A New Way of Testing Fan Airflow

[peTeMelster]

It seems like you're suggesting the S-Flex over the Noctua for use with the Ninja? Because you're saying the S-Flex would be able to push air through tighter spaces? I'm curious though, where you read SPCR was coming to the consensus that Noctua fans aren't able to push air through tight spaces as well?

Would the S-Flex 1200 RPM be at the same performance with the Ninja as the S-Flex 1600 RPM?

The reason I asked about the Noctua 1200 RPM fan vs. the Scythe S-Flex 1600 RPM is because SPCR is saying that above 1200 RPM, there is no cooling performance gain. This leads me to believe that the Noctua 1200 RPM fan will max out the Ninja's cooling capabilities, but I just want some confirmation. I don't want to buy the Noctua 1200 RPM fan and realize that the Scythe S-Flex 1600 or even the Scythe S-Flex 1200 (if the S-Flex is indeed able to push more air through tight spaces than the Noctua fans) can cool the Ninja even more.

As far as CPU temperatures go, I know 38-40C at idle isn't bad, but I expected better with the Ninja. Since according to AnandTech, the stock Intel HSF can idle the X6800 Core 2 Duo CPU at 41C. That's almost no cooling advantage. So now I suspect my Noctua 800 RPM fan is at fault, not pushing enough air to push the Ninja to its highest cooling capabilities. Which is why I'm asking which fan to get.

Here are the choices, which would you choose?
1. Scythe S-Flex 1200 RPM
2. Scythe S-Flex 1600 RPM
3. Noctua 1200 RPM
4. Stock Scythe fan (unsure of RPM or CFM)
5. SilenX Ixtrema 14 dBa version (although it's definitely no where close to 14 dBa)

The things I want are, in order of priority:
1. Pushing the Ninja to its upmost cooling capabilities.
2. Fan to be as quiet as possible while still achieving #1 (aka pushing the minimum amount of air needed to push the Ninja to its best cooling abilities)
3. Fan to have pleasant noise characteristics

Basically, I'm hoping to drop CPU temperatures by about 10C by changing fans from the Noctua 800 RPM fan. I'll try the stock Scythe fan in a bit and see how that goes, but I think it's too loud, and it definitely doesn't push the Ninja to its best cooling abilities (at least it doesn't push it to the best cooling according to AnandTech).

[jaganath]

your temps seem absolutely normal, see this thread:

viewtopic.php?t=41105

also there is some suggestion that CoreTemp/TAT are not reporting the correct temp readings.

[peTeMelster]

Well, that guy (the original poster) wasn't running a fan with the Ninja, and he's getting around the same CPU temps at me. This worries me greatly. With a fan, temperatures should be MUCH lower...shouldn't they?

Although those other guys with fans get around the same temps as me.

Then, how in the world did AnandTech get their Ninja to cool their Intel X6800 CPU (stock speed, not overclocked) to 27C idle with the SilenX Ixtrema 14 dBa version fan? Were they using some other CPU monitoring program that was reporting significantly lower temps?

I use TAT 2.05, Core Temp 0.95, and SpeedFan 4.32 and all three running simultaneously report similar temperatures (maybe varying from each other by 1-2C, but often, reporting exactly the same temperatures). Not sure why people are saying that SpeedFan temps are 15C lower than temps actually are.

[NeilBlanchard]

Hello,

Your ambient room temps are probably different, and actually, this is the beauty of the Ninja -- it works very well with just a little ("passive") air flow, and it does not take a high air flow fan to work well. Obviously, the trade off is that it "stops responding" to higher air flow fans.

If your temps are safe, and the machine is quiet enough, then stop worrying about it! I'll guess that the stock HSF at 41C idle is far from quiet...

[peTeMelster]

Well, I plan to OC my computer. That's why an idle temp of around 40C worries me. Sure, the Ninja heatsink + attached fan is super quiet right now, but it's not cooling to the best of its abilities. What I really want is to push the Ninja to its highest cooling abilities with a stronger fan, but not attach so strong a fan that the extra airflow doesn't benefit, and actually hurts because it becomes too loud.

I'd just like to know if SPCR's claim that at 1100-1200 RPM for a 120mm fan the Ninja's cooling performance is at its peak is true. AnandTech was able to push the Ninja to its peak by using a hi-flow SilenX Ixtrema 14 dBa fan, and pushed the Ultra 120 Extreme a lot too using a S-Flex 1600 RPM fan.

My question is whether using a Noctua 1200 RPM or S-Flex 1200 RPM would push both the Ninja and Ultra 120 Extreme to their max cooling performance level? I'm asking if AnandTech basically overdid it by using fans far faster than 1200 RPM, when they could have reduced the speed (and subsequently, also lowered the noise) from 1600 RPM or whatever down to 1200 RPM while STILL achieving EXACTLY the same peak cooling performance.

