Do people set their chassis fans to PWM?

[SpyderTracks]

Question for you all as I'm truly interested.

This has just come up in a thread where a user was saying that his case fans were ramping up in games and becoming too loud.

Now I've never ever set any case fans to PWM, I've got a few PWM fans mounted purely because they were the highest quality airflow movement.

But I would never bother setting case fans to variable speeds as in my experience it's never really given any particular cooling improvement compared to the noise it generates.

Am I in the minority here? I know loads of you have Corsair cases which will have high quality PWM fans on, do you alter the speed based on thermals or some other metric?

 

[Martinr36]

Yep, run mine through iCUE, front 3 on one custom curve for inlet, and rear 3 set up as exhaust, slightly slower, so as to pressurised the case slightly

 

[Ekans2011]

Question for you all as I'm truly interested.

This has just come up in a thread where a user was saying that his case fans were ramping up in games and becoming too loud.

Now I've never ever set any case fans to PWM, I've got a few PWM fans mounted purely because they were the highest quality airflow movement.

But I would never bother setting case fans to variable speeds as in my experience it's never really given any particular cooling improvement compared to the noise it generates.

Am I in the minority here? I know loads of you have Corsair cases which will have high quality PWM fans on, do you alter the speed based on thermals or some other metric?

When I had access to my desktop (which is currently MIA in Ukraine ), I always kept the chassis fans (both intake and exhaust) at full design speed.
Noise has never been an issue for me as an engineer officer, and I always wear headphones.

 

[SpyderTracks]

When I had access to my desktop (which is currently MIA in Ukraine ), I always kept the chassis fans (both intake and exhaust) at full design speed.
Noise has never been an issue for me as an engineer officer, and I always wear headphones.

What's your pc doing in Ukraine? Were you over there?

 

[SpyderTracks]

Yep, run mine through iCUE, front 3 on one custom curve for inlet, and rear 3 set up as exhaust, slightly slower, so as to pressurised the case slightly

If you ever fancy it, do me a huge favour

Do a stress on both cpu and gpu wirh furmark and cinebench r23 with the way it's currently setup

Would you say 30 mins was probably enough time for case thermals to normalise?

Then disable PWM just on the case fans and leave them on native speed, and run the test again.

Take full sensor readings before finishing the test

Your setup is far more geared to full PWM than mine is, I'd be so interested in the temp readings at the end of the tests

I'll do the same on mine too but will have to rewire some bits.

 

[Ekans2011]

What's your pc doing in Ukraine? Were you over there?

Yep, I had been living there since my marriage in 2016, but everything fell apart two years ago while I was at work (in UK ), and I had to 'force' my wife to move to Italy so we could see each other again.

Anyway, when this nonsense is over, we'll go back to Ukraine. Life was very good for me there

 

[SpyderTracks]

Yep, I had been living there since my marriage in 2016, but everything fell apart two years ago while I was at work (in UK ), and I had to 'force' my wife to move to Italy so we could see each other again.

Anyway, when this nonsense is over, we'll go back to Ukraine. Life was very good for me there

Oh mate, had no idea you lived there. I'm glad your wife got out. I sincerely hope all this madness ends soon, Ukraine was such a beautiful country, largely due to the passion of the people who live there and the respect they have for their heritage. I have no doubt they'll rebuild it just as beautiful.

 

[TonyCarter]

My front fans are connected to Commander Pro and are on a profile to ramp up with temps…mainly because the bottom fan was very noisy due to blowing air directly against the HDD cage in front of it.

I’ve set the bottom one on a static 1200rpm which is silent (or quiet enough that I can’t hear it) and the middle one ramps up the earliest, followed by the top if the temps get really hot...so the system stays quiet even in the tiny 220T case with a 4090 in there…until I play something really demanding, but even then the fans aren’t obtrusive.

 

[sck451]

I have all my fans (case and cooler) on PWM from motherboard headers. When idle, this means my system is silent. Under load, the temperatures are pretty reasonable without the noise being incredible.

But I'm running (by this forum's standards) an unusual setup, with an overclocked (well, PBO and Curve Optimizer) 5900X on an air cooler. No way could it cool the CPU without ramping up the fans.

 

[SpyderTracks]

But I'm running (by this forum's standards) an unusual setup, with an overclocked (well, PBO and Curve Optimizer) 5900X on an air cooler. No way could it cool the CPU without ramping up the fans.

So you're saying that having the case fans not on PWM would cause the CPU to overheat? That doesn't make sense? The case fans have no impact on the CPU cooling, only in moving air in the case out of the case?

