Cooling questions

[steve wilson]

Hi,
I have a case were the PSU goes at the bottom. It has a vent next to where the PSU goes. Should I position the PSU to blow out of the vent, or into the case? Here's a pic for clarity.
https://imageshack.com/i/kmQLExsGj

My CPU is running very hot at the moment. I took my heatsink off to help a friend test his CPU to see if it was his CPU or MOBO that was broken and I have reseated it. It is reaching 70 degrees C under load. One of the screws is not quite tightening all the way, so it could be that, or maybe I scewed up the thermal paste. If I have to buy a new Heatsink which one would you guys recommend? I don't want to go for a water cooler, so air only please.

Also is a push/pull configuration worth it, or should I stick to one fan? My CPU is 2500k OC to 4.3Ghz @ 1.4v.

 

[MagnusTheBrewer]

PSU fan towards the outside of the case.
Every time you remove the heatsink, you should clean and reapply thermal compound.
Most case/fan gurus seem to feel little advantage in push/pull versus single fan but, as usual, under extreme temps and case restrictions, your mileage may vary.

 

[lynna]

Fan is very important to prevent overheat which may cause damage to your computer. I suggest maybe you should allow it to check by a computer technician so he can suggest which brand of fan will be best suited for your computer issue.

 

[MagnusTheBrewer]

Fan is very important to prevent overheat which may cause damage to your computer. I suggest maybe you should allow it to check by a computer technician so he can suggest which brand of fan will be best suited for your computer issue.

Not trying to be a jerk but, paying a computer technician to solve this kind of problem is down right silly unless, you've got money to burn.

 

[BonzaiDuck]

Not trying to be a jerk but, paying a computer technician to solve this kind of problem is down right silly unless, you've got money to burn.

I've noticed a new cohort of users growing in the last few years.

It seems the OP's case would only allow two orientations -- as do mine and the cases of many others with the bottom mounting.

You will notice the rectangular hole and its fittings: you can orient the PSU either way -- intake from case exterior, or intake from case interior. I happen to choose the latter. If I wanted to pressurize the case more, I'd flip that PSU and attend more to PSU dust removal.

70C degrees at load is good or bad depending on what kind of load and what kind of processor you have.

I'm not going to recommend some air-cooler until I know the answers to these questions. If you want a good thermal paste, try Coollaboratory Liquid Ultra or similar. Another option is INdigo Xtreme. Another good option is IC Diamond. Read up on the quirks and frustrations about using any of these formulations, before you proceed to install them. Some "metal-pads" require heating up the CPU with enough thermal wattage to melt them. You may have to specially arrange to do that while controlling the temperature to -- say -- 80C. I'd say your best bet is either CLU or ICD.

 

[steve wilson]

A computer technician is not necessary... I'm good with building, it's just I don't overclock often.

I mainly just play computer games...Counter strike GO mostly and a bit of starcraft 2.

I'm going to try reseating my heatsink again and yes I do clean and reapply thermal compound. I'm using Artic Silver 5 as I have always done for the past 15 years or so... probably time I updated to a better compound though.

Could really do with some heatsink recommendations for a 2500k please.

Bonzai, I don't really understand the bit you have written about PSU blowing into or out of the case. Please can you rephrase it?

 

[krnmastersgt]

The big fan on the PSU you're thinking of is actually an intake, your PSU won't ever dump hot air into your case but depending on the environment and the internal temps of your system the orientation can affect a PSU. I personally have my PSU intake fan mounted with it drawing air from inside the case since I have a huge amount of intake fans and only 1 exhaust so I use the PSU as a sort of exhaust fan as well. That and I was using it on the carpet for a while and it draws less dust in this way.

Most cases are designed to mount it the other way (and most people do so) having the PSU draw fresh air from outside the case as the internal temps from running a GPU won't affect the temps of the air your PSU draws in. Also why a lot of cases nowadays come with fan filters to block a lot of the dust from that particular intake.

Which is better? Well assuming you have decent airflow in your system you'll likely not see much of a difference in performance temperatures for your PSU, it depends more on your enviroment and which way you prefer to mount it for the given situation. The slightly better way is probably externally drawing in the air but make sure you clean out the filter pretty routinely as most people forget that they even have one as it sits outside your normal viewing range.

As for TIM compounds, AS5 is still pretty good but I've switched over to ICD for some time and prefer it for the most part. Slightly better temps and no real curing time for me were my primary reasons for switching. Had a run in with the Indigo Extreme stuff but between all the steps and then requiring a re-mount as I couldn't reach something with my CPU cooler mounted I just said forget it and slapped some ICD7 on and tossed the Indigo stuff (while it hadn't melted yet it didn't pull off cleanly for me). I still have another 5 applications of the stuff left and might play around with it in my next few builds/rebuilds but not in a hurry (you actually want to read the instructions for applying it).

 

[BonzaiDuck]

+1 and :thumbsup:

There's some consensus, I guess . . .

2500K? Look at the Noctua NH-U14S, and measure your case innards. That's a good match for the processor, unless you want top-end air, then look at the pros and cons for a D14 or D15.