And if AnandTech could have achieved max cooling performance with a slower/quieter fan such as the Noctua 1200 RPM or S-Flex 1200 RPM, then which of those two fans would be better for the Ninja?

First and foremost, I want the best cooling ability with a fan attached to the Ninja at the expense of louder noise. But if what SPCR is claiming, that beyond 1200 RPM, the Ninja's cooling performance doesn't increase, then obviously, both the Noctua 1200 and S-Flex 1200 fans are equals in terms of cooling abilities.

Yet, I'm hearing now that the Noctua fans don't push air well through tight spaces. Would that mean the Noctua 1200 fan wouldn't help the Ninja cool as much as the S-Flex 1200 RPM fan?

If on the other end, the Noctua 1200 RPM does cool the Ninja equally to the S-Flex 1200 RPM fan, then which fan is quieter at their highest speeds (at the speed that would push the Ninja to the best cooling performance, around 1200 RPM purported by SPCR)?

All my questions in a nutshell:
1. Is SPCR's claim that the Ninja's cooling abilities do not improve beyond 1100-1200 RPM 120mm fans really true?

2. If SPCR's claim is true, then wouldn't that mean AnandTech used fans that were far too powerful and far too loud than necessary? They could have used quieter/slower fans to achieve the SAME MAXIMUM cooling peak on the Ninja, right?

3. If SPCR's claim is true, then wouldn't that mean the included stock Scythe fan with the Ninja Plus B is spinning SLOWER than 1200 RPM since obviously, the Ninja's performance was greatly improved by switching to a much higher airflow fan (higher RPM, louder) according to AnandTech's tests.

4. If SPCR's claim is true, then wouldn't both the Noctua 1200 fan and the S-Flex 1200 RPM fan be able to push the Ninja to its max cooling abilities? Yet, now SPCR is claiming that the Noctua can't push air through tight spaces very well? So what is up with that?

5. If the Noctua 1200 fan indeed can't push air as well through tight spaces, then doesn't this imply that the S-Flex 1200 RPM would help the Ninja cool more than the Noctua 1200 fan would?

In other words, is the S-Flex 1200 RPM fan better at cooling than the Noctua 1200 fan when both fans are used (one at a time) with the Ninja Plus B?

6. When at max speed, which fan is quieter at 1 meter: the Noctua 1200 or S-Flex 1200 RPM fan?

[HammerSandwich]

More airflow always provides more cooling, but diminishing returns set in quickly, especially for the noise-conscious.

If you're mostly concerned with getting the biggest possible OC, you'll need to start with a high-CFM fan. Find your OC, then reduce cooling while running Memtest/Prime95/whatever. When the system starts crashing, you'll then have to decide how to balance noise versus clockspeed.

[peTeMelster]

HammerSandwich, from what I've read in this published SPCR, Mike Chin is claiming that more airflow DOES NOT improve cooling ability. He is claiming (although it is a simplified claim) that beyond 1100-1200 RPM, 120mm fans will not cool CPU heatsinks any more, and will in fact start detracting because they will be louder than their 1200 RPM counterparts. He says this is because the rate at which heat is transferred from the CPU to the heatsink fins and passed off into the air is limited at a fixed rate, and once the fan's airflow matches this heat transfer rate, increasing airflow WILL NOT increase cooling since the heat isn't being transferred to the heatsink fins that fast.

That's why all my questioning.

[HammerSandwich]

I need to expand on my previous comment about diminishing returns.

Dump a given amount of heat into some air, and the air gets hotter. More air implies less temperature rise, because the heat is spread more thinly.

The CPU temperature vs. airflow curve is asymptotic. Starting from passive cooling, adding just a bit of forced airflow makes a big difference. More airflow continues to reduce the temperature, but at a declining rate. At some point, the curve turns a corner, and you need MUCH more airflow to achieve very little temperature reduction. The curve is nearly horizontal by this stage. What's happening is that the airflow's contribution to overall C/W eventually becomes irrelevant when compared to the fixed heat load.

Assuming we cool 100W, let's see how much the air's exit temp increases (rounded):

Code: Select all

CFM deltaC
  2   88
  4   44
  8   22
 16   11
 32    5
 64    3
128    1

SPCR's test with the high-speed JMC fan demonstrates the point. Doubling the quiet fans' airflow, the JMC probably did provide greater cooling, but it was only a fraction of a degree C. This difference was below the testbed's resolution, so SPCR found no difference.

The other possibility is that the 7W drawn by the JMC heated the cooling air significantly, even before it reached the heatsink. (Given that this was about 0.15C, I doubt it...) For a more extreme example, go ask some watercooling gurus if you should use a 100W pump to cool a 50W CPU.

