Help me i7 4790k very hot !!!

[MrFuny]

Because the Intel Burn Test is excessively harsh, unnecessary and will reach 10c higher temps than any other stress test. if you have bad surface contact with the cooler you could very quickly put the chip in danger. There are many other stress test that are more appropriate for testing your system.

OCCT, AIDA64 are some of my favorite

I like stress with IBT, AIDA64 not good for ocer - AIDA64 pass, render miss bug.

 

[Toothless]

@MrFuny It all is within safe limits. If you can drop the vcore to 1.25v and you'll be fine. If you really don't want to mess with it then leave it as is.

 

[MrFuny]

@MrFuny It all is within safe limits. If you can drop the vcore to 1.25v and you'll be fine. If you really don't want to mess with it then leave it as is.

-I will lower volt, step by step 1.25,1.2,...
-with my setting( all default, only set turbo for all cores up to 4.4G - default 4.2g), power consumption show up 184W normal or problem?

 

[Toothless]

-I will lower volt, step by step 1.25,1.2,...
-with my setting( all default, only set turbo for all cores up to 4.4G - default 4.2g), power consumption show up 184W normal or problem?

That's fine. Honestly with a CPU running with turbo will draw power as voltage goes up. Remember it's an 8-threaded chip that hauls ass.

My 4790k can sit on 1.75v just fine but I'd rather run at 1.2v to keep it more stable.

I know I'm running standard but the program keeps crashing on maximum. Very high gives same temps.

Specs in the little tab under my name is up to date so you can see what cooler I'm using.

 

[EarthDog]

@trog - you may be surprised at how little temperatures increase with just clockspeed increases and no voltage bumps. It is mainly voltage that increases temperature.

Also consider that typically you can't raise the clockspeed much without voltage control..what couple hundred mhz?

You should test it and see. I did years ago when I thought the same thing (was told to test it just like you in fact!). While temps may go up a degree or so with a couple hundred mhz bump, it PALES in comparison to what voltage does. Voltage alone raises temps a tonew without a clockspeed change.

 

[trog100]

@trog - you may be surprised at how little temperatures increase with just clockspeed increases and no voltage bumps. It is mainly voltage that increases temperature.

Also consider that typically you can't raise the clockspeed much without voltage control..what couple hundred mhz?

You should test it and see. I did years ago when I thought the same thing (was told to test it just like you in fact!). While temps may go up a degree or so with a couple hundred mhz bump, it PALES in comparison to what voltage does.

i have spent more time "testing it and seeing" than you can shake stick at earthdog.. he he..

plus it all fits in with my simple logic.. if a chip is running 25% faster its doing 25% more work therefore it generates 25% more heat.. everything else being equal..

look at it this way.. heat at 100% say 60 C.. heat at 125% say 75 C.. when you add in the extra voltage needed to get to that 125% the heat ramps up even more.. you can soon get to 100 C and throttling..

to be honest i started off thinking it was mainly just extra voltage that caused extra heat then i realized it was actually a combination of both voltage and work being done..

the faster the chip runs the more work it does and the more heat it generates.. i think the same logic applies to running prime95 set on small FTT blocks.. when it all fits in the CPU cache the chip dosnt have to hang about waiting for data delivery gets more work done and generates more heat..

having said all that only silly torture tests do this kind of stuff and i think users really do need to keep all this in mind..

i know this is an overclocking forum but in real life that extra couple of hundred mhz on a cpu clock (over and above its "happy" speed) really aint worth bothering with.. i think 4.4 is a "happy" speed for the average 4790 cpu.. it will live a long and easy life.. he he

trog

 

[MrFuny]

i know this is an overclocking forum but in real life that extra couple of hundred mhz on a cpu clock (over and above its "happy" speed) really aint worth bothering with.. i think 4.4 is a "happy" speed for the average 4790 cpu.. it will live a long and easy life.. he he

trog

Yes, I think too

I just get new mobo: Gigabyte Z97X-UD3H - LGA 1150, 180W will be OK, he he

 

[EarthDog]

plus it all fits in with my simple logic.. if a chip is running 25% faster its doing 25% more work therefore it generates 25% more heat.. everything else being equal..

