Review: Scythe Ninja 4 CPU cooler

[lehtv]

忍者四 SCYTHE NINJA 4: the noble art of processor cooling​

I upgraded my rig from Ivy Bridge to the new Skylake, and decided to also switch to a new cooler. The previous Scythe Mugen 2 would surely have done the job adequately, however I felt that new parts deserve a bigger and prettier cooler. Because when it comes to CPU cooling, big is pretty:



I bought the Ninja 4 for 47.90 euros, which isn't too high of a price given the size of the cooler. Before we get into the meat of the matter, let's absorb some background information:

Scythe Ninja 4 features​

• 780 g heatsink with six 6 mm heatpipes and a nickel-plated copper base
• Scythe Glidestream 120mm PWM sleeve bearing fan with a wide RPM range
• good compatibility with different sockets, and being only 155 mm tall, it fits many of the narrower tower cases as well
• beautiful, distinctive, symmetrical shape, its face embossed with a Japanese shuriken

Comprehensive specifications: Scythe product page.

Installation​

Ninja 4 is fairly straight forward to install by following the instructions, however, the final part where the heatsink is secured on the installation bracket requires a bit of finesse. The screw holes are accessible only through the narrow vertical gaps in the heatsink. While attaching the cooler, the motherboard has to be horizontal, and a flash light is quite useful. All this is slightly awkward, but I can't think of a better way to install a cooler of this size and shape, so no points are taken away.

It comes supplied with a Phillips 2 screwdriver, which is good news for those building their first rig: the user doesn't need to bother acquiring a screw driver for installing other parts.

On my motherboard, the heatsink itself covers the first DIMM slot, while the attached fan overshadows the rest:



This is not a problem in and of itself, as standard profile memory fits with no issue. The first DIMM slot should be just about accessible without removing the heatsink.

Test bed​

Serving as the test bed, we have a test bench modded from an old ATX tower (more information in my thread on the Finnish site MuroBBS):

• CPU: i5-6600K (not delidded)
• Motherboard: Asus Z170M Plus
• RAM: Kingston 2x4GB DDR4-2400
• GPU: integroitu Intel HD 530
• SSD: Samsung 840 EVO 250GB
• HDD: WD Black 3TB
• PSU: Seasonic G-650 (SSR-650RM)

There's a 92 mm Enermax 1500 RPM fan installed, simulating the situation where the cooler is installed in a tower case with a rear exhaust fan.

Test methods​

Thermal paste: Arctic MX-4
Monitoring: HWiNFO64 and a digital thermometer for monitoring ambient temperature

The fan profile for both the case fan and the CPU fan is set to Standard in the motherboard BIOS. The CPU fan RPM does not reach its 1500 RPM maximum, instead it caps out at 950-1200 RPM depending on temperature in these tests.

Temperature testing was done mostly with Prime95 v27.7. Here are the tests I ran:

• TEST 1: Temperatures at different clock speeds: AUTO, 4100, 4200, 4300, 4400, 4500 sekä 4600 MHz. For each overclock, the lowest stable voltage was first acquired.
• TEST 2: 4500 MHz overclock investigated with Prime95 v28.5, IntelBurnTest 2.54 and x264 Stability Test programs. The first two stress the CPU more than P95 v27.7, whereas the x264 test reflects the maximum temperature you can expect in real world use. For more info on Skylake stability testing, see the OCN thread.
• TEST 3: Temperatures with the lowest fan speed, and with no fan at all (rear 92 mm fan still installed)

I ran each stress test for 10 minutes and took the average of the four core temperatures which I then normalized for 24.0 celsius ambient temperature and rounded to an integer.

Test 1: Temperatures on different OC settings​

Alright, bring it on!

I have to admit, I didn't quite expect such performance from a sub 50 euro cooler. One reason for this may be that in my previous setup I was used to the HyperThreaded i7-3770K. HyperThreading considerably increases heat output. In any case, temperatures stay well within safe margins in an unrealistically high load, even at a voltage which borders on the unsafe for this CPU and my entry-level Z170 board. Intel allows a maximum of 1.45V for the i5-6600K, but with my 4-phase VRM motherboard I'm not really prepared to go past 1.4V.

Temperature differences between cores were very small (2-5 degrees celsius depending on OC setting), but this is probably not to the credit of the Ninja 4. I think I just happened to get a CPU which heats up evenly, possibly due to thermal paste being spread evenly under the IHS.

So, what else can you hope from a cooler than temperatures low enough that the cooler doesn't become a bottleneck when overclocking? Well, you can certainly also hope for quiet operation in heavy loads. Scythe Ninja 4 is by no means silent - the maximum RPM at the highest clock speed was 1200 RPM which could put off those seeking a super quiet cooler, but should not bother most users. At idle, the fan (at Standard profile) runs at about 750 RPM which is very quiet (the WD Black hard disk is noisier). With Asus "Q-fan tuning" (which calibrates a new minimum RPM) I was able to lower the minimum to 600 RPM. It's possible to lower this even further, see Test 3.

Test 2: 4.5GHz in different stress tests​

Stress test          °C
-----------------------
x264 stability test  61
Prime95 27.7         69
Prime95 28.5         73
IntelBurnTest 2.54   73

4.5GHz stays as cool in the real world load that is the x264 benchmark, as it does at 4.4Ghz in Prime95 27.7 (see Test 1). The same can be expected when comparing 4.6Ghz to 4.5Ghz.