Either way, MikeC's conclusions are correct for the practical user.

[peTeMelster]

That makes sense.

So then for all practical purposes, the SPCR claim is true.

Then, again, my questions:

1. Since SPCR's claim is true, then wouldn't that mean AnandTech used fans that were far too powerful and far too loud than necessary? They could have used quieter/slower fans to achieve the SAME MAXIMUM practically measurable cooling peak on the Ninja, right?

2. Since SPCR's claim is true, then wouldn't that mean the included stock Scythe fan with the Ninja Plus B is spinning SLOWER than 1200 RPM since obviously, the Ninja's performance was greatly improved by switching to a much higher airflow fan (higher RPM, louder) according to AnandTech's tests.

I wonder if the stock Scythe fan really is spinning slower than 1200 RPM though.

3. Since SPCR's claim is true, then wouldn't both the Noctua 1200 fan and the S-Flex 1200 RPM fan be able to push the Ninja to its max cooling abilities? Yet, now SPCR is claiming that the Noctua can't push air through tight spaces very well? So what is up with that?

4. If the Noctua 1200 fan indeed can't push air as well through tight spaces, then doesn't this imply that the S-Flex 1200 RPM would help the Ninja cool more than the Noctua 1200 fan would?

In other words, is the S-Flex 1200 RPM fan better at cooling than the Noctua 1200 fan when both fans are used (one at a time) with the Ninja Plus B?

5. When at max speed, which fan is quieter at 1 meter: the Noctua 1200 or S-Flex 1200 RPM fan?

6. And finally, an additional question. Which 120mm fan (Noctua, S-Flex, and SilenX Ixtrema) pushes the most CFM given a fixed RPM (say 1200 RPM)?

[Erssa]

I'd just like to know if SPCR's claim that at 1100-1200 RPM for a 120mm fan the Ninja's cooling performance is at its peak is true. AnandTech was able to push the Ninja to its peak by using a hi-flow SilenX Ixtrema 14 dBa fan, and pushed the Ultra 120 Extreme a lot too using a S-Flex 1600 RPM fan.

The reason for this is very obvious. The Scythe B fan was mounted the wrong way, it was pulling instead of pushing air. This can be seen clearly when the results are compared to the almost identical OCZ Vindicator. Vindicator has 200rpm slower fan then Ninja, but it performed considerably better at stock. Yet Ninja had much better performance with the SilenX.

I emailed Anandtech about this, but never recieved an answer.

[Spanki]

Rather than answer all of the above questions, let me try to put you out of your misery...

The one piece of the puzzle you're missing (and others haven't pointed out) is that...

Since SPCR's claim is true...

...is true, for each given level of input heat. Transfering heat from one source and/or medium to another (via convection?) is largely governed by the difference in temperatures between any two sources in the pipeline:

cpu->IHS->TIM->HSF baseplate->heatpipes->fins->fan-fed case air->room air

...be sure not to gloss over this:

What this tells us is that for this combination of CPU, load, and heatsink, the airflow of an 1100 RPM 120mm fan matches the rate of thermal conduction from CPU to heatsink fins.

...if you generate more heat at the cpu level (using the same HSF), you're either going to have to move the air across the fins faster or lower the temperature of that ambient air.

So, SPCR does thier tests with X amount of input heat being generated (they're not interested in results of other heat levels, according to a post on the first page of this thread). What they are saying is that for X amount of heat, with a similar ambient air temp, using a similar case and/or airflow configuration, you reach the point of diminishing returns above ~1200 rpm on a Ninja.

Note that 1200rpm is not some magic number for the Ninja in the general sense. Thier results do not say anything about a CPU putting out twice as much heat, or a situation with higher ambient temps, or one with a more constricted airflow. Each of those conditions would have a higher "rpm point of diminishing returns".

This explains why AnandTech is getting more cooling with faster fans, when SPCR didn't see the same effect - they're testing two different configurations.

Having said the above, that 1200rpm figure for the Ninja (under the SPCR test conditions) may or may not be the right number for the Tuniq Tower, Ultra 120 or any other heatsink. I believe that the general conclusion that SPCR reviews have come to is that the Ninja gives the best amount of cooling, for that speed (generally related to noise level) fans. A Tuniq or Ultra 120 might require a faster fan (because of tighter fin spacing) to cool as well as the Ninja (again, at the tested configuration).

The flip-side of that coin is that that "point of diminishing returns" on fan speed is reached earlier on the Ninja than the heatsinks with more tightly spaced fins (all other factors being equal).

In short, while what they are saying may well be true - for thier test configuration - it can't be applied "literally" to other configurations - it's just one data point.