The problem is, that simple logic is flawed. Whatever testing you did was not done correctly. Sorry. First off, I highly doubt you can test that 25% theory on most chips as most chips won't overclock 25% without a voltage change. But let's say you took a 6700K at 4Ghz and overclocked it to 4.4GHz. That is a 10% overclock, right? Let's say P95 Small FFT load gets it up to 60C. 10% on top of that is 66C. It will not get 6C warmer with no voltage change and a 400Mhz overclock, though it will go up some (1-2C or so in my past testing). It is mainly voltage. Leave the clocks the same and add .1v more and watch your temps go up. Same amount of work, but those temps will go up. They go up A LOT faster with voltage than clockspeed.

Sorry Mr Funy about the side stuff... Perhaps this needs another thread.

EDIT: Something I found:

"As clock speed increases, voltage is constant and temperature increases linearly." If you increase clock speed without increasing the chip's supply voltage, the chip temperature will increase. That is because the heat per cycle being generated does not change, but the heat per second does - more cycles per second. And the thermal conductivity of the materials is fixed, so the material heats up as more heat per second is generated. May not be exactly linear, but the effect is less severe than the effect of voltage increases. However, as above, people who increase clock speed often follow that with an increase in supply voltage to regain stability, and the combined effect is significant. That's why overclockers spend so much effort on heat removal.

Let's see if this works... I bet @lilhasselhoffer would want to jump in here and throw down some mathszzzzz

I could be wrong, but, my testing years ago showed otherwise (it is NOT a linear rise as you state).

 

[trog100]

its not a linear raise and i wasnt being "exact".. i just used the figures i used for easy round figure reference..

other things affect it.. and i wont ague that voltage increases may have more effect than speed increases.. my chip ups the voltage as it ups the clock speed.. it isnt easy to figure out which part does what.. they do both play a part though..

i can pick say a fixed vcore of say 1.28 and run my chip at 4 gig and 4.6 just to see.. i run dynamic voltage with plus or minus off-sets normally.. guessing at the off-set is trial and error..

doing it with a fixed vcore is easier.. if i just change the clock speed and keep the vcore fixed it should tell me exactly what a clock speed alteration does in term of temps.. i do expect to see something roughly linear percentage wise.. he he

trog

 

[Toothless]

Yes, I think too

I just get new mobo: Gigabyte Z97X-UD3H - LGA 1150, 180W will be OK, he he

You don't need a new board. Just saying.

 

[EarthDog]

No, not 'may' have, voltage DOES have more effect, PERIOD.

doing it with a fixed vcore is easier.. if i just change the clock speed and keep the vcore fixed it should tell me exactly what a clock speed alteration does in term of temps.. i do expect to see something roughly linear percentage wise.. he he

And you wont see that. As I said, I did this testing before, YEARS ago, when I was under the same impression you had.

You opinion is formulated because of this:

my chip ups the voltage as it ups the clock speed..

ANd you can PM me the results, be it right or wrong so we don't torture this poor thread. Perhaps I will message the mods to split our conversation to a different thread...

 

[Aquinus]

The thing is that there are a few factors that come into play. The biggest source of leakage on a CPU is from transistors changing states, being turned on, off, or going high-Z. Between these states, transistors have increased leakage because resistance is low enough to allow current to flow but not so high to prevent the flow of electrons. As a result, you have a transistor acting as a pseudo-resistor for a short period of time. All in all, both @trog100 and @EarthDog are correct because there are two things that are happening here.

1. At the same voltage, higher clocks will result in higher leakage because of the number of transitions being made increases. Since state transitions is where most transistor leakage comes from, leakage at any constant voltage will increase as a direct proportion of clock speed. This only doesn't hold true if the amount of time it takes for transistors to change states decreases at higher clock speeds without a change to voltage (unlikely.)
2. Increasing voltage does two things.
   1. Decreases the amount of time for a transistor to be switched on or off however, depending on the type of transistor, this can be highly variable.
   2. Increases the amount of current going through a transistor, so when a state change occurs, leakage increases exponentially in relation to voltage, which is directly related to current.