Also, I don't understand why your i5-2500K has a voltage set to 1.40V in Turbo(?) or otherwise to achieve 4.3 Ghz. Are you using "AI" or "tuning" software that came with the motherboard? I'm pretty sure that 4.3 for that processor will have a rock-solid voltage of 1.3 or less. If you "auto-tuned" this overclock, it's typical to get those type of over-volts for clock speeds in that range.

I've got the i7-2600K, Hyperthreading enabled. I think I was still experimenting with 4.3 Ghz using the "auto" voltage features. Or maybe it was 4.2 . . . . I don't think that even stock voltages upped by a higher VID based on the speed would go as high as 1.4. Not at all.

Just a minute here . . .

From my logs and notes, you should be able to do 4.4 Ghz with settings which realize a voltage reading around 1.26V under stress load -- as with Prime95 sFFT, IBT or LinX. Some of my notes were "prelim" tests, but that value came from a page that didn't have any BSOD after four hours running . . .

 

[steve wilson]

I am a bit of a noob overclocker, so probably went too high with the voltage. I will try to get it lowered. I didn't use the Auto tuner, I did it myself following a guide.

Thanks for the advice guys.

 

[BonzaiDuck]

I am a bit of a noob overclocker, so probably went too high with the voltage. I will try to get it lowered. I didn't use the Auto tuner, I did it myself following a guide.

Thanks for the advice guys.

I'm only guessing you need to spend some time with it. Usually, you find a point where the computer will boot and run Prime95 sFFT test for 10 minutes or so without crashing, locking up or BSOD. After that, incremental voltage adjustments will get you a longer duration of stability to a point with no discernible failure for several hours.

If you're trying to find a maximum 24/7 overclock, you can probably do it with incremental voltage and speed adjustments that give you initial stability for between 30 minutes and a couple hours. After that, you'd be adjusting only the voltage, watching your temperatures (in case you want to reconsider the target), and doing longer stability tests.

Once you find a point where the stress-test will go eight hours or more, add another 4 to 6 millivolts and it should be ample. Better before that point though, to use something like LinX in 20-pass runs. It shouldn't take as long. You might want to "validate" your settings with a 50-pass duration before you add the few extra millivolts.

Others may say to run Prime95 for 24 hours, but I don't think that's necessary with the stronger tests like LinX or IBT. It would be a matter of probability: for each extra hour the test runs, the chance of a failure after that declines. Some have experienced Prime95 failure after 12 or 18 hours, but that's why its "about probability." If you add the few millivolts after confirming a longer test-run, you'll have likely squeezed off that remote chance of something happening after an even longer test.

It's not something you can do in less than an hour and expect an optimal setting for lengthy stress -- however hard or mild it may be.

 

[steve wilson]

Thanks for the info... what temps should I be keeping the CPU under? I've always tried to stay below 65 with 70 being my max.

 

[steve wilson]

My Heat sink is screwed ... well one of the screws won't tighten properly and the temps were spiking above 70 after 1 min of prime95. I've ordered a Noctua NH-U14S and then I'll try and get the voltage down as low as I can.

Thanks again for the advice/info.

 

[Dylanharvey999]

Fans blow Air. I blow air, therefore, I am a fan

 

[BonzaiDuck]

Thanks for the info... what temps should I be keeping the CPU under? I've always tried to stay below 65 with 70 being my max.

The processor has a spec called "TJ-Max." This is the temperature at which the processor throttles or shuts down to save itself from overheating. As far as I KNOW, and Haswell folks can correct me, the spec should be about 100C.

There is another spec targeted at case design and ventilation called TCASE. It is a "mythical" specification -- a temperature taken at the center of the IHS (processor's metal heat-spreader), where no sensor exists. It is generally thought to be about 10C lower than the average of the core sensors at load.

I tend to terminate my overclocking expectations when the voltage setting gives a load VCORE close to a certain level. Intel used to publish specs for "safe range" and "operational range" (which included "safe.") The last such spec was for the Nehalem i7 cores, AFIK. So -- as the lithography for newer chips have changed, one would adjust those voltage limits accordingly, but everyone's guess is as good as mine. Some people think nothing of overvolting their processors to load levels above 1.40+V. This is a personal choice -- You accept the risks, otherwise -- don't do it.

I also tend to terminate my efforts when the temperatures under severe load exceed ~80C to ~85C. Again -- this is a personal choice. Of course, at 85C, one would deduce a TCASE around 75, and the specs are usually about 72.5C.

Ultimately, it's your ballgame -- not mine. Your choices -- and someone else might say I'm too cautious. Truth be told, once you've "certified stability" for a particular processor, it is never going to reach the extremes of temperature under which it was tested.

That leaves the issue of voltage, and a given level of voltage at mild load seems less likely to have the same impact as the same level at higher load. But tracking my own processor while playing an older game, I find that settings that give an "unloaded" Turbo voltage of 1.36 are only showing a VCORE fluctuating around 1.30V during the game play.

 

[steve wilson]

Thanks again for the help/info. The Noctua NH-D14 just fit in my PC thank god and was fairly straight forward to install. Temps are 65 degrees at the highest after 4 hours of prime 95 and I got the voltage down to 1.35.