G3258 ... 60.1w ... 81c?

[VirtualLarry]

I'm trying to figure this one out. I'm guessing it's poor case cooling, because there's no intake fan in the case, but bear with me.

Got a G3258 / GA-H81M-DS2V combo. In a Topower case with an Antec Earthwatts PSU (top-mounted). Case has an exhaust fan. Currently has side off, in a desk cubby.

Cooler is the copper-cored stock Intel 1150 heatsink. Supposedly it's good for 95W.

According to CoreTemp RC6 in Win7 64-bit, running PrimeGrid on both cores, and Einstein@home on a 7790 1GB (51C according to GPU-Z), the CPU reaches temps of 80C on Core 0, and 70C on Core 1. The 10C difference is consistent. Power usage is 60.1W.

How come, a heatsink designed to dissapate 95W, is getting to 80-81C on one core, if the CPU is only using 60W?

 

[ShintaiDK]

How did you measure 60.1W?

Whats the fanspeed? My fanspeed for example is not max at 80C.

I assume you run stock speed with stock voltage?

 

[VirtualLarry]

Sorry, forgot to mention in this thread that it is OC'ed, from 3.2 to 3.8, 1.2009v (VID according to CoreTemp) / 1.201v (voltage according to CPU-Z).

CoreTemp shows a realtime indicator of how much power the CPU is drawing (according to the onboard PCU sensors on-die), and it shows 60.1W.

 

[DrMrLordX]

The delta T between core 0 and core 1 is significant no matter what the power draw. Could be a problem with the TIM under the IHS. I'm going to assume that you'd used the proper amount of TIM and that your mount is good. It would be worth checking that if you haven't (though you probably have).

I had some problems like yours years ago on my old Propus (well, sort of) when I applied an insufficient amount of TIM to the IHS surface. Basically, the IHS was developing hot spots where it couldn't properly spread heat to certain parts of the HSF base. So, my NB was getting too hot and was unstable, and I seem to recall having at least one core showing abnormal temperatures.

With a die size as small as the G3258's, I would think such an issue would be far less likely.

 

[VirtualLarry]

I know, it sounds like a bad mount, right? I've never seen such a large delta between core Tjunction temps on an Intel CPU, especially under load.

But I know that I assembled the heatsink to the mobo outside the case, and I always turn the mobo over to make sure that all of the "tips" have seated correctly on the underside of the board. It's kind of bizarre, really. Maybe the heatsink or mounts were just defective somehow.

Anyways, I'm not especially worried about the temps (as long as they don't get hotter), but I'm just surprised to hear that a heatsink that supposedly handles 95W chips is getting this warm with a chip that is reporting that it draws 60W. I would think that would be a walk in the park for this heatsink.

Edit: The 60W may be only the "core" power levels, and not the "uncore". I don't know how much that draws.

 

[Yuriman]

I've seen around 10c difference between my hottest and coldest cores, before delidding.

 

[PG]

So lets summarize things: low amounts of air flow, a gpu which is also probably dumping some heat into the case, cpu overclocking, 100% cpu and gpu load, and with all that the cpu is still well below the point where it would throttle. I think it would throttle around 95 or 100, but not 100% sure. An i5 or i7 with that same heatsink hits higher temps at stock speeds under load so I"m not too surprised at your results.

I think you are fine and I don't see anything too unexpected.
You could take the side off the case for an hour or so and see how the temps look.
I would not remount the heatsink, but I am assuming you used the stock heatsink and whatever thermal paste Intel put on there. It works fine and you know you got it on right since you saw the tips on the bottom.
As you said, if you can improve airflow that would be your best bet.

 

[dbcooper1]

Isn't that CPU including gpu rated at less than 60 watts at stock voltage and speed?

 

[PG]

Yes, but he has it overclocked to 3.8.

 

[Greenlepricon]

My g3258 hits 80C on the stock heatsink at 3.9GHz. Sounds like a pretty normal occurrence to me. If I turn the fan all the way up, it can manage the temps pretty well however, and stays in the mid 70s. The issue there is of course noise. I've remounted it to see if that was the problem, but same results.