With that said, heat scales to voltage a lot more because of how heat scales to current as an exponential function, clock speeds increase the amount of heat linearly, more or less (which is really over-simplifying the problem.) So even if clocks and voltage scale at the same rate, we're talking about a linear function versus an exponential function. There may be a point where clocks speeds increase heat more than voltage but, the thing is, if you add 0.1v to 1.1v, you're going to see less heat than adding another 0.1v to 1.2v and that every time you add a little more voltage, that "step" produces more heat than the last.

So you're both right but, @EarthDog I would say is more correct because there comes a point where voltage impacts heat a whole lot more than clock speed but, they both play a role.

When we're overclocking, we're putting a lot more voltage through these ICs which instantly makes voltage the biggest contributor.

 

[EarthDog]

Exactly.. some science behind it. Thank you Aquinas. I admit temperatures do increase with clocks, but it isn't 1:1 as Trog described. Voltage is, by far, the biggest factor for increasing temperatures. For a couple hundred Mhz with the same voltage, I consistently saw around 1-2C.

Anyway, I digress.

 

[BiggieShady]

leakage increases exponentially in relation to voltage

Just one small correction, since we are in a scientific mood, it's quadratic not exponential.

 

[Aquinus]

Just one small correction, since we are in a scientific mood, it's quadratic not exponential.

Fine, a less aggressive parabolic curve. You're right though because it's proportional to the square of the current, the current isn't the exponent. Words fail me.

 

[trog100]

i will spend half an hour running the 4 gig and 4.6 gig comparison at the same vcore.. obviously i will have to use a vcore voltage that runs the chip at 4.6..

i will post what i see..

trog

here we are.. it didnt take f-cking long.. he he

i see a 10 C difference not exactly linear from a percentage point of view.. but the core speed everything else being equal shows a fair increase in tempts at the higher speed..

i would also like to point out that at no time did i suggest that raising the vcore would not increase temps i simply said earthdog was wrong when he said a speed increase without a voltage increase would not increase temps.. quite clearly it does..

both voltage and speed play a part in the tempts..

trog

 

[EarthDog]

i would also like to point out that at no time did i suggest that raising the vcore would not increase temps i simply said earthdog was wrong when he said a speed increase without a voltage increase would not increase temps.. quite clearly it does..

I never stated, inferred, nor alluded to you thinking that. Also, I never said a speed increase without a voltage increase would not increase temps.

Here is what I said...

you may be surprised at how little temperatures increase with just clockspeed increases and no voltage bumps...

For a couple hundred Mhz with the same voltage, I consistently saw around 1-2C.

I also believe your severe overvoltage for the clockspeed (4Ghz) is playing a part in that. All my testing was at stock voltage and raising the clocks on stock voltage... or a voltage that started out appropriate for the clockspeed. 1.28v is a lot for stock. You would not be able to raise clockspeeds 600 Mhz on most CPUs without changing the voltage so this is probably not the best test.


EDIT: Your testing method also leaves a lot to be desired.. First, the length of the test is way too short.. its also different lengths of time. by over 50%. The cooler one was run for 54s, while the hotter one ran for 1:23s. My suggestion is to run Intel XTU stress test. It has a time on it that will stop it automatically (5 mins is default). Run that with Realtemp and you may have something worthwhile.... but after that, you should try it at stock voltages and a couple hundred MHz.

EDIT2: Not that it will make a difference in this testing, but you do know you are running RT for hex cores right? Run the regular EXE and you wont have that extra space on the side.

EDIT3: Part of it is my fault as I didn't share an empirical testing method and agreed to the raised voltage and 600MHz... my fault. After thinking about it, that makes sense considering what Aquinas said in how it works. The more voltage used, the higher the temps will raise (quadratic). So it makes sense the higher the voltage you have to start, the more the temps will raise with clockspeed increases. With less starting voltage, the temperature increase will not be as dramatic.

