You picked the hottest test of Aida64. If you want to check for stabiltiy tick CPU/FPU/CACHE/MEMORY. FPU is the most stressful test for Aida64.

At 1.29 vcore I hit 90's instantly when running FPU alone.

 

Thanks for your reply.

I'm still amazed at the jump in wattage, for such a small voltage increase.

Default CPU settings, result in a similar wattage during the Aida cpu test, which is well outside Intel's spec of 88watt.

Xerphiel.

 

Additional note. Running with the stock cooler, cpu settings set back to default, Aida fpu test, temps hit 98c within a minute.

Is this normal? Stock cooler seems grossly inadequate. Maybe its not specced for synthetic benchmarks...

 

your temps are normal for 4790k. The stock intel cooler was designed for 4770k stock tdp. 4790k is really a factory overclocked 4770k and intel doesnt expect overclockers to use stock cooler, and it wasnt redesigned for 4790k. Though intel stock cooler will work fine for 4790k, as long as not running synthetic testing programs.

But everyone that buys a 4790k has 2 options. 1) accept the high temps or 2) delid and accept the risk, but get 20+C lower temps.

If you rma your 4790k, you will get another with similar temps, since your temps are normal for 4790k.

 

Thanks for the information, reps up.

 

Also avoid avx based stress test because this lastest avx instructions set is able to use the cpu 100% which will overheat your cpu

 

If the vcore is set on auto that will add to the heat issue at stock. Try setting the vcore manually to 1.10v

If its stable there you can try to lower it again to 1.050v and it will definitely be cooler.

Intel got a little high with the stock voltages on most of theses cpus.

 

Just to put things in perspective, my Sandy Bridge 2500k 5.0ghz 1.58v was same temp as a Haswell/Devil Canyon 4690k @ 4.5ghz 1.2v

 

Quote:

Originally Posted by Wirerat

If the vcore is set on auto that will add to the heat issue at stock. Try setting the vcore manually to 1.10v

If its stable there you can try to lower it again to 1.050v and it will definitely be cooler.

Intel got a little high with the stock voltages on most of theses cpus.

This. Every guide I've read on overclocking recommends NEVER using auto. It gives you a lot more juice than you need.

 

Quote:

Originally Posted by prescotter

Just to put things in perspective, my Sandy Bridge 2500k 5.0ghz 1.58v was same temp as a Haswell/Devil Canyon 4690k @ 4.5ghz 1.2v

Haswell/DC is 22nm, versus sandy 32nm, and die shrink of haswell increases power density which increases temps given same volts.
Haswell/DC has the fivr on die which significantly increases temps over sandy.
Haswell/DC has paste tim, Sandy is soldered, which is about 20+C difference.

So one would expect a 30-40+C difference between Sandy and Haswell/DC just based on the above, not to mention other architectural differences.

As for running avx or fma3, that is only a problem if havent delidded and replaced tim1 with liquid metal. Since only real issue is paste for tim1 isnt suited for tdp of ~130W and above.
24 hours prime 28.5 with avx/fma3 at 4.7 and 1.29v on my DC but tim1 replaced with CLP, max temps are 78C, most time around 60's. ambient temp is 25C.


and 10 mins with prime 27.9 (avx only) at 5ghz, 1.47. prime 27.9 has lower temps than prime 28.5. Cant run prime 28.5 at 5ghz because trips motherboards overcurrent shutdown even set to highest setting.

 

Really wish I had the guts to delid

 

Quote:

Originally Posted by Pikaru

Really wish I had the guts to delid

It doesn't look too bad with a vise and block. Doing it with a razor qualifies a person for the Darwin award if you ask me.

I wouldn't do it myself, but, I don't even OC my 4790k. The thing is a beast at stock.

 

Yes i know about the IVR, the bad TIM and architecure differences..

But still it comes down to roughly "double" the heat for 10% increased performance going from Sandy to Haswell.

Wish Intel would just solder them but i guess it would give us too high of a OC margin.