Something wrong with haswell cpus? High temps.....

[cytg111]

prime95 avx version, http://www.mersenneforum.org/showthread.php?t=16535.

You running stock cooler ?

Oh and you got a devil? Different ballpark then.

 

[tidy]

.

 

[notty22]

I have a 2 month old 4670k and GA-Z87X-HD3. No throttling at stock.
Using a H50.
Can run 4200mhz (all cores) o/c no throttling from heat. Have a watt setting raised in bios.
Intel® Extreme Tuning Utility (Intel® XTU)

is helpful. You can mouse over the throttling reading and get a overview of throttling. %'s and such. With correct settings to wattage and such I get 0%.
Now I can o/c higher up to 4500-4700mhz but the H50 is not up to it, during the summer. Under Prime, it will cascade up to 100 after 10 minutes. Normal loads it won't.

 

[cytg111]

I was already using a newer 28.X version, does that still have the AVX tests?
Noctua NH-U12S.
OP states a 4970k (albeit mistyped), isn't that what we're talking about here?

Sorry my bad, yes of course. I suppose it should be AVX enabled then yes.


notty22 -> HT may be a culprit as well, ill ill give it a go with both 4 and 8 threads.

 

[BonzaiDuck]

I'm a bystander on all this -- including the Haswell refresh and Z97 chipset motherboards. I was "planning" to build a Haswell-E system next year. So I had done some quick searches over the last week to assess whether I'd rather use the i7-4790K or the i5-4690K instead of an "E" processor.

I stumbled across an Asian's post of his experience with a custom water loop -- which seemed adequately "prodigious." He didn't bother to test the new "polymer" TIM: he just went ahead and delidded the processor, replacing the TIM with CLU. I'm pretty sure he cited a load temperature of 69C.

This seems to be the opposite direction we would've expected on this chip. We'd think they improved the fabrication process with this new "polymer" material, but with the results I mentioned in the last paragraph, something else has made for a worse outcome.

. . . and now I have to say this about "planning." "The best laid plans of mice and men " etc. etc. Or -- you can "plan" all you want, but it only matters in successful "execution."

 

[notty22]

Sorry my bad, yes of course. I suppose it should be AVX enabled then yes.


notty22 -> HT may be a culprit as well, ill ill give it a go with both 4 and 8 threads.

Sure, HT does historically show higher temps at a given clock speed. Stock, though me thinks it should not throttle. Unless you have possible MCE enabled and no wattage setting disabled. Of course you are not stock then, but the bios's in these enthusiast boards have many settings, and the 'default', at times is not quite to the letter of Intel stock definition. Leading to maybe more end user performance or throttling in your case?
Possibilities.

 

[Xed]

Exactly what test and version of Prime are you using to get those temps? I will run it on mine to compare.



What constitutes a 'ridiculous level'?

Nearly 1.5v.

Unless I missed it, I don't think the op has mentioned what voltage is being applied or if he's updated the bios.

 

[cytg111]

Here 8 and 4 threads yields the same endresult, throttling, 8 threads will get there under 10 seconds, 4 takes a little longer :

Throttling will take it down around 1.5 to 2.5GHz

Running 1 thread i can push it to 100 degress too with turbo, it will back down a little to 3.5-3.8 and run in that area.

Running 2 threads and we are back in throttle terretory pretty fast again, now ranging 2.4-2.8GHz

Running 3 threads+ would be redundant

reproduction
- should get version 28 for avx2 punishment - http://mersenne.org/download/
- run tourture test (small FFTs, max heat)

So, how do your haswell temps stack up ?

 

[flapjacky]

So, how do your haswell temps stack up ?

Dont know where you check core voltage on hwmonitor but if its around 1.2 like VID its not normal.

either a bad chip or bad heatsink installment. I have the same noctua u12p + 4670K and i im miles ahead of you (under you) in temps

edit: for a reference im getting max 85c @ 1.2V 4,4Ghz -- prime95 large TTL's

 

[EightySix Four]

I know this is odd, but can you take a picture of how much thermal paste you use?

 

[Techhog]

Here 8 and 4 threads yields the same endresult, throttling, 8 threads will get there under 10 seconds, 4 takes a little longer :

Throttling will take it down around 1.5 to 2.5GHz

Running 1 thread i can push it to 100 degress too with turbo, it will back down a little to 3.5-3.8 and run in that area.

Running 2 threads and we are back in throttle terretory pretty fast again, now ranging 2.4-2.8GHz

Running 3 threads+ would be redundant

reproduction
- should get version 28 for avx2 punishment - http://mersenne.org/download/
- run tourture test (small FFTs, max heat)

So, how do your haswell temps stack up ?

Hold up. What is up with that VCORE? That has to be some kind of glitch, right?

 

[crashtech]

I thought I must have been reading it wrong. I have to keep my 4770K below 1.3V or the temps just go insane. Currently I have it maxxing out at about 1.28 and it will still throttle a bit under testing because the ambient temps are so hot. I need a better cooler.

 

[Techhog]

I thought I must have been reading it wrong. I have to keep my 4770K below 1.3V or the temps just go insane. Currently I have it maxxing out at about 1.28 and it will still throttle a bit under testing because the ambient temps are so hot. I need a better cooler.

HW Monitor's VID is what we normally call vcore, though (as in, the VID = CPU-Z's core voltage). I have no clue what VCORE is.

