CPU Temperature running a bit high

[Timbercottage]

Hi all,

About 6 months ago I bought an i7 930 and stuck a new Corsair H50 watercooler on it. When new the idle temperatures were running around the 20 - 25C region and I was able to OC the CPU to 3.8GHz quite comfortably with the temperatures never going much above 70C. Now however, (back on stock speeds) the idle temp is close to 50C! Something is clearly not quite right here, but as this is the first time I've used watercooling I'm not really sure what I can do to try and bring the temperatures back down.

Any suggestions?

 

[Gorman]

Idle temps are wildly inaccurate by upto 20c at any given point, they only get accurate when they get to the upper limit of the sensor (100c), its just how they work.

So adding in the obvious 5-10c ambient increase that spring has given us i would like to know the temps under load.

My I7 920 will only idle at about 30c during the coldest of winter months and i have a significant water cooling solution on there.

 

[Timbercottage]

Ah ok, I'll go run a test on Prime and see what that gives me and post the results up here.

 

[Gorman]

excellent, its the best way to get a good idea of how well its doing

 

[Timbercottage]

A quick 15 minute burst of Prime with the CPU running at 100% load got me temperatures hovering around the 80C mark, max was 82C.

 

[abdulhamid]

Are you able to check what speed your H50 pump is operating at? I use either speedfan or HWmonitor to do this.

The reason I ask is I have the ECO ALC (similar thing to the H50) and when I first received the machine the pump was connected to a Chassis fan header with Q-Fan enabled in Bios. This meant that the pump was not working at the recommended 100% rpm. Try disabling all Q-fan in bios and see what temps and pump speed you get.

 

[pengipete]

Running an i7-930 at 185X21 (similar overclock to your own) with a Titan Fenrir - air-cooled but gives very similar cooling to an H50 - I will get ~75-80C unless the heating is turned up high or the sun is shining directly onto the black, metal case when it can go up into the mid 80's

Idle temps are almost meaningless. The best way to check them is not when first switching on - it's after ending a Prime95 test and waiting at loeast 15 minutes to see where they stabilise. As Gorman says, the "i" series have a sort of sliding scale of accuracy - they're very accurate at high temperatures become less accurate the cooler they get but reported idle temps for 920-930's are anywhere from high 20's to low 50's.

If temperatures start to increase and you are absolutely certain that there's no external cause - ambient temperatures are the obvious one but even changes in air-pressure will affect cooling (don't forget that "water cooling" on an H50 is something of a misnomer - the liquid merely moves the heat to a radiator where it can be air-cooled) - then you should check that the fans, vents and radiator aren't clogged with dust and fluff.

Ultimately, there are no genuinely stable temperatures when overclocking - apart from anything else, Prime95 doesn't really run the cores (and HT virtual cores) at 100% so there will always be small variations in the maximum recorded temps. If you get to the stage where it worries you - you should consider lowering the overclock if you know how to do it yourself.

I ran mine at 3.9GHz for a while - stable but needed almost 1.3V and was getting into the 80's in Prime95 (in winter) so I decided to experment to see how high I could overclock whilst also "undervolting" - not the most accurate description as there isn't really a definitive stock voltage for these i7's. I'm now running at 21 x 175 on less than 1.2V with the 1600 RAM moved up to 1845 (it wouldn't overclock with bclck at 185) - the PC feels much nippier (benchmarks support that) and temps never go above 70C. In fact, my CPU fan has yet to run at full speed since changing the overclock so it's a "better" overclock in every respect.

 

[Timbercottage]

Thanks for the advice guys!

@Abdulhamid, the CPU fan is running at 1400rpm which I'm fairly sure is the recommended speed for the h50

Just out of curiosity, what does everyone say is the max 'safe' limit for the cpu go to, I've read people say anywhere between 70 up to 90 is the maximum you should really be seeing. Just wondered what you guys opinions are?

 

[Gorman]

Constant 90's is fine as far as the integrity of the chip goes but you really want to be seeing 80's under load. The cut off point is 112C i believe.

 

[DeadEyeDuk]

The cut off point is 112C i believe.

What...what...what happens then? :helpsmilie:

 

[Gorman]

The machine turns off to prevent damage.

 

[pengipete]

The machine turns off to prevent damage.

...and before that happens the CPU throttles back - effectively slowing down until the temperature drops to safer levels. This can have an effect where adidng that bit of extra voltage and speed actually slows the PC down during demanding tasks as it runs for a lot of the time at much slower speeds than it would with a lower overclock or voltage.

