Corsair H90, 550D Case, with 3930K Overclock Temp/Stability

[TechRaven]

Build Specs
----------------------------------------------------------
Case: Corsair 550D
Motherboard: Rampage IV Extreme
CPU: Intel 3930K (OC @ 4.2Ghz)
RAM: 32GB (4x8GB) Corsair Dominator 1866MHz (Planning on added the other 4x8GB for 64GB soon)
Cooler: Corsair H90 (top mounted, intake)
HD: 4x 3TB WD RED NAS in Raid5
HD: 2x Samsung 840 Pro 256 (not raid0, the risk of full failure is too high, just separated usage patterns)
----------------------------------------------------------
Voltage Offset : -0.025 (0.882 Idle, 1.288 Full Load)
C-State: All enabled
Speedstep: Enabled
VSSCA: 1.15 Manual
CPU PLL: 1.85
LLC (Load Line Cal): Regular
----------------------------------------------------------
Stablility Testing: Newest version of Prime95 and IntelBurnTest.
Memory Testing: Ran through 2 passes of Memtest RC1 at 1600 and 1866
----------------------------------------------------------

At first I used the stock paste on the corsair and with the radiator mounted as a exhaust on the case with little luck, temps hit of 88c and up pretty quick (within a single very high IBT run). I moved the radiator to an intake at the top of the case and took off the cover which unfortunately increases the noise quite a bit more then I was hoping, but with the radiator as the rear exhaust it was a good 5-8c hotter.

I remounted the heatsink and used Artic Silver 5 paste (I spread a thin layer with my finger as per directions, and prior use, I know some people say this causes issue but i've had ok luck with prior builds and doing this).

One oddity is that CPUID's Hardware Monitor seems to report temperatures 10c above RealTemp/HWInfo/CoreTemp, I'm unsure which one I should be trusting? I'm learning towards CPUID is reporting things incorrectly (Atleast I hope, or i've gotten the CPU to hit 92-96C which seems that it should have been in thermal throttling according to intels spec).

I also see that Core 1/Core 5 run about 10C hotter then the other cores, at idle currently its at 38, 30, 29, 29, 32, 37, I've seen lots of people report they have differences between the cores and its too be expected, but most other people seem to be worried about a 3-4C difference (which I wouldn't be shocked about), i've not seen anyone showing a 10C difference. The Full Load temperatures seem to be similar, core 1/5 will be at 82c the rest at 72-75c.

Going anywhere north of 4.2GHz and the system will BSOD pretty quickly with a Watchdog timeout (windows 8's awesomely info lacking BSOD). Trying to use a BStrap of anything other then MHz and the system will not post.

From what I've seen around most people are at higher voltages with worse coolers and lower temperatures, although sometimes its hard to tell as some people run old versions of Prime95 that don't stress the CPU very well at all.

My last build was a i7-920 which was able to hit 4.0 GHz relatively easily on air, I had posted several times up to 4.5Ghz, but usually ran it at 3.9GHz just for not taxing the CPU. I was hoping to be able to get this 3930k to atleast run at 4.4 to 4.6'ish.

Did I just get semi-unlucky in the CPU lottery? Any thoughts/ideas?

[HiVizMan]

I would suggest that your CPU cooler is not mounted optimally. That is where I would start. Do a remount and see how that helps. How have you mounted your fans?

If possible show the direction of the air through the radiator.

[TechRaven]

Quick/Dirty MSPaint of the case/airflow:



The Raid5 is doing its initialization and is at 75% currently so I won't be able to remount the CPU for probably another 12 hours. But I'm pretty sure the remount I did was pretty decent, the temperatures drop pretty quickly from 82c to 40c when load is removed which would lead me to believe (incorrectly or correctly) that the thermal contact is decent.

Is the AS5 an ok thermal paste? I've seen lots of contradicting opinions on the mater, unsure if I should order a different paste.

[kkn]

i would say that termal paste is overall good, its when your starting to benck hard whit high OC you will start to notice the difrence and the 1degree diffrence counts

[TechRaven]

Going by http://www.madshrimps.be/articles/article/1000393/Corsair-Hydro-90-H110-Review-140mm-Cooling-Power/7 I may try reversing the installation, but I'm starting to regret the choice of the H90. The HardOCP review I showed didn't show nearly as much difference from the H90 to the H110 but othersites show closer to the madshrimps review.