Done editing, LOL!

EDIT4: (I lied, didn't want to double post) - I will get screenshots later tonight, but I ran XTU stress test for 5 minutes at 4.25 Ghz 1.19v then 4Ghz 1.19v ("couple hundred" MHz difference). Here were the differences between the cores - 0/1/0/-1/1/0 (where -1 means the core was warmer at slower speed. I then dropped to 3.75Ghz 1.19v and here were the differences against the 4.25 GHz... 2/2/3/3/2/1. So 500 Mhz difference was within 1-3C. Again, PROPER SS's later with the proper screens (CPUz, etc) longer testing. This was quick and dirty and a sanity check for me, LOL!

 

[Schmuckley]

That's a bit warm for 4.4 Ghz.
Try umm..leaving EIST and all that on and offset voltage..
Hashwell..hmm..maybe can limit volts with input voltage.
That clock should need about 1.18-1.23v max.

 

[EarthDog]

Ok.. here is the testing. This is Xtu Stress test for 10 minutes each run with a 10 minute cool down back to ambient between runs (water temp back to ambient CPU margin of error close). It is cooled by the custom loop in the system specs for the 5820k.

Something worth mentioning.. The last test, at the slowest speed, looks odd doesn't it, with the highest temps? Its my luck that the heat in my house kicks on when the test starts. It raised the intake temp by almost ~2C by a couple minutes in (I'm under the vent). GAH.

So, I tried. But you can see for yourself on my system what 500 Mhz did at the same voltage.... not much at all. 1-2C give or take +1/-1 for error of instruments...

EDIT: This testing isn't perfect, I understand, but I think it was better data than the two minutes worth above...

 

[trog100]

its impossible to test clock speed changes without using a vcore that will run the chip at all the speeds involved in the comparison tests.. i used the minimum voltage necessary for stability at 4.6 gig.. my ambient temps didnt change while i was doing it ether.. he he..

one thing i can tell you is that using a voltage (1.11) more appropriate to the 4 gig clock speed the cpu tempts go right down from 65 C to 45 C..

45 to 65 is close to 50% more heat.. 50% more heat for a 15% performance gain aint good a swap.. once the chip is taken above 4.6 things get really silly..

one other thing is for sure.. if heat is a problem the easiest way to solve that problem is to lower the vcore and lower the clock speed.. in the real world the performance loss wont be noticed but the heat loss surely will be..

my own 4 gig to 4.5 gig comparisons show a 50% heat gain.. it seems to be split fairly evenly between extra core speed (extra work done) and extra voltage.. going over 4.5 gig the ratio between the two starts to change.. it does all come down to how hard any chip is being pushed and how close it is to its limits

trog

 

[Toothless]

Y'all scared the op away. Poor thing probably wasted cash on a new board when they didn't need it.

 

[EarthDog]

its impossible to test clock speed changes without using a vcore that will run the chip at all the speeds involved in the comparison tests..

Look at my testing. My stock voltage is 1.14 on this cpu. I added .05v to it in order to complete this testing with nearly a 25% overclock. You, like around .15v to .2v. BIG difference there. As I said the more voltage the higher temps will be and the more temps will go up with the same amount of work.

it seems to be split fairly evenly between extra core speed (extra work done) and extra

Your testing didn't show that. You didn't change the voltage to show how much it goes up comparatively.

@Toothless - I'm reporting the thread and asking the staff to split this off to a new thread.

 

[sneekypeet]

Pretty sure it does not work that way.

 

[EarthDog]

Pretty sure I do it a lot on a vbulliten based forum at least on a weekly basis. Is this a xenforo can't or something else?

Thanks in advance, and apologies for the drawn out aside and your assistance.

 

[sneekypeet]

Pretty sure I do it a lot on a vbulliten based forum at least on a weekly basis.

So you want me to start a thread....
Give it a title like Earthdog and trog....go!
All for the sole purpose of either trog or yourself winning that conversation? Please just take it to PMs if you need to continue with that topic