One thing to bear in mind when looking at discussion of "safe" maximum temperatures (apart from the obvious fact that there's a lot of garbage on the internet posted by people who lose nothing if you damage your hardware) is that there are a number of different types of measurement and they often get crossed - people see one figure and incorrectly apply it to their own measurements.

Basically, Intel place a sensor inside the CPU package - the whole "box" that contains the processors, circuits, heatsink etc - and that's what they use when refering to temperatures. It's as near as damn it at the very top of the package - generally called the Case Temperature (as in CPU case rather than PC case), in the centre of the square heatsink which means that it's quite some distance from the hottest part of the CPU - the processor cores (that's right inside the actual processors themselves) - close in the sense that everything CPU related is measured in millionths of a millimetre or less. Most decent temperature monitoring applications - the ones favoured by overclockers and enthusiasts - actually report the reading from within the processor cores - the hottest part of the CPU. The difference between the case temperature and the core temperatures is not actually known (and probably varies quite a bit from chip to chip due to minor variations in materials and manufacturing) but is commonly held to be around 5C.

What this means is that Intel may specify a maxiumum temperature of - for example - 70C but overclockers will regard 75C as safe because they're monitoring cores not cases. That's fine as long as everyone is using the same monitoring apps, those apps are well written and you know and trust the person who tells you that 75C is safe. All too often, some invented figure gets used in some random post and before you know it, everyone "knows" it's true and nobody ever goes back and checks it at source (not helped by the fact that most overclocking ex-spurts are far anything but).

Also, when reading about other people's overclocks and temps, you have to allow for huge amount of bulls' smelly-stuff and badly written reports. People exaggerate their result or "forget" to mention that there PC crashes ever five minutes and after a while you begin to notice that the big-overclock boys usually post their claims for a few weeks at most before disappearing - maybe because their $2,000 PC just became a doorstop.

With i7's, there's also a very specific and hugely important factor where temperatures are concerned when overclocking or "optimising" BIOS for performance - and that's Hyper Threading. HT requires a lot of extra voltage - above and beyond what's needed to run the "real" CPU cores and that means a lot more heat. When you read of people hitting 4GHz at 70C or hitting 4.5GHz at 1.3V, chances are they've disabled Hyper Threading. Not specifying that when reporting their overclocks is like boasting about a car's top speed without mentioning that you were driving down a very steep hill at the time. There's also a highly debatable "trick" involving disabling the voltage spike protection in the motherboard's settings that gives a higher actual voltage at any given setting - again, that should be reported if you want to make fair comparisons.

 

[pengipete]

What...what...what happens then? :helpsmilie:

Have you never seen this video? It's a bit old now - not exactly the latest CPU's - but it's still fascinating to see how Intel CPU's protect themselves but AMD ones just burn out. There's one floating around where the AMD chip literally bursts into flames - I couldn't find it on a quick search so you'll have to settle for a bit of smoke.

Apart from giggling as the AMD chips start to fry, watch the game graphics in the background - note how they slow down as the Intel chips throttle back - but it keeps running.

[video=youtube;zgwu6X-QZ7U]http://www.youtube.com/watch?v=zgwu6X-QZ7U&feature=related[/video]

 

[DeadEyeDuk]

I like to learn, thanks pete!

 

[Timbercottage]

Wow, awesome post. I guess I'll be taking most OC advice with a massive pinch of salt in the future.

I've definitely taken away from this that for the majority of the time I'm going to be fine as far as CPU temps are concerned due to the amount of safeguards in place to prevent any lasting damage. This is good to hear! I'm going to be taking my PC apart in the next few days and give it a decent Spring clean, the amount of dust, hair, fluff and god knows what else in there is disgraceful!

Thanks again for the great advice

 

[pengipete]

Abdul swapped the fan on his Coolit - similar set-up to the H50 - and knocked around 3C of his temps (check post 15 for side by side comparisons). It's not a huge improvement but worth considering for such a relatively small outlay. If you want to check that out, go for a fan with high static pressure rather than high CFMs or check around the net and post back if you want any advice before spending. Just bear in mind that more powerful fans generally means more noise though if the existing fan isn't top-notch, you could get both - better cooling and lower noise.

Just for info - the reason for going for static pressure over CFM is that you need to be able to force air through the densly packed fins on the radiator. High CFM fans shift a lot of air but they don't give it oomph to fight past the fins - they're great for cases and fine for CPU air coolers - just no much cop for radiators. It's a bit like comparing a Porsche with a Land Rover - the Porsche is faster on the flat but the Land Rover will go up a steep hill without having to raise the revs to melting point.