Test 3: Performance at low RPM and passively cooled​

The Glidestream fan has a handy manual switch with three settings, High, Medium ja Low:



According to official specs, M and L lower the maximum RPM to 1150 and 800, respectively. Apparently M lowers the operating voltage from 12V to 7V while L lowers it further to 5V.

The minimum fan speed in my testing was 400 RPM for both M and L settings, and 600 RPM for the default H setting.

According to this test, even the hefty 4.5GHz overclock stays cool enough at a practically silent fan speed. For comparison, the Standard fan profile achieves an RPM of 1050 and 14 celsius lower temperture.

With no fan at all, the cooler performs admirably in cooling a light overclock at low voltage. The downside with passive cooling, however, is that the heatsink stores more heat and the processor stays quite warm (about 50C) for several minutes post-load. The surrounding components also suffer, as there is no fan to move air around the heatsink. Even though passive cooling is easily feasible, I recommend people to use a fan at low RPM. If you have even one other fan or one hard disk installed, passive cooling does nothing to benefit the PC's noise output compared to running a very low RPM CPU fan.

Conclusion​

Scythe Ninja 4 is without doubt a high performance cooler. With i5-6600K, it is practically impossible to create high enough temperatures in normal use without at the same time supplying unsafe voltages to the CPU. To properly bring out the limits of this cooler, you need an overclocked i7. For CPU's less demanding than an i7, Scythe Ninja 4 makes it possible to run a nice overclock at a practically silent fan speed.

With respect to cost, Ninja 4 is mid tier. Unfortunately, pricing info doesn't seem to be available for North America. In Europe, you might be able to find the same performance for less cost (Thermalright HR-02 for instance), but at the same time the Ninja 4 is not too highly priced because one of its main selling points is the distinctive, beautiful design.

The Good
+ excellent performance for heavy i5 overclocking or somewhat lighter i7 overclocking
+ appearance
+ compatibility
+ PWM-fan with Low/Mid/High settings
+ compatible with other 120 mm fans
+ supplied with a Phillips 2 screwdriver

The OK
* cost
* installation

The Bad
- I'd gladly pay a little more for a long-life FDB or other bearing fan
- not compatible with 140 mm fans
- short 1 year warranty
- availability

Score: 8/10

 

[ClockHound]

Great review. Thanks.

Ninja 4 has the updated base and mounting system like the Mugen 4, Kotetsu and Max.

Would love to see what Scythe could do with a twin tower concept.

Did you see the SPCR Ninja 4 review? They stated using their reference 140mm fan resulted in 3 degree worse temps. Maybe they'll make a Ninja Max edition with a 140mm fan. But, that will blow the 155mm max height.

Again, thanks for sharing your Ninja experience.

 

[lehtv]

Did you see the SPCR Ninja 4 review? They stated using their reference 140mm fan resulted in 3 degree worse temps.

Yeah I bought it partly thanks to that review.

They must've used some other fan clips to install the 140mm fan because the supplied clips only work for a 120mm fan. The reason for the worse temps probably has to do with how the sides are grooved to fit a 120mm fan snugly within the corners of the heatsink. A 140mm fan will be too large to fit these grooves and so you'll get greater distance between the fan blades and the heatsink, resulting in worse temperatures.

Maybe they'll make a Ninja Max edition with a 140mm fan. But, that will blow the 155mm max height.

Now that would be awesome.

 

[aigomorla]

when taking measurements of heat, you need to find the Delta T from ambient.
This helps people get a grip of what a sink can do at a wider range of ambient.

Example... you had a 75C temps and say ambients were 25C.
That would give u a delta T of 50C.

This 50C allows us to ball park how efficient the sink is, because we can super impose this number +10-15C ambient. Anything more then that range the sink changes due to how a heat sink is more efficient at a higher temp load.


And if u really want to get perfect review quality, you also need to do at least 3 mounts.
You can never end a review with a single mount, because, no one can ever replicate a single mount over and over again without any form of variation.

It also helps the reader know how easy the sink is to install by seeing the differences in temps on each mount...

ie... if the temps are all near each other, the mounting on the sink is great, which is noob proof, as every mount is almost identical....
Vs.. seeing a large varation, which means, the heat sink has a poor mount, and care must be given when tightening the retention system.

 

[lehtv]

@aigomorla

Really... Nice of you to focus on the positive things for a change

when taking measurements of heat, you need to find the Delta T from ambient.

As mentioned in the review, results are normalized to 24C ambient. You can easily subtract that from the results to get delta T if you wish. Though it would've been a good idea to mention the ambient T in the graphs, maybe I'll fix them later ;P

And if u really want to get perfect review quality, you also need to do at least 3 mounts.

No, because this is not a comparison review. Even if I managed to completely mess up the mount, the conclusions in this review would still stand. In any case, who said I'm looking to get perfect review quality? I'm not getting paid for this, I'm just sharing my thoughts with the community.

It also helps the reader know how easy the sink is to install by seeing the differences in temps on each mount...

That's not how I would judge the difficulty of installation. A difficult-to-install cooler might use a ton of different mounting parts and require precision and extreme care to put together, with screws holes in difficult to reach locations etc, but in the end would line up with the IHS perfectly and evenly when all is done, and result in exactly the same temperatures with each mount. That is to say, ease of installation correlates with simplicity of installation, not so much with consistency of results.