My new 4770k rig - temperatures

[Nec_V20]

Just built my new Haswell rig:
Mobo: Gigabyte X87Z-UD5H (BIOS F7)
CPU Core i7-4770k
Cooling: Corsair H100i

I ran the test IntelBurnTest 2.54 on my rig which has not been optimised in any way, everything is stock including the regulation of the fans which I have hanging off the 100i and I have it attached to a SATA connector instead of the one of the board four pin connectors.

I did the measurement with CoreTemp 1.0 RC5. The ambient temperature was 26-27 degrees Celsius.

My idle temps were:
Core1 33
Core2 30
Core3 32
Core4 33

I did the test with hyperthreading active and the test done with IntelBurnTest set to Stress Level Maximum. The maximum temperatures were:

Core1 79
Core2 77
Core3 76
Core4 73

I then turned hyperthreading off and used the same IntelBurnTest settings as above and the maximum temperatures were:

Core1 73
Core2 72
Core3 71
Core4 69

 

[Arkaign]

Stock clock?

 

[Nec_V20]

Stock clock?

Sorry about writing "IntelBurnIn" instead of "IntelBurnTest". Every time I write the word "Burn" I automatically write the word "In" when I talk about computers.

Yes, stock clock as well.

I just wanted a baseline of where the system is at present and I will be doing some optimisations to see if I can affect any improvement.

The clock on the cores was at 3.7GHz during the test (100*37).

I should add that I have had the system running 24/7 for the past two days (just basically idling) to bed in all the components.

 

[Jaydip]

looks a bit on the higher side to me.My 4770K after 15 mins of prime blend sits @ ~70C

 

[Nec_V20]

looks a bit on the higher side to me.My 4770K after 15 mins of prime blend sits @ ~70C

I used IntelBurnTest 2.54 because from what I have garnered this is meant to stress the CPU more than Prime95. I will give Prime95 a go now and see if there is any difference.

 

[ShintaiDK]

The IntelBurnTest is obsolete isnt it. Not even AVX2 support. SO dont use it.

You want to download and run this for maximum stress:
http://software.intel.com/en-us/articles/intel-linpack-benchmark-download-license-agreement

For stock operation. The Tjmax is 100C. Everything below that is perfectly fine and guaranteed for atleast 3 years.

 

[Nec_V20]

OK I have had Prime95 (I ran it with "In-place large FFT's" - the second setting - which is meant to maximise stress) running for about 20 minutes now and the temperatures with HyperThreading activated were:

Core1 69
Core2 68
Core3 66
Core4 64

ShintaiDK, even if IntelBurnTest is obsolete it still stressed the CPU more that Prime95 did.

 

[ShintaiDK]

Use Linpack 11 for maximum

 

[spat55]

My i5 3570k rests at 57c on prime65, with a Hyper Evo 212, funny thing is when playing Battlefield 3 it hits 60c, but I think that is slightly poor ventilation.

 

[JimmiG]

OK I have had Prime95 (I ran it with "In-place large FFT's" - the second setting - which is meant to maximise stress) running for about 20 minutes now and the temperatures with HyperThreading activated were:

Core1 69
Core2 68
Core3 66
Core4 64

ShintaiDK, even if IntelBurnTest is obsolete it still stressed the CPU more that Prime95 did.

The information in the Prime95 stress test screen is outdated. With Haswell, you'll get higher temps with Small FFT's.

Your temps seem a little on the high side but not abnormally high. Try checking for a BIOS update for your mobo. Some early BIOSes would set the votage incorrectly.

 

[Nec_V20]

The information in the Prime95 stress test screen is outdated. With Haswell, you'll get higher temps with Small FFT's.

Your temps seem a little on the high side but not abnormally high. Try checking for a BIOS update for your mobo. Some early BIOSes would set the votage incorrectly.

F7 is the latest BIOS and if you read the thread this, http://forums.anandtech.com/showthread.php?t=2337488
you'll know why I know this for sure

If you look further up this thread you will see that I stated that the tests were done with EVERYTHING AT STOCK, so that I have a baseline against which to measure any optimisations I will be undertaking.

I would however disagree with you, my results are on the lower side, if anything, of test results I have seen - short of delidding.

The five degree difference in temp between the the cores is probably due to the crap that Intel uses instead of a proper TIM and the amount of it they dump onto the CPU under the lid .

 

[T_Yamamoto]

Those temps look pretty high for someone using a AIO cooling system.

 

[Nec_V20]

Those temps look pretty high for someone using a AIO cooling system.

I would suggest you look at other results involving an H100i and the 4770k.

