Need help with cooling for OC'ing

[ziggy1971]

Hello all,
Thus far I've OC'ed my i7 2600K to 4.9GHz @ 1.440v. I've run Prime95 with the standard tests over 2 hours and the temps (using RealTemp 3.67) maxed out at 83C. When I cranked up the Vcore to 1.500 and the mutilplier to 50, the temp maxed out at 92C within the first 6 minutes (no BSOD or shutdown) I just stopped it because I'm not sure just how far I can push the temps on this CPU.

I'm using a Corsair H100 in an Antec 1200 with all fans set to low (except the rear 2 fans, where the radiator is mounted, which are set to high) and are pulling air from the inside to outside.

With the exception of the CPU temp, I'm very impressed with how cool the total system runs. (Compared to my old dual-core AMD system where the case gets warm while idling using the stock HSF system).

So, my question is just how far can I (safely) push the temps on this Core i7 2600K? I've followed the OC'ing guide as described here http://hardforum.com/showthread.php?t=1578110 for the most part; I'm using the X.M.P. of the RAM, BCLK @ 100.0, Turbo and Hyperthreading enabled, etc. Most settings are as described in the article or left at their defaults.

Are there any other specific "tweaks" I can use for OC'ing? or to reduce the temps of the CPU?

I'm more or less just curious what this CPU can do. I don't plan on using anything beyond 4.8GHz for everyday usage as that is already a huge improvement over stock values and quite impressive.

Please note, I do have 3 more 120mm fans that I could install into the case to help with cooling other parts of the motherboard, if necessary.

Thanx in advance,
Cheers,
ziggy1971

I thought my other PC details would show up in a sig, but I guess not, so please see my profile for full details on my rig.

 

[Twiggz]

83C is too hot. I would keep it around no higher than maxing at low 70's while under full load for 24/7 and that is still the high side IMO.

 

[Automata]

Why are you upping the voltage from 1.44 to 1.5v? Voltage increases heat load exponentially. So, if you can run a lower voltage at 5.0 GHz, you would be much better off.

 

[ziggy1971]

After doing some more research, I found the following.

http://www.anandtech.com/show/5054/corsair-hydro-series-h60-h80-and-h100-reviewed/4

Real Temp version 3.67 is used to monitor and log temperatures with samples taken once every second. We run each test for 15 minutes, take the average temperature of all four cores, and use a three minute rolling average to calculate the results. The final temperature is the average of the last three minutes from the test. We then take the final calculated temperature and subtract the room’s average ambient temperature to get our delta temperature. This method has the least temperature fluctuations and is the most consistent.

http://www.anandtech.com/show/5054/corsair-hydro-series-h60-h80-and-h100-reviewed/5

Given that the temps are listed as a Final Result (ie, Result = Real Temp 3.67 - the room’s average ambient temperature).

Let's do some math:
Suppose you had an ambient room temperature of 20°C. (Delta Temperature)
The calculated from Real Temp 3.67 (as listed). Let's use the the highest value, the H80 (Low) @ 61.2°C)
By adding 61.2°C + 20°C = 81.2°C

Back to the article @ http://hardforum.com/showthread.php?t=1578110

Cooling recommendation and test performed with CoolerMaster Hyper 212+ with Single Fan, this is the minimum recommendation for multis above 46x. For 50+ multis we recommend a dual fan configuration with this cooler or improved cooling.

A quote from unclewebb @ http://www.overclockers.com/forums/showthread.php?p=6765342#post6765342

When the core temperature reaches TJMax, the Intel docs say, "the processor has reached its maximum safe operating temperature." That's hot but it's not "very dangerous" and it doesn't start to thermal throttle until it hits TJMax.

TJMax can vary from one CPU to the next. Many Sandy Bridge CPUs seem to be set to 98C. Intel typically sets the THERMTRIP thermal shutdown temperature to TJMax + 25C. When significantly overclocking, it's not unusual for a CPU to reboot when it starts running too hot but that usually happens way below the Intel set THERMTRIP temperature.

I'm just an average computer user so this is what I think.

Aside from what documents or reviews say, I think that Intel's engineers would be smart enough to impliment some simple features that prevent damage to the CPU's they produce and then provide a 3 year warranty on top of that. With a lot of people pushing their CPU's to the very extreme (including liquid nitrogen), just think of all the RMA's Intel would have to deal with if they didn't provide a high quality product.

So, I guess what I'm trying to say is, based on what I've found on the net, 83°C isn't "extreme" at all. In fact, 92°C still seems doable seeing that the TJMax is set at 98°C and the CPU has a THERMTRIP sensor (among others). Although I won't be pushing that 24/7, (I don't know of anyone who could do that in the "consumer" market), it's still nice to know that it's there and available should the need arise, that's all.

I hope that this is clear and accurate so that it may help others as well; if not please feel free to add/modify this post with more accurate or conclusive information.

Thanx again,
ziggy1971

 

[Twiggz]

After doing some more research, I found the following.

http://www.anandtech.com/show/5054/c...100-reviewed/4


Quote:
Real Temp version 3.67 is used to monitor and log temperatures with samples taken once every second. We run each test for 15 minutes, take the average temperature of all four cores, and use a three minute rolling average to calculate the results. The final temperature is the average of the last three minutes from the test. We then take the final calculated temperature and subtract the room’s average ambient temperature to get our delta temperature. This method has the least temperature fluctuations and is the most consistent.

http://www.anandtech.com/show/5054/c...100-reviewed/5

Given that the temps are listed as a Final Result (ie, Result = Real Temp 3.67 - the room’s average ambient temperature).

Let's do some math:
Suppose you had an ambient room temperature of 20°C. (Delta Temperature)
The calculated from Real Temp 3.67 (as listed). Let's use the the highest value, the H80 (Low) @ 61.2°C)
By adding 61.2°C + 20°C = 81.2°C

You're using the H100 not the H80(you choose the H80 value from that link) which puts the temps at a max 74.5C if you're using the worst value on the graph and 68.6C if you are using the best value (both of those values being figured up with the 20C ambient you choose). That is at 4.8ghz, your OC is at 4.9 and is hitting 83C, that doesn't seem very good too me. So standing by those figures you found for us, i'd say what I recommended to you is completely accurate. If you want to use higher temps and be okay with that, then that's up to you of course.

 

[ziggy1971]

I was using the information I found as reference material for the actual temperatures and capabilities of the CPU that others have tested.

I chose the H80 solely based on the highest temperature listed in the chart, nothing more (if they had the stock Intel cooler result I would've used that as the reference temperature.)

I wasn't comparing one cooling system to another (ie. my H100 to the H80 listed), only the temps of the system being tested on that site.

The H100 results are helpful for me though because I have something that I can compare my own results to, instead of having to estimate what I should be getting.

Without drawing this out too much further unnecessarily; I can also draw the following conclusion:
If the H80 can hold a temp of 61.2°C + 20°C = 81.2°C running 4.8GHz @ 1.4v
The H100 should be able to hold the same temp with slightly higher voltage and GHz. right?

Anyway, I was hoping this would be informative and helpful, not an argument, so I'm bowing out of the conversation.

Cheers,
ziggy1971