 

[ShintaiDK]

Here 8 and 4 threads yields the same endresult, throttling, 8 threads will get there under 10 seconds, 4 takes a little longer :

Throttling will take it down around 1.5 to 2.5GHz

Running 1 thread i can push it to 100 degress too with turbo, it will back down a little to 3.5-3.8 and run in that area.

Running 2 threads and we are back in throttle terretory pretty fast again, now ranging 2.4-2.8GHz

Running 3 threads+ would be redundant

reproduction
- should get version 28 for avx2 punishment - http://mersenne.org/download/
- run tourture test (small FFTs, max heat)

So, how do your haswell temps stack up ?

Something is terrible wrong with your system if thats what you see. Specially if you throttle even with 1 thread. And I assume its your VID since its showing 1.2V. Mine peaks at 1.107V. Possible also your cooling solution. 1000RPM CPU fanspeed?

No AC here, summer and using stock cooler in a MiniITX chassis. Ambient case temp is 30C using a EVGA z87 Stinger +i5 4670. Ambient temp peaked at 39C under test.
1 thread with Prime95 28.5 peaks at 70C at 1.107V.
4 threads with Prime95 28.5 peaks at 92C at 1.092V

Non AVX loads is 0.100V lower.

 

[tidy]

.

 

[cytg111]

Dont know where you check core voltage on hwmonitor but if its around 1.2 like VID its not normal.

either a bad chip or bad heatsink installment. I have the same noctua u12p + 4670K and i im miles ahead of you (under you) in temps

edit: for a reference im getting max 85c @ 1.2V 4,4Ghz -- prime95 large TTL's

- Well, you are runnning a noctua u12p, while I am using the stock cooler

I know this is odd, but can you take a picture of how much thermal paste you use?

- The exact amount that the stock cooler is equipped with

Something is terrible wrong with your system if thats what you see. Specially if you throttle even with 1 thread. And I assume its your VID since its showing 1.2V. Mine peaks at 1.107V. Possible also your cooling solution. 1000RPM CPU fanspeed?

No AC here, summer and using stock cooler in a MiniITX chassis. Ambient case temp is 30C using a EVGA z87 Stinger +i5 4670. Ambient temp peaked at 39C under test.
1 thread with Prime95 28.5 peaks at 70C at 1.107V.
4 threads with Prime95 28.5 peaks at 92C at 1.092V

Non AVX loads is 0.100V lower.

- Everything is stock and auto, the only thing I've messed with is MCT in the BIOS.
Yes I could set a fixed vcore get a better cooling solution etc. however I dont see the problem in regular workloads, ie. will play a AAA title maxed just fine no throttling, so the actual *need* is an artificial one.

"4 threads with Prime95 28.5 peaks at 92C at 1.092V" ... now add 0.1V and hyperthreading .. you dont see that hitting 100C ?

 

[ShintaiDK]

"4 threads with Prime95 28.5 peaks at 92C at 1.092V" ... now add 0.1V and hyperthreading .. you dont see that hitting 100C ?

1.092V is with +100mv for 256bit AVX. Else it runs at 0.992V.

And since you can throttle with even 1 thread. Something is wrong with your system. And if you use the stock cooler as you say. Then the CPU FAN shouldnt be 1015 RPM as your screenshot shows, but more around 2800 RPM.

 

[cytg111]

Something is wrong with your system.

- Sort of the point, indeed, and moreso how many else has this issue with haswell "everything at stock". The problem is determining this demographic - finding ppl with stock systems and getting them to run prime95 plus a hardware monitor. however judging from all the "haswell - my temps are insane" stories around, i am betting i am not the only one.

 

[ShintaiDK]

- Sort of the point, indeed, and moreso how many else has this issue with haswell "everything at stock". The problem is determining this demographic - finding ppl with stock systems and getting them to run prime95 plus a hardware monitor. however judging from all the "haswell - my temps are insane" stories around, i am betting i am not the only one.

Well your VID is too high, your fanspeed is way too low. Unless its something entirely different HWMonitor reports.

And then we can guess on all the other parts that Asrock might have been so friendly to set outside the stock values for you.

Its interesting to see if its more an issue with motherboard makers than with the CPU.

 

[tidy]

.

 

[Durp]

I can't understand how you guys are managing to get these temperatures. Are you configuring Prime in any way other than just starting the default small ftt test on all cores?

Your ambient temps are lower than usual. You also might have gotten a proper tim/glue application from Intel.

Others get too much glue and air gaps in their TIM and have 25-26c ambient.

 

[tidy]

.

 

[cytg111]

Its interesting to see if its more an issue with motherboard makers than with the CPU.

Yes, I guess that is why intel is dropping FIVR again come skylake right? . Nothing to see here, pass the street please. But hey, only idiots dont fess up to their mistakes so hats off for that.

 

[ShintaiDK]

Yes, I guess that is why intel is dropping FIVR again come skylake right? . Nothing to see here, pass the street please. But hey, only idiots dont fess up to their mistakes so hats off for that.

Yet its back in Icelake.

You should rather focus on whats pointed out, but it seems you dont have any interest in fixing your system that even throttles with 1 thread.

 

[cytg111]

oh I will, trust me, I will. I aint in front of the system right now thou'

Back in icelake? Lets hope a single beta run is enough then.