Also you will have noticed I said that I took the results with everything at stock (except for the fact that I replaced the Corsair fans with the much better Noctua NF-F12 fans). This includes configuring the Corsair Link software to work with the fans as it still does not have the fans running full speed at full load.

 

[Aikouka]

Those temps look pretty high for someone using a AIO cooling system.

On my i7-4770k, I hit 50C during Cinebench 11.5 at stock clocks, and I have a custom water cooling solution (includes 2x GTX680 in the loop) with 9 120mm fans and high-quality radiators. 70C doesn't really seem that far-fetched when you consider that it hasn't been delidded (nor has mine).

 

[Nec_V20]

On my i7-4770k, I hit 50C during Cinebench 11.5 at stock clocks, and I have a custom water cooling solution (includes 2x GTX680 in the loop) with 9 120mm fans and high-quality radiators. 70C doesn't really seem that far-fetched when you consider that it hasn't been delidded (nor has mine).

And I have also not yet optimised the fans on the H100i to run with the Corsair Link software (they don't yet get spun up to full speed under load). I have also not yet optimised the Mobo to run with the 4770k either.

I want to see if this is the fault of the Corsair Link or the fans. If push comes to shove I will try them out on the fan headers of the MB.

My results are only meant to be a baseline for me and I will see if I can do something to lower the temps.

When I do I will be documenting the results here so that others can gain from my success or witness my failure.

 

[Kenmitch]

Stock clocks as in optimized defaults with voltages on auto?

If so your temps are influenced by the xtra .1v Haswell feeds during anything with avx extensions.

Haswell = the 1st chip that needs a fixed vcore to get any decent overclocks with reasonable temps.

Did you monitor your vcore when you were testing.

My 4670k using stock settings uses the same vcore that it takes to get 4.2ghz stable using a fixed vcore. It's the xtra .1v that creates this situation.

I'm also using the H100i as I didn't want to hang 2lbs of heatsink off my motherboard.

My chip is stable at 3.8ghz with 1.00v's. Temps are decent at such a low clockspeed. My H100i is running in silent mode. Still need to get quieter fans for it.

I like to use AIDA64 on Haswell as it'll test more parts of the cpu. I still use Intel Burn Test to compare gigaflop output at different clocks.

Linpack 11 is the cooler killer absolute worst case scenario test I think. Not sure Haswell could run stock with the stock cooler without throttling in Linpack 11.

 

[JimmiG]

I would however disagree with you, my results are on the lower side, if anything, of test results I have seen - short of delidding.

The five degree difference in temp between the the cores is probably due to the crap that Intel uses instead of a proper TIM and the amount of it they dump onto the CPU under the lid .

My temps maxed out at 56C in Prime95 @stock, no delid. However I have a CPU that loves low clock speeds and voltages and absolutely hates being overclocked. The important thing (for overclocking) is how the CPU reacts to increased voltage and clock speed.

Yeah, after delid, my cores now all max out at exactly the same 81C in IntelBurnTest at 4.4 GHz / 1.24V. Before, they were all over the place.

My 4670k using stock settings uses the same vcore that it takes to get 4.2ghz stable using a fixed vcore. It's the xtra .1v that creates this situation.

It's a PITA. I've found that AVX and especially AVX2 actually do require a higher voltage than everything else, but not an increase of .1v like Intel seems to think.

The end result is that using adaptive VCore is like trying to juggle while riding a unicycle. You need to keep the idle Vcore high enough to avoid instability in the low power state, the VCore under "normal" load needs to be high enough to maintain stability, and the VCore in AVX loads needs to be kept low enough to avoid the temperature going out of control. Sometimes, this is simply impossible to achieve with a single VCore offset option. If a mobo manufacturer would find a way to implement an "AVX Voltage Offset" option, that would be really cool.

 

[Arkaign]

81C delidded @ 4.4Ghz?

:'(

I guess that cements it, I'm never going back to Haswell, unless they have a particular batch that is easy to identify for cold running 4.6+ on a 212+. As it is, even with an H100 my 4770Ks were all insanely hot.

 

[Nec_V20]

Stock clocks as in optimized defaults with voltages on auto?

If so your temps are influenced by the xtra .1v Haswell feeds during anything with avx extensions.

Haswell = the 1st chip that needs a fixed vcore to get any decent overclocks with reasonable temps.

Did you monitor your vcore when you were testing.

My 4670k using stock settings uses the same vcore that it takes to get 4.2ghz stable using a fixed vcore. It's the xtra .1v that creates this situation.

I'm also using the H100i as I didn't want to hang 2lbs of heatsink off my motherboard.

