Crosshair V / Phenom II 1100T - temps getting out of control

[skubany]

Hi,

Just got myself a Crosshair V, Phenom II 1100T cooled by Thermalright SI-128 SE with fan @ 2400 rpm. Goal is to reach 4Ghz, but for starters I've set 3,6Ghz, just to have it higher than my old Phenom II 940 Deneb @ 3,5Ghz.

I've got some issues and I'm wondering why they occur.

To stress the CPU I'm using SP2004 x 6, 4/6 have option Blend - stress ram and cpu, 2/6 have option Large - test some ram, because I don't have enough memory for all of the SP2004 instances to run with Blend.

Measurements taken with HWMonitor64 and Asus AI Suite II's EPU status for CPU wattage monitoring.
I have a Kill-A-Watt (KaW) monitoring wattage of all my computers.

OC'ed via BIOS by changing the multiplier only, from Auto to 18. Set Digi+ VRM to Regular/100%/Auto/Optimized/T.Probe - basically the defaults.

Stock results:
idle - 22C cores, don't remember the socket temp
load (10min) - 48C socket, 41C cores
wattage:
idle - 580W (KaW), 35W (EPU)
start of the testing (about 1 min after starting the test) - 660W (KaW), 76W (EPU)
10min into testing - 662W (KaW), 76W
20min - same as 10min, even after 20min the values at 10min hold without change.

OC'ed at 3,6Ghz:
idle - 37C socket, 27C cores
load (10min [socket/cores in C]) - 55/52, (20min) - 64/66, (23min, manual stop) - 68/71
wattage:
idle - 580W (KaW), 40W (EPU)
start - 678W (KaW), 87.91W (EPU)
10min - 694W (KaW), 98-99.5W (EPU)
20min - 708W (KaW), 99.99W [max that can be displayed apparently] (EPU)
23min - 715W (KaW), 99.99W (EPU)

Loading the CPU pretty much reaches and then holds 55 socket and 52 core temps for a few minutes. Once these temperatures are exceeded then it's a gradual rise to a point where I'm forced to stop the test. Above 20min the temps increase at an even more rapid pace. It seems like the MB is feeding the CPU more and more and more juice, but should it? Is this normal or abnormal?

Could a 300Mhz increase lead to the observations I'm making? My HS is perfectly capable of keeping a 3,3Ghz at 48/41 for over 20minutes, but it can't keep a 3,6Ghz CPU below rated max temperatures after 20min period?

Is the MB regulation circuitry bad? Why would MB feed more power to the CPU over time when it held very wall over couple minutes at 55/52 temps.

The case is ventilated well. Could it be that too much AC5 was applied?

[Detection]

Voltage is the biggest factor in heat - keeping the vcore as low as possible while still stable keep the heat lowest

I think those temps are too high - I have my Phenom II 965 @ almost 4.1Ghz with a vcore of 1.48v - My cooler is the Tuniq Tower 120 Extreme and thermal paste is Arctic Silver 5

Idle around 30c

Load around 48c

You only need a very thin layer of thermal paste - all it is for is to fill the impurities in the metal surfaces - not to be a sandwich spread between them - re-apply your paste using a credit card to spread it perfectly over the CPU as thin as possible edge to edge

Clean the heatsink and CPU with dry tissue first and get them shining clean

[skubany]

What do you think about those runaway temps?

That is the biggest concern for me. It seems to hold the temps and then once 55/52 is breached it goes up forever. This seems very unusual.

If the stock can hold the temperature why can't OC'ed do the same. When the temps reach 55/52 I'm observing the EPU CPU Wattage value and if it held about 88-95Watt then there wouldn't be a problem. And it does hold that for a few min, but then it breaks down and the Wattage increases and increases until the CPU hits max temp.

What makes the MB think that CPU needs more watts later in the testing?

vCore actually decreases as the CPU gets hotter. Starts at up to 1.5 on idle and then goes down to 1.34 at load.

I'm using BIOS 0903 but will flash to the newer one, maybe today even.
Also I'm probably going to re-seat the heatsink. I am aware of proper cleaning/preparation for heatsink placement. I'm a fan of the dot method for AC5 application. I've actually run a bunch of tests yesterday with various AC5 amounts, but on a different system of mine.

[skubany]

MarshallR@ASUS
I'll do that and post back.

"elevated voltages for OCing"
But according to my monitoring software the voltage (vCore) actually drops as the test progresses. I am thinking of setting the voltage manually and see if it makes any difference. I'm basically hoping that the auto settings on the MB make it behave so.

I have an external temp probe attached to the top of the HS base, its reading does increase, but not as much as the cpu temps. It changes maybe 10-15C max.

I'll also try a 100Mhz increase to see how the MB/temps behave.

[skubany]

Tried 3,4Ghz, seems to stay at 44C core.

Back to 3,6Ghz, voltages:
CPU voltage decreases from 1.4xx to mostly 1.356 and then 1.344 when cores are at 68C and higher.
CPU/NB fluctuates back and forth between 1.151 to 1.164. 1.164 held longer when cores are 68C and higher. 1.144 at idle.

And everything else is pretty much unchanged. DRAM occasionally flips from 1.521 to 1.528, briefly. My RAM is set to 1T and otherwise to it's stock timings, which were incorrectly set by default by this MB. MB temp stays at 31C, there is a large fan blowing on it.


I'll reseat the HSF now, attach external probes to HSF base, cpu side MB heatsinks and ram heatspreader.

And unfortunately, I can't check this CPU anywhere else.

[X-ROG]

As your CPU gets hotter its transistors leak more. Basically what you're seeing is prolonged pressure at elevated voltages for OCing causes rampant leaking (Auto setting will usually raise the voltage if you OC iirc).

Can you check the CPU in other boards?
Monitor the voltages as it stays under load - do they progressively increase?
Check if your heatsink is getting hotter as it says the temp is increasing. If the heatsink isn't the same temp as the CPU it's not doing its job effectively (as the system should move to equilibrium if nothing else changes).

[Detection]

You got your voltages set on auto ?

[skubany]

Detection:
Yes, voltage is on auto for now. I just wanted to get a quick 3,6Ghz before I really get into OCing this thing.

[skubany]

Unfortunately, still can't hold 54/53 and again, once this breaks down it's just time till it all explodes. Got to 66/71 in 22min. HS probe this time was 51C, so much higher than the ~45C before lapping.

Maybe the heat transfer to HS is still hindered.

And I had such high hopes.

It's interesting that the max temp is reached always about the same time, around 20min mark.

My next step is to try and lower the heatsink mounting assembly so that I can get a tighter fit. But I'm worried that maybe my CPU or MB is just not very OC friendly.

[DaemonCantor]

Maybe you could set an Air conditioner inside your case?...Just thinking out loud again...try a better Heatsink like http://www.xoxide.com/gelid-solutionsgamerseriesgx7-cpucooler.html or going with http://www.xoxide.com/antec-k-hler-h2o920-liquidcooling.html...Just my humble opinion...