 

[sck451]

So you're saying that having the case fans not on PWM would cause the CPU to overheat? That doesn't make sense? The case fans have no impact on the CPU cooling, only in moving air in the case out of the case?

They move cool air that's not in the case into the case. That's where the air comes from to cool the CPU, no?

You have me curious now. I'll try it tomorrow and see what the difference in temperature is with the case fans set to minimum (which is around 330RPM) and not ramping up.

 

[AgentCooper]

I set my fans based upon which ones I think may have been slacking in previous weeks.

Same with my weekly recycling, I’ll put anything in that I think deserves a second chance.

 

[SpyderTracks]

They move cool air that's not in the case into the case. That's where the air comes from to cool the CPU, no?

You have me curious now. I'll try it tomorrow and see what the difference in temperature is with the case fans set to minimum (which is around 330RPM) and not ramping up.

Yeah, absolutely they bring cool air in which the CPU cooler expels it's hot air into, thereby raising the case air.

But you have 2 kinds of fans, static pressure for coolers and radiators, those aren't about volume of airflow, they're about the force they push it under, the idea being they can force that pressure through the fin stacks more efficiently. Then you have high airflow fans for case fans, those are all about pushing high volumes, these are for case fans to get that airflow generated from input to output in the case.

But the whole point of case fans is that they're high airflow fans, designed to move large quantities of air measure by CFM (cubic feet per minute).

On any modern case fan (of reputable standard), the volumes they push are almost always around 60CFM which is roughly 1 cubic feet per second. Most case are around 1 cubic foot, maybe 2 on a large case, so the reality is that every second or two you've got entirely new air in the case assuming you've got an optimal fan configuration that's not causing dead spots, and that the case has decent airflow design.

So if the internal case temps are actually rising when you have the fans at their optimal speed (all fans state their optimal speed) there's something really odd going on and there's some kind of dead spot that needs sorting (assuming the cooler is suitable for the CPU for instance)

Setting a case fan at optimal speed (again, assuming a decent manufacturer) that speed is normally the optimal noise vs performance setting.

So if you set them static at optimal speeds, you shouldn't hear the case fans at all, and still be recycling the case air every second. The only fan you should ever hear is the cooler fans on GPU and CPU.

Caveat here, it's more an more becoming the case that you now have one fan design that's both effective for static and air flow. Noctua's really pioneered this with the NF12, Corsair top end fans now are very effective with static pressure as well, so it's starting to blur the lines between needing different fan types per application. But the same principle is true.

 

[SpyderTracks]

They move cool air that's not in the case into the case. That's where the air comes from to cool the CPU, no?

You have me curious now. I'll try it tomorrow and see what the difference in temperature is with the case fans set to minimum (which is around 330RPM) and not ramping up.

I'd be really interesting what you find, I'll do mine too, although my case really isn't a particularly great airflow case.

 

[sck451]

Right, I have done my testing.

First, the setup. I am running a 5900X on a Tuf X570 motherboard. The chip is overclocked with PBO and Curve Optimizer. On an all-core load it runs at about 4.5GHz, pulling up to 220W. The case is a 275R Airflow (the one with the asymmetric front panel). There are two NF-A14 fans fitted to the front panel; I have no other case fans. The cooler is a NH-D15S (the one-fan version). If it matters, the graphics card is a 3060 Ti FE and the memory is Kingston Fury DDR4-3600 C16 2x16GB. Sadly, the performance of my setup is hampered because the only RGB I have is on my mouse.

I have used two setups. The first is my regular setup: the "Silent" ASUS BIOS preset, which has the case fans at 350RPM (i.e. silent) at idle, then ramping up to 1500RPM at full load (this is fairly loud, though not unpleasant). Then I have tried it with the case fans locked at 350RPM and only the cooler fan ramping up. (If anyone's interested, I used the excellent Fan Control app and set the case fans to 22%.)

First, single-run Cinebench R23 multicore scores.

Regular setup
Cinebench: 22669
Max temperature: 88.1°C

Locked fan setup
Cinebench: 22680
Max temperature: 89.1°C

So far I think we'd agree "nothing to see here". But my thesis (opposed by @SpyderTracks) is that the extra fan speed on the front will stop heat build-up in the case, which means there will be more cooler air for the big fat Noctua cooler to use to cool the CPU. So I did a 10-minute run of Cinebench for each of the same setups. In each scenario, the CPU was thermally limited at 90°C, so the interesting factor is what levels of power draw and CPU frequency could be maintained.

I logged this with HWiNFO64. Rather than uploading big CSVs, here are some quick Excel charts. Be aware that, in both cases, I have shortened the Y axis to show the difference more clearly. (Sorry for the X-axis labelling, by the way. Blame my shoddy Excel skills.)

