

Brand Name: "Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz"

Family: 6, Model: 60, Steppping: 3

CPU TCase_MAX: 74.04 C

CPU TCase_MAX + Tolerance: 81.444 C

CPU TjMAX 100 C

Performance Bias: BALANCED

Non-Turbo Frequency 4000 MHz

[Active 1 Core] -> Max Turbo Frequency: 4400MHz

[Active 2 Cores] -> Max Turbo Frequency: 4400MHz

[Active 3 Cores] -> Max Turbo Frequency: 4300MHz

[Active 4 Cores] -> Max Turbo Frequency: 4200MHz

Test in progress...(SPEED_PENALTY+MAX_TURBO_LIMIT)++

60++++

63++++

65++++.++++

67++++

68++++.++++.++++

69++++

70++++.++++.++++

71++++.++++

72++++.++++

74++++.++++.++++.++++

75++++.++++.++++

76++++

77++++.++++.++++

78++++.++++

79++++

80++++

81++++.++++

82++++.++++

83++++.++++.++++.++++.++++

84++++.++++.++++.++++

85++++.++++.++++.++++.++++

86++++.++++.++++

87++++.++++.++++

88++++.++++.++++.++++.++++.++++.++++

89++++.++++.++++.++++

90++++.++++.++++.++++.++++.++++.++++.++++.++++

91++++.++++.++++.++++

92++++.++++.++++.++++.++++.++++.++++.++++.++++

93++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++

94++++.++++.++++.++++.++++.++++.++++.++++.++++.++++

95++++.++++.++++.++++.++++.++++.++++.++++.++++.++++

96++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++

97++++.++++.++++.++++.++++.++++.++++.++++

98++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++

99++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.(SPEED_PENALTY@THERMAL_STATUS)+++.++++.++++.++++.++++.++++.++++.++++.++++.++++.++++.+++@.++++.++++.++++.++++.@+@+.++++.++++

100+++@.@@@+.++++.++++.@+++.++++.++@+.+@++.@@@+.@+@@.@@+@.@@@+.@@+@.@+@@.@@+@.@@@@.@@@@.@@+@.@@@@.@@@@.@@@@.@@@@.@@@@.@+@+.@@@@.@@@@.+@+@.@@@@.@@+@.@@@@.+@@+.@@@@.@+@@.@@@@.@@@@.@@@@.@@@@.@@@@.@@@@.@+@@.@+@@.@@@@.@@@@.@@@@.@@@@.@@@@.@@@@.@@+@.@@@@.@@@@.@@@+.@@@@.@@@@.@@@@.@@@@.@@+@.@@@@.@@@@.@@@+.@@@@.@@@@.@@@@.@@@@.@@@@.+@@@.@+@@.@+++.@@@@.@@@@.@@@@.@@@@.@@@+.@@@@.@@@@.@@@@.@@@@.@@@@.@@@+.@@@+.@@@@.@@@@.+@@@.@@@@.@@@@.@@@@.@@@@.@@@@.@@@@.@@@@.@@@@.@@@@.@@@@.@@@@.@@@@.@@@+.@@@@.@@@@.@@@@.@

TEST FAILED

!!!! YOUR CPU EXCEEDING DESIGNED THERMAL PROFILE BY INTEL !!!!!

---- THERMAL TEST FINISHED ----

Small abbreviations of this magic @@, ++, numbers and ...

1. Number in each row is new highest noted package (not core) temperature

2. Dot (" . ") in each row, means this temperature lasting for unit of ONE SECOND (5 dots, 5 seconds, 10 dots 10 seconds, etc)

3. Special symbols (!@# $%^&*|) means current PERF_STATUS register (only on haswell). If your turbo speed will be clamped, this will help you see what is currently clamping reason. Each symbol means 0.25s time. First occurence of PERF_STATUS clamping is also noted as (SPEED_PENALTYMEANING), in this case @ means clamping because of too high temperature... and each @@@@ symbol means clamping lasts for 1second.... At last minutes of tests, speed is constantly clamped because of internal thermal protection mechanisms.

