Managing 2600K temperature (above 5GHz overclock)

[velis]

I'm in a bit of a dilemma here: my 2600K will overclock great - I can go as high as 5.5 (though the voltage required raises above 1.5 there). Currently running 5.0@1.4V. However, here's the issue:
The same moment I run Prime, CPU spikes from 30 to 70 - 80 degrees celsius. My water cooling then keeps it pretty much at those temps (never goes an additional 10 degrees higher).
However, these temps are a bit too much for the proc and it will start throttling after a while.
Of course I have tried lowering voltages, but it seems the automatic ones my MB assigns are already the optimal ones. Reducing them one step will just give me a BSOD shortly after starting Prime torture.

So is there any particular thing I can do that will make my proc not heat up as much under full load?
Also: are 2600Ks typically like this? Did I just get a bad one which does overclock, but has temps to show for it --> would a better overclocker just overclock at lower voltages?

Using ASUS P8Z77M- PRO + 2600K + 2x4GB Some Patriot 1600 memory.
BTW: this board rocks. The first board I have after 486 that does POST in one go without a gazillion option ROMs POSTing on their separate screens with plenty of flickering to spice it up. Overclocking a breeze as well and I haven't even tried the TPU switch yet

 

[Puppies04]

Your temps are pretty much normal for what you are doing, I can tell you that there is nothing wrong with your chip in fact a lot of chips won't get anywhere near 5ghz on 1.4v. If you want lower temps then lower your overclock that is your answer plain and simple.

I know you have probably been selectivly reading posts from people who are running 5.2ghz on their chips but you need to accept the reality that this is the upper limit on high end cooling. If you take a look around these forums you will see most people settle for a 4.4-4.8ghz overclock because they aren't comfortable putting any more voltage or heat through their chips.

I'm not trying to burst your bubble here but you need to accept that is you are going to run at 5ghz you have to accept the temps (and possibly the degradation) that comes with it. You might lose a few degrees by investing it some higher quality TIM/ tightening your CPU block down a little more/ running a higher flow pump but 60-70C @5ghz under stress test conditions isn't going to happen unless you want to start looking at below ambient cooling.

 

[T_Yamamoto]

5ghz is the upper limit for air cooling

 

[BURNINGchicken3]

most 2600ks wont go over 5.7 without L2D or DICE or have very high temps. only thing you can do is lower your overclock or get more rads/fans.

 

[Puppies04]

5ghz is the upper limit for air cooling

*On a good chip/decent mobo/decent PSU*

 

[borisvodofsky]

OK, seeeeeee. THIS IS where you DISABLE thermal throttling. EEEEYYYEAA......................

Because 80C is nothing. This processor can take 80C for a good 2 years, no problem.

 

[guskline]

I'm in a bit of a dilemma here: my 2600K will overclock great - I can go as high as 5.5 (though the voltage required raises above 1.5 there). Currently running 5.0@1.4V. However, here's the issue:
The same moment I run Prime, CPU spikes from 30 to 70 - 80 degrees celsius. My water cooling then keeps it pretty much at those temps (never goes an additional 10 degrees higher).
However, these temps are a bit too much for the proc and it will start throttling after a while.
Of course I have tried lowering voltages, but it seems the automatic ones my MB assigns are already the optimal ones. Reducing them one step will just give me a BSOD shortly after starting Prime torture.

So is there any particular thing I can do that will make my proc not heat up as much under full load?
Also: are 2600Ks typically like this? Did I just get a bad one which does overclock, but has temps to show for it --> would a better overclocker just overclock at lower voltages?

Using ASUS P8Z77M- PRO + 2600K + 2x4GB Some Patriot 1600 memory.
BTW: this board rocks. The first board I have after 486 that does POST in one go without a gazillion option ROMs POSTing on their separate screens with plenty of flickering to spice it up. Overclocking a breeze as well and I haven't even tried the TPU switch yet

Do you want a cpu OCing too high just for bragging rights or do you want one that will run Prime95 24/7/365 without errors?

BTW try IntelBurnTest at 5 to see if you have errors. These safety features on the CPU are there for a purpose. Also, even though you are water cooling, is your radiator fan running high enough to dissipate all the heat ? You may need a second fan or higher speed fans on the radiator.

I would lower the 2600k to 4.8 and see where you are. Personally, I found a safe sweet spot on my 2500ks at 4532 (44 x 103). No matter what diagnostic test that OC speed is safe and solid AND a big OC from stock. Your silicon might be better but running at 5 is near the limit, even for the 2600k.

 

[Ajay]

What is your LC setup and what are ambient temps in your computer room?

 

[velis]

Thanks everyone for the replies.
I'm just looking at how high this chip can go. I'm not planning on staying this high, but I also noticed that even @5GHz, there are very few tasks that actually heat up my CPU, so I feel pretty safe up there. When fewer cores are loaded, the temps don't reach nearly as high...
As I said, my WC setup will keep the temps in check. If the CPU starts at 70, it will only go as high as 77 - 78 after whole night of prime. Using a custom EKWB config (2x 240mm rads - only 3 fans) with Silverstone GS05 case (need to do the blog post on assembly for a bit of showing off )

And just to clear this up: I don't have cooling issues as such. The question was about how I could lower the jump in temps that happens immediately when I load the cores. I'm talking 1-2 seconds when my chip will go from 30 to 70. (30 to 55 when left @stock) I guess power saving really works with these chips.

 

[BD231]

You could slow the rise down with a bigger water block, but being that you can do 5ghz @ only 80c max load I wouldn't bother, that's an awesome temp for 5ghz.

 

[Ferzerp]

2600k is not going to throttle at 80C. That's 20 from Tjmax where it would throttle.