My chip is stable at 3.8ghz with 1.00v's. Temps are decent at such a low clockspeed. My H100i is running in silent mode. Still need to get quieter fans for it.

I like to use AIDA64 on Haswell as it'll test more parts of the cpu. I still use Intel Burn Test to compare gigaflop output at different clocks.

Linpack 11 is the cooler killer absolute worst case scenario test I think. Not sure Haswell could run stock with the stock cooler without throttling in Linpack 11.

Yes, everything is on "Auto". As I said, I wanted a baseline and I wanted that baseline after the system was basically bedded in. For instance with regard to the thermal paste on the H100i, I don't know if that stuff needs any time to cure - do you?

I want to eliminate any false negatives and conversely any false positives. I am in no hurry.

One of the things I want to do over the coming days and maybe even weeks is to thoroughly read up on various aspects of the new platform as it effects my own system.

The reason for this is simple. I don't want to change anything without knowing exactly what that change will mean for the future.

I am a great fan of "RTFM" anyway, so I will look at what you have written above and will look at other things as well and see what changes will give me the biggest long-term benefit and then look at which changes will do the same for the short term.

It is only when I know what I am doing that I will start implementing changes.

 

[santilmo]

Would like to join this thread as I am also pondering on Haswell but before I do, I'd also like to get the "facts" about Haswell temps for real.

At stock clocks/voltage only, is Haswell generally hotter than Ivy? I am particularly interested with 4670K vs. 3570K idle and load temps...

Thanks!

 

[Kenmitch]

Would like to join this thread as I am also pondering on Haswell but before I do, I'd also like to get the "facts" about Haswell temps for real.

At stock clocks/voltage only, is Haswell generally hotter than Ivy? I am particularly interested with 4670K vs. 3570K idle and load temps...

Thanks!

Short answer Yes.

During normal use they would be similar.

Anything hitting the AVX extensions will cause Haswell to run hotter.

 

[Aikouka]

So, I decided to try to stress test my processor some more today, and I ran the Intel Burn Test on it. I recall reading that it's not the best option, but... it certainly turned the heat up a few notches on my Haswell. I ended up topping out at about 70C at stock on my custom water loop. Maybe I just did a poor job at applying the PK-3 this time. I did put a bit less of it than normal, because it seemed like I put too much on it the previous times.

 

[santilmo]

Short answer Yes.

During normal use they would be similar.

Anything hitting the AVX extensions will cause Haswell to run hotter.

Ok, great! But pardon my ignorance though, in what instances that AVX extensions will be used? Are there any particular software or application used for this or by this?

 

[Rvenger]

81C delidded @ 4.4Ghz?

:'(

I guess that cements it, I'm never going back to Haswell, unless they have a particular batch that is easy to identify for cold running 4.6+ on a 212+. As it is, even with an H100 my 4770Ks were all insanely hot.

I had the same exact experience as you. My last one was a real scorcher at under 1.2v.

 

[Nec_V20]

Would like to join this thread as I am also pondering on Haswell but before I do, I'd also like to get the "facts" about Haswell temps for real.

At stock clocks/voltage only, is Haswell generally hotter than Ivy? I am particularly interested with 4670K vs. 3570K idle and load temps...

Thanks!

Shortly after starting this thread the pump on my Corsair H100i went.

However it did show me something with regard to the Haswell that I would probably not have realised otherwise:

I have built up my new system based on the i7-4770k.

Mobo: Gigabyte Z87X-UD5H
CPU: i7-4770k
Cooler: Corsair H100i
PSU: Corsair AX860
Chassis: CoolerMaster HAF XB

I noticed that the light on the H100i cooler was no longer on whilst I was working on it configuring Windows and migrating data. I checked and the pump was no longer working.

Windows however had not crashed, so I loaded CoreTemp and the 4770k was only at around 45 degrees Celsius.

I thought, "That's weird", and loaded up FireFox and did a bit of browsing and the temperatures remained stable.

I double checked that the pump was indeed dead - which it was - and I could notice the temperature going up in the BIOS. I still wasn't convinced, so I took the pump off the CPU entirely.

When I rebooted into Windows again I kept an eye on the temperatures for all of half an hour.

I am now working on the machine and I am writing this post from it without the cooler (which I am sending back to get replaced) on it any more, just the bare CPU.

It goes to show, I have gotten so used to CPUs needing cooling or frying that I never thought of the low energy consumption of the Haswell CPUs having this bonus. The temperatures are still stable at around 45 degrees and the CPU itself is warm to the touch, not hot.

So all in all the first few days of the new 4770k system have been interesting to say the least.