Meaning of all symbols:

- new package temperature point

"." - unit of 1 second

"!" - PROCHOT is asserted (processor hot)

"@" - THERMAL STATUS is asserted (internal mechanism for avoiding crossing TjMax)

"# " - GRAPH_DRV

"$" - AUTONOMOUS_UTILIZATION

"%" - VR_THERM_ALERT

"^" - ELECTRIC_DESIGN_POINT.. that means CPU can't handle more than 95Amps (in case of 4790k).. that means VCore is lower than it should be (for example 1.09v instead of 1.16v - check bios and voltage source)

"&" - CORE_POWER_LIMIT - exceeded core power limit. usually set by kernel os (for example during powersaving plans activation)

"*" - POWER_PLANE_1_LIMIT - same as above, but for plane1

"|" - POWER_PLANE_2_LIMIT - look up

"+" - MAX_TURBO_LIMIT - related to 0x1ad msr register limits. usually normal situation, especially on K cpus (can rewrite 0x1ad register to higher values (overclock), but is actually clamped at factory 4.2GHz value). Maybe not ok when OS kernel will limit that for some really low values. Readout of 0x1ad register is during test startup at "[Active 4 Cores] -> Max Turbo Frequency: 4200MHz", and this is actual clamp value.

"-" - TURBO_ATTENTUATION_STATUS

[LPala6]

IntelThermalTest v0.1 by karwos - startup

Calling opcode 0xa20f, eax=0...

This is GenuineIntel, good...

Calling opcode 0xa20f, eax=1...

Brand Name: "Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz"

Family: 6, Model: 60, Steppping: 3

CPU TCase_MAX: 74,04 C

CPU TCase_MAX + Tolerance: 81,444 C

CPU TjMAX 100 C

Performance Bias: HIGH PERFORMANCE

Non-Turbo Frequency 800 MHz

[Active 1 Core] -> Max Turbo Frequency: 4400MHz

[Active 2 Cores] -> Max Turbo Frequency: 4400MHz

[Active 3 Cores] -> Max Turbo Frequency: 4400MHz

[Active 4 Cores] -> Max Turbo Frequency: 4400MHz

Test in progress...

59

61

63.

64

65.

66.

67.

68.

69..

70..

71..

72......

73..

74........

75..

76.......

77.....

78....

79.............

80...................

81....

82..................................................................

83.............................................................................................................................

TEST FAILED

!!!! YOUR CPU EXCEEDING DESIGNED THERMAL PROFILE BY INTEL !!!!!

---- THERMAL TEST FINISHED ----

Batch# L421B969

I have already tried to open up the case, and add an external table fan to enhance airflow. This helped with 10°C degrees. At the moment I am thinking about buying a water cooler or a heatpipe solution, but I am not sure if it will solve the problem. The system hangs up randomly with several settings. I have a Z97-A Asus motherboard with Kingston 2400 RAM, but the system is useless. The seller offered replacement, but I am afraid of getting a similar one. Maybe some month later Intel will come out with a new series?

[SPark48]

karwos,

When running your program, I'm not clear about the results. What does it mean when I'm not getting any special characters and only "."? Is this a good thing or a bad thing? With the previous processor (before RMA), I was getting a LOT of "+" with a few "." in between, but now with new processor, I'm not getting any "+" and ONLY "." Can you help me understand what this means? Thank you for your time.

[BSant8]

Hey Sampark!

i'm glad to hear your new CPU runs better than the last one. I can see you even managed to squeze a little more from it and still keep it in normal temp ranges.

here is what the simbols mean:

1. Number in each row is new highest noted package (not core) temperature

2. Dot (" . ") in each row, means this temperature lasting for unit of ONE SECOND (5 dots, 5 seconds, 10 dots 10 seconds, etc)

3. Special symbols (!@# $%^&*|) means current PERF_STATUS register (only on haswell). If your turbo speed will be clamped, this will help you see what is currently clamping reason. Each symbol means 0.25s time. First occurence of PERF_STATUS clamping is also noted as (SPEED_PENALTYMEANING), in this case @ means clamping because of too high temperature... and each @@@@ symbol means clamping lasts for 1second.... At last minutes of tests, speed is constantly clamped because of internal thermal protection mechanisms.