A large part of it is efficiency of the chip, and of course factors such as the IHS giving problems. In summary though, don't worry that it's not removing as much heat as it says it should.

 

[crashtech]

Do you have to run 1.2V to get 3.8GHz?

 

[Techhog]

Do you have to run 1.2V to get 3.8GHz?

Yeah, that sounds ridiculous, though I doubt that board has voltage control.

 

[crashtech]

Yeah, that sounds ridiculous, though I doubt that board has voltage control.

Pretty sure it does. I have owned one.

 

[Techhog]

Pretty sure it does. I have owned one.

Oh. That's way too high then. My 4670K is running 4.2GHz at 1.18V, and I think it can even go lower than that. 1.2V for the Pentium seems really high.

 

[Greenlepricon]

My pentium does 4.1 on ~1.1. It's the heat that kills that overclock. 1.2V should get it higher than 3.8 unless that chip is a complete dud

 

[crashtech]

Some chips are duds. I got one G3258 that would do 4.9, and another that couldn't even reach 4.0.

 

[Techhog]

Some chips are duds. I got one G3258 that would do 4.9, and another that couldn't even reach 4.0.

That's quite the variance, even for Haswell.

 

[crashtech]

I can't rule out some factor that might be skewing my results, the second G3258 will be tested again with a nice Z97 board in the next few weeks.

 

[VirtualLarry]

Pretty sure the CPU I got is a dud. Requires something like 1.15v at stock (3.2Ghz). I just have vcore on AUTO. Board is limited to 1.2v.

 

[crashtech]

Sell it and try again!

 

[DrMrLordX]

I know, it sounds like a bad mount, right? I've never seen such a large delta between core Tjunction temps on an Intel CPU, especially under load.

But I know that I assembled the heatsink to the mobo outside the case, and I always turn the mobo over to make sure that all of the "tips" have seated correctly on the underside of the board. It's kind of bizarre, really. Maybe the heatsink or mounts were just defective somehow.

Anyways, I'm not especially worried about the temps (as long as they don't get hotter), but I'm just surprised to hear that a heatsink that supposedly handles 95W chips is getting this warm with a chip that is reporting that it draws 60W. I would think that would be a walk in the park for this heatsink.

Edit: The 60W may be only the "core" power levels, and not the "uncore". I don't know how much that draws.

Sounds like you've practiced due dilligence, then.

I've seen around 10c difference between my hottest and coldest cores, before delidding.

My mind was moving in the same direction. You may be suffering from a TIM gap under the IHS. Whether or not it's that important to you to drop the other core down 10C (or more, actually) is strictly up to you. Obviously you don't want to delid if you're going to sell the chip and try again. But, if you want to practice a Haswell delid, it won't get much cheaper than that. Got any CLU on hand?

There is also the distinct possibility that bringing the temps down will buy you little to no extra headroom.

 

[VirtualLarry]

My biggest disappointment with this CPU, is with DC. With my i3-3240, I could have BOINC crunch on three cores, and it still left enough CPU time due to HT for the 0.5 cores that the GPU compute thread needed to feed the GPU.

With the G3258, if I run two PrimeGrid CPU WUs, and a 0.5 CPU GPU WU, the GPU WU gets starved out.

Perhaps I'll end up getting a pair of Haswell i3 CPUs down the line.

 

[Toro 45]

Pull that H/S off and redo it, there's no reason for that cpu to be that hot with 1.2v. unless your room is sweltering.

 

[sm625]

I have two G3258. I run them at 1.3V, but I did do a bit of testing at 1.2V. 75C average loaded core temp is about right. And I also noticed wild fluctuations in core temps, one core, both cores, you name it. More than 10 degrees C, actually. At 1.3V its even bigger, nearly 15C deltas. I suspected throttling but never saw any actual evidence of it. Benchmark scores are consistent, and intel XTU didnt detect throttling.