Now, I'll be the first to admit that this test could be more scientific. I'm sure there are things I haven't controlled properly! If anyone wants to suggest some improvements to my testing method (maybe using Prime95 would be better?) I'd be happy to do it. I'd also quite like to run some graphics card checks, though I doubt I'd be thermally limited in any scenario with my graphics card.

I'm also confused at why the "ramping" test seemed to build up power draw and CPU speed over time. This feels unlikely to me. Perhaps the CPU was too hot before the test began? But I think over 10 minutes it will have evened out.

Conclusion
Frankly, this is a bit of an ambiguous set of testing. First, I was right! Having the case fans ramping up does make a difference. At the end of the run, the fans set to ramp with PWM allowed the CPU to draw an extra 15W of power, and around 75MHz of higher frequency.

However, @SpyderTracks was right! This difference is less than 2%, and the fans ramping up to 1500RPM was quite loud. I'm not sure the noise level is worth it.

 

[SpyderTracks]

While that’s interesting, setting the fans as low as 350 is nowhere near their optimal, so they wouldn’t be pushing their designed airflow, any case fan should set at it’s rated speed to be sure you’re pushing the rated CFM.

So the noctua NF- A14s would need to be set at 1200rpm (assuming you’re using the low noise filter which why wouldn’t you with a noctua), at this speed there should be very little noise while optimising airflow.

It would be interesting to see the temp difference between 1200rpm with low noise and 1500rpm where you can audibly hear them.

That graph output is superb btw!

The reason the cpu is getting higher frequencies with the case ramping is because it on air cooling, this is an example where these chips really do need an AIO to achieve higher boost clocks. So your case fans are supplementing the cooler (which shouldn’t be needed). The fact it’s still thermal throttling shows the cooler is not adequate for the cpu.

With an AIO, the chip wouldn’t be thermal throttling even under sustained loads, and the frequency would be higher. The case airflow speed wouldn’t make any difference.

 

[sck451]

While that’s interesting, setting the fans as low as 350rpm is nowhere near their optimal, so they wouldn’t be pushing their designed airflow, any case fan should set at it’s rated speed to be sure you’re pushing the rated CFM.

But they're silent. That for me is the sine qua non of the whole setup: silent at idle, able to cool properly under load.

 

[SpyderTracks]

But they're silent. That for me is the sine qua non of the whole setup: silent at idle, able to cool properly under load.

But they’d be silent at far higher RPM as well.

At 1200rpm with the low noise filter or without, they should still be remarkably quiet.

I’ve got the older NF-F12s as a few case fans set to static 1200rpm, and it’s like a person breathing in the background. The cpu cooler is the predominant noise on idle.

 

[Martinr36]

Ok I've run some tests here as well, Test 1 is with the case fans set on quiet, and the cooler fans set on its normal custom curve, with the pump set on balanced (pump and cooler settings the same for all 3 tests)

Case fans on quiet no load

Case fans on quiet 10 minute stress test

Test 2 is with case fans set on balanced

Case fans on balanced no load

Case fans on balanced 10 minute stress test

Test 3 Fans set to the normal custom curves I use

Custom Curves no load

Custom Curve 10 minute stress test

Looking at those results, I should run my case fans on balanced or tweak my custom curves slightly as the package temp on balanced is a couple of degrees cooler than my custom curves

 

[Scott]

I definitely wouldn't run at 350RPM. In the other thread I had mentioned that I tested and found a nice middle ground where the acoustics from the case fans was effectively nil. I can't remember what RPM this was at, but off the top of my head I have 1000 in my head.

I found pretty much exactly the same situation as you did @sck451, there was a little difference but there was a difference between minimum and maximum. At the 1000RPM mark (or whatever I had it set to) the difference was effectively nil.

In your situation I think it would be worth hammering both the CPU and the GPU though, as they will have an overall impact in the case temp delta and they will both heat soak to a point over time if the airflow isn't enough to provide them with fresh air.

Regarding the CFM discussion, it's not the fresh air that's the limitation. It's the CFM of the air cooler that needs to be considered. If the cooler can flow 60, the case needs to be able to recycle more than 60 to keep it fresh. The GPU cooler comes into play as well, as it needs provided with fresh air too. This is where the pressure also needs balanced though as it's all well and good battering in 120cfm but if you can't expel it you make a positive pressure with nowhere for that air to go so it leaks out of various places and the 120cfm ends up not actually entering the case (hence why slight vacuum is preferred to show well balanced cooling).

I'm glad I don't need to worry about this any more