Meaning of all symbols:

- new package temperature point

"." - unit of 1 second

"!" - PROCHOT is asserted (processor hot)

"@" - THERMAL STATUS is asserted (internal mechanism for avoiding crossing TjMax)

"# " - GRAPH_DRV

"$" - AUTONOMOUS_UTILIZATION

"%" - VR_THERM_ALERT

"^" - ELECTRIC_DESIGN_POINT.. that means CPU can't handle more than 95Amps (in case of 4790k).. that means VCore is lower than it should be (for example 1.09v instead of 1.16v - check bios and voltage source)

"&" - CORE_POWER_LIMIT - exceeded core power limit. usually set by kernel os (for example during powersaving plans activation)

"*" - POWER_PLANE_1_LIMIT - same as above, but for plane1

"|" - POWER_PLANE_2_LIMIT - look up

"+" - MAX_TURBO_LIMIT - related to 0x1ad msr register limits. usually normal situation, especially on K cpus (can rewrite 0x1ad register to higher values (overclock), but is actually clamped at factory 4.2GHz value). Maybe not ok when OS kernel will limit that for some really low values. Readout of 0x1ad register is during test startup at "[Active 4 Cores] -> Max Turbo Frequency: 4200MHz", and this is actual clamp value.

"-" - TURBO_ATTENTUATION_STATUS

SO, basicly, when you only get a "." you are ok. The CPU is not throttleing. It just recorded that temp for one second. Another second = another dot. So you are fine.

COuld you please run another test for me? I would be interested to know your results when you run the Small FFT test in Prime95 (ver 28.5) for 15 min. What i'm most interested in are the temps and the Vcore Voltage. You could monitor with Intel XTU while Prime95 is running in the background. Could you please post the results? I'm asking because i also get the same temps while using any other stress test software, but when i run Small FFT test in Prime95 my CPU starts throttleing due to current. Thanks!

[SPark48]

Woolphy,

Sorry this reply took so long to get back to you, but I have been out of town and have not been able to test the machine. First of all, thanks for the explanation on the Thermal Test. That's a relief that the new processor is not throttling. I guess that meant that the previous processor (which had mostly +'s with a few periods in between) was throttling very frequently.

As far as the Prime95 Small FFT test.... all I can say is OUCH! 99C within the first 1-2 seconds of turning it on, after which I turned off the test immediately. It's very strange, though, because my previous processor (pre-RMA) was giving these temps in OCCT as well as Prime95. The new one, however, runs perfectly fine on OCCT, yet Prime95 is again giving 99C. I'm curious what specific instructions are run by the Prime95 Small FFT tests and if there are any applications that use the same instructions that will have a chance at overheating the processor... You can be sure that Intel has called GIMPS to let them know how unhappy they are with the Prime95 testing. Regardless of these results, I firmly believe there is something inherently wrong with some of the 4790k processors that Intel produced. Luckily the second one I got runs a lot better, even if it does "fail" the Small FFT. Considering that I'm passing the other stress tests (OCCT, EVGA OC Scanner, and Intel XTU are the ones I have tried so far), I'm pretty satisfied with where I'm at, at this point.

[RKimb]

Forgive me for being cautious, but is the program likely to cause any damage? I wouldn't want to void the warranty at this stage.

[BCicc]

Hi,

tried to download karwos' utility but the link is now broken. Is it possible to obtain a working link from someone here?

I am dealing with overheating too, I even opened a separate thread on this community.

Thanks in advance!

[aanwa1]

but when running AIDA64 it reaches 85-90 and when running prime it goes slowly till 93 degrees.

[jband2]

I also have the same problem with overheating i7-4790K.

When idle the temps are between 25-30 with cooler stock 1150 rpm +/-

I looked at the batch and it is indeed L4.(L425B824)

I use ASUS Z97-A motherboard and recently updated to the latest bios (2012), without improvement.

[idata]

Is anybody recording the bad batches? Mine is L419B634

There may be a pattern.

[SRene]

I'm facing the same problem. My batch is L420B887.

To run the tests I had to reduce vcore to 1.125 and deactivate the Turbo Boost (with stock cooler).

For those who are getting 100C temps I suggest you try to reduce the vcore a bit and disable the turbo boost.

I'm searching another cooler to buy to check if I can reactivate the turbo boost.

ken-Intel you should be embarassed for that post. This discussion started 2 months ago, we have a lot of posts detailing the problems, and you say what everyone already knows?!

We want solutions! It's a K processor! We want to reach at least the turbo speed with stock conditions and without being throttled! That is what we bought!