I went with the H90 because on corsairs site they say the H110 wont fit in the 550D but http://www.overclock.net/t/1256647/corsair-550d-owners-club/380 says it will. And looking at the case and how the H90 is mounted I think It would. So I might have to order the H110 as well and just throw the H90 in garbage or something cause I don't think I can return it. And I'm guessing no ones going to buy a used watercooler on ebay.

[HiVizMan]

Folks will buy it I can assure you.

Do not reverse your air flow it is correct that way.

I would just make sure that your water pump is indeed on (simple touch will tell you if it is) and let things settle down. It just might be you have a hot CPU.

Until your RAID array is finished building there is no point to mess with stuff.

[Zka17]

For those voltages (Vcore at 1.288V) the temps seems pretty high...

With my H100 I was able to keep core temps below 80C on the 3930k at 5GHz with 1.5V... - used four Cougars in push-pull... I would expect the H90 to be a little bit worse, but not so much...

First, what is your ambient temp? Second, pay attention only to the CPU Core temps in HWMonitor - it's known that CPUTIN gives wrong readings... Also, have only one temp assessing application installed at a time, more of those likely will give you misreadings...

When applying TIM, make sure that you apply only the very minimal amount... usually people says to use the pea-size drop on the center method, but if you want to spread it, still stick to that pea-size amount... (too much TIM actually will increase the temps)

[TechRaven]

Zka17 wrote:
For those voltages (Vcore at 1.288V) the temps seems pretty high...

With my H100 I was able to keep core temps below 80C on the 3930k at 5GHz with 1.5V... - used four Cougars in push-pull... I would expect the H90 to be a little bit worse, but not so much...

First, what is your ambient temp? Second, pay attention only to the CPU Core temps in HWMonitor - it's known that CPUTIN gives wrong readings... Also, have only one temp assessing application installed at a time, more of those likely will give you misreadings...

When applying TIM, make sure that you apply only the very minimal amount... usually people says to use the pea-size drop on the center method, but if you want to spread it, still stick to that pea-size amount... (too much TIM actually will increase the temps)

I used a pretty small amount, just enough to get a thin coat of it, I realize a big pad you'll make it worse. Always hard to know for 100% without taking it off, but then you have to redo it so 🙂

I found the VSCCA Load Line Cal setting, it was on 'auto' and seemed to be going to 'extreme' for some reason, I forced it to regular and it dropped my temperatures a tiny bit.

I'm reading the CPU Core temps not the TIN. And I usually only have one app open at a time, was just looking at the different ones.. HWMonitor is always +10C compared to HWInfo/RealTempGT/CoreInfo/etc, the fact that it says the temp is at 94C vs (84C) which should mean throttling is on and throttling isn't on I am going with the assumption HWMonitor is not correct.

MB Ambient's like 32-35C, I think I may have just gotten a relatively hot CPU and 4.2 is going to be my wall for being stable (better then nothing I guess, I got super lucky with my 920, can't win them all though). If I put the V Core anywhere over 1.3 it heats up to 80C extremely quick. Or maybe my H90 is defective, it is pretty damn loud (I don't have a decibel meter unfortunately but relative to the Windowed AC right next to me its still quite obvious, which I wouldn't think is right from the reviews i've seen)

[Arne_Saknussemm]

Hmmm I am wondering about your cooler too. That voltage and clock should not be giving you those temps. I suspect either the contact is bad or the cooler is bad. You will have to examine the contact carefully..is the pump making a lot of noise like there is air in the loop or anything like this?

Core temps can vary by 10C easily..that is no big deal but 80-90C for 1.28 is high

Also installing so many monitoring programs may give you problem readings....I'd un-install a few

[Zka17]

Those temps are insanely high!

Does the H90's pump working at all? Go in BIOS and check the CPU temp there... if that is higher than 10-15C then your ambient/room temp, then your cooler isn't working properly...

When I'm talking ambient temp, it's about the room temp where you have your system - if that is 32-35C, then I wouldn't play with OC'ing...

Also, when I say to have only one temp assessing application, it's meaning one application installed - some of those applications can run in the background, still polling the sensors...