6700k temps?

[ioni]

I just picked up a 6700k a few days ago and just OCd to 4.5 tonight and not sure if my temps are on the high side or not. I know these chips tend to run a little on the warm side, but I thought my nepton 280L would do better than 70-75C after running IBT. Just wondering if those are reasonable temps. This was with all my fans set to full blast as well.

 

[YBS1]

That's pretty much the best you can expect without a delid. Nothing to worry about.

 

[BonzaiDuck]

That's pretty much the best you can expect without a delid. Nothing to worry about.

I have to agree.

I got a delidded and CLU-relidded chip from Silicon Lottery. It had been binned as validated running at 4.8 Ghz proven with ROG RealBench for 1 hour at a voltage of "1.424V (or less!)"

I opted against water-cooling, as a calculated guess of what would happen using a really good heatpipe cooler with a temperature reduction of 12C from the CLU-relid. That part proved quite true: I discovered an i7-6700K review with a stellar testbed motherboard and an EXOS dual-fan water-cooling system -- an "external" radiator assembly as we would recall from EXOS.

The reviewer was able to get to 4.8 Ghz, but not for "24/7" use. At that speed, his temperatures passed 90C; the system became unstable from the heat. At 4.7, the reviewer was gratified, noting that his temperatures (with a factory-original chip and no CLU_relid) were summed up in a peak package temperature of 82C.

With my "special" chip and a good air-cooler, I can run IntelBurnTest, OCCT: Linpack, and several milder tests at 4.7 Ghz and my peak package temperature never exceeds 69C.

But . . . . despite the folklore passing around about Prime95, I found better advice about it: it's destructive potential is limited to the high temperatures it creates and the length of time you choose to run it. It is no longer as feasible for "marathon" tests; it's best use is finding errors early that would turn up later through lengthier but milder tests.

So I found that Prime95 at 4.7 also pushes my temperature to between 69C and 70C if we're talking about maximum package temperature. My cooling solution was quite effective, and water-cooling might net me only another 5C in improvement -- if that.

I still figured, however, that you should be able to pass Prime95 "Blend" for some period of time. And to get it to stay stable and error-free for a 2-hour run, I needed to bump up the voltage about 8 millivolts at 4.6 Ghz. For 4.7 Ghz, the adaptive voltage would have to be 1.384 to pass the other tests, but I had to bump it up to 1.408 or higher to pass even an hour of Prime95.

I also noted that the mobo makers overvolt the processor at stock speeds so monitor software will show an unloaded vcore of 1.39V, but stock settings will run well at around 1.20V. So I figure advice to keep voltage below 1.40V was wise enough to heed.

This all boils down to a conclusion that my cooling is quite effective, but cooling doesn't buy you everything. For me, passing a couple hours in Prime95 means bumping up VCORE to 1.35/1.36V for a speed of 4.6. All the other tests will pass for several hours with VCORE showing at around 1.33 to 1.34. And if I can't do 4.7 at 1.40 or less, then we're just not going to choose 4.7 as a good 24/7 speed. Yet, my temperatures are stellar at 4.7: even the most strenuous stress test won't push the temperature beyond 2C exceeding the processor TCASE spec.

And also -- the motherboard choice will limit the overclock. 16-phase power design will (and has --) proven better overclocks than 12-phase. so a top-tier board may buy you 100 Mhz -- maybe more, probably not less.

But, as I already said, good temperatures will get you a dime for a cup of coffee if you choose to limit voltage reasonably and still pass tests which many of us now simply avoid.