 

[tweakboy]

Must sound like a Jet in your room. Whats up with the H20 ,, make sure pump is moving the water properly and setup was proper. @ 1.40v , your H20 can handle that easy get maybe 60's Celsius at the most...on full load. Reseat the waterblock possibly.. add more coolant or water ,,,, Anyhow @ 1.40v that is nice @ 5Ghz thats all you need, you go higher and heat will slowly kill the components.... gl

 

[Puppies04]

Must sound like a Jet in your room. Whats up with the H20 ,, make sure pump is moving the water properly and setup was proper. @ 1.40v , your H20 can handle that easy get maybe 60's Celsius at the most...on full load. Reseat the waterblock possibly.. add more coolant or water ,,,, Anyhow @ 1.40v that is nice @ 5Ghz thats all you need, you go higher and heat will slowly kill the components.... gl

Care to back this up with some facts? Show me a 2600K running 5ghz circa 1.4V hitting 60C MAX on ambient water cooling after prolonged stress testing. I won't hold my breath....

 

[BonzaiDuck]

I'm in a bit of a dilemma here: my 2600K will overclock great - I can go as high as 5.5 (though the voltage required raises above 1.5 there). Currently running 5.0@1.4V. However, here's the issue:
The same moment I run Prime, CPU spikes from 30 to 70 - 80 degrees celsius. My water cooling then keeps it pretty much at those temps (never goes an additional 10 degrees higher).
However, these temps are a bit too much for the proc and it will start throttling after a while.
Of course I have tried lowering voltages, but it seems the automatic ones my MB assigns are already the optimal ones. Reducing them one step will just give me a BSOD shortly after starting Prime torture.

So is there any particular thing I can do that will make my proc not heat up as much under full load?
Also: are 2600Ks typically like this? Did I just get a bad one which does overclock, but has temps to show for it --> would a better overclocker just overclock at lower voltages?

Using ASUS P8Z77M- PRO + 2600K + 2x4GB Some Patriot 1600 memory.
BTW: this board rocks. The first board I have after 486 that does POST in one go without a gazillion option ROMs POSTing on their separate screens with plenty of flickering to spice it up. Overclocking a breeze as well and I haven't even tried the TPU switch yet

Scanned through the posts here while I wait for the squirrels to get careless in my garden so I can nail the little b*****ds.

I'm not going to fault you for "trying" to OC to 5+ Ghz, but the stats on the Sandy were pretty well determined a year ago. Depending on your self-imposed voltage and heat limits, some percentage of the 2600K's will go as high as 4.8, some percentage will get you to 4.9 or 5. You may be able to push further, but what's the point?

I can tell you that you risk the greater damage from excess voltage. I can also tell you that -- assuming room-ambients in a range around 75F and with air-cooling (good . . . air-cooling) -- a carefully done OC to about 4.65 or even 4.7 won't push the TCASE temperature much beyond 70F (and it might even be lower than that). The sensor temperatures, on the other hand, are likely to show values around 80C or higher.

Water-cooling without "chilled" water may buy you 10 or 20C on the temperature -- I haven't looked at the data on that for some time.

Every time I go through this cycle -- buy, build and overclock -- I may spend several weeks pushing it to within parameters which I myself set. It may be safe to run the "turbo-OC" voltage up above 1.4V, but I chose for Sandy Bridge to keep it below 1.35. The fact that load voltage will be below "peak" even with LLC settings, or that "turbo" means it will not be at those values (Ghz and Volts) all the time -- may make it even safer.

Once I get it "perfect" -- then we're talking about daily use, and I won't go back and diddle with it more for the trouble I might cause myself. Or -- for the disruption in dependable usage.

And I feel skeptical about "how much better" the system will perform at 200 or 300 Mhz higher (say 4.9 or 5.0 instead of 4.6+).

 

[borisvodofsky]

Care to back this up with some facts? Show me a 2600K running 5ghz circa 1.4V hitting 60C MAX on ambient water cooling after prolonged stress testing. I won't hold my breath....

I wondering, is this not possible for some reason other than Cost?

I mean, I ask because I've seem those triple x triple 120 radiators with 9 fans on both sides. Surely that can get 60C

 

[BonzaiDuck]

I wondering, is this not possible for some reason other than Cost?

I mean, I ask because I've seem those triple x triple 120 radiators with 9 fans on both sides. Surely that can get 60C

Rule of Thumb: Ask AigoMorla -- if he is still among our brethren (and sister-en) here. He has practically been crowned water-cooling guru.

Over the years on this forum and "Cases and Cooling," we've discussed some pretty exotic cooling rigs -- Rube-Goldberg as they were.

There are designs for evaporative cooling towers using 4"-dia. PVC pipe. There are designs using RubberMaid tubs filled with plastic drinking-water bottles frozen solid. There are designs utilizing a commercial water-cooler/water-chiller. There are designs using TEC plates to cool a reservoir. And of course, one could order a phase-change cooling rig from "Frozen CPU:"

http://www.frozencpu.com/cat/l1/g49/Phase_Change.html?id=HBzWvUjU

One guy even got a heavy-duty waterpump to throw in the mix so he could run water through a copper pipe under his lawn. I think he lived in some place like Arizona. All he had to do was turn on the lawn sprinkler system -- night or day -- and the evaporation would chill his water-coolant that went through his CPU waterblock. But not easy to move his computer -- even if only to change its location in the same room.

I get more satisfaction out of implementing the KISS principle and making an "optimized trade-off" between dollars, speed and complexity.

But it doesn't much matter if you use liquid nitrogen while bumping up the voltage close to 1.5V. Just freezing the CPU and keeping the temperatures at 10C will not prevent electromigration due to excess voltage.