Scythe Katana

Cooling
Viewing page 4 of 5 pages. Previous 1 2 3 4 5 Next

TESTING


On the test bench...

Test Platform

  • Intel P4-2.8A The Thermal Design Power of this P4-2.8 (533 MHz bus) is 68.4 or 69.7W depending on the version. As the CPU is a demo model without normal markings, it's not clear which version it is, so we'll round the number off to ~69W. The Maximum Power, as calculated by CPUHeat & CPUMSR, is 79W.
  • AOpen AX4GE Max motherboard - Intel 845GE Chipset; built-in VGA. The on-die CPU thermal diode monitoring system reads 2°C too high, so all readings are compensated up by this amount.
  • OCZ DDRAM PC-3200, 512 MB
  • Seagate Barracuda IV 40G 1-platter drive (in Smart Drive)
  • Seasonic Super Tornado 300 (Rev. A1)
  • Arctic Silver Ceramique Thermal Compound
  • Nexus Real Silent 92mm fan
  • Two-level plywood platform with foam damping feet. Motherboard on top; most other components below. Eases heatsink changes and setup.

Measurement & Analysis Tools

  • CPUBurn processor stress software
  • SpeedFan version 4.25 software to show CPU temperature
  • A custom-built fan controller that allows us to dial in exactly what voltage is powering the fan

Noise and airflow measurements were made with the fan powered from the fan controller while the rest of the system was off to ensure that system noise did not skew the measurements. The Katana was tested both with its stock fan and our standard 92mm reference fan, a Nexus. Airflow measurements were made while the fans were mounted on the heatsink, and are lower than they would be if they were measured in free air.

Load testing was accomplished using CPUBurn to stress the processor, and the graph function in SpeedFan was used to make sure that the load temperature was stable for at least ten minutes.

The ambient conditions during testing were 18 dBA and 24°C.

TEST RESULTS

Scythe Katana with Stock Fan
Fan Voltage
Airflow
Load Temp
°C Rise
°C/W MP
°C/W TDP
Noise
12V
25 CFM
47°C
23
0.29
0.33
33 dBA/1m
9V
21 CFM
49°C
25
0.32
0.36
26 dBA/1m
7V
16 CFM
51°C
27
0.34
0.39
22 dBA/1m
5V
11 CFM
54°C
30
0.38
0.43
~19 dBA/1m

Airflow: Measured in Cubic Feet per Minute mounted on the HS
Load Temp:
CPUBurn for ~20 mins.
°C Rise: Temperature rise above ambient at load.
°C/W MP / TDP: Temperature rise per Watt, based on CPU's Maximum Power (79W) or Thermal Design Power (69W) rating (lower is better)
Noise: SPL measured in dBA/1m distance with high accuracy B & K SLM

12V / 25 CFM: The stock fan Katana is quieter than most heatsinks at full speed but is still too loud to really be called quiet. The cooling performance at this level is perfectly acceptable for our CPU, which means it is adequate for most AMD based systems. Current AMD CPUs dissipate about as much heat as our 2.8 GHz Northwood, with the high end models getting a bit hotter. However, Intel's more recent Prescott and Pentium D processors run much hotter, and will probably require more cooling than the Katanda can provide. We can extrapolate that a CPU with a 115W TDP, such as the high end Pentium 6xx models, would see a 38°C rise over ambient.

9V / 21 CFM: Dropping the fan voltage to 9V makes a dramatic difference to the noise level, which was measured at 26 dBA/1m. The load temperature only rises by 2°C, so the reduction is well worth the thermal penalty.

At this level, the fan is quiet but not silent, so the quality of the noise is quite important. Most of the noise comes from the motor, which gives off a soft hum somewhere around middle C. A tiny amount of buzz from the bearings can be heard on close listening, but otherwise the sound is very clean and constant.

7V / 16 CFM: For another 2°C rise, the fan noise can be reduced to a level that should be around ambient in most environments, although it is still clearly audible in our lab. The noise character doesn't change much except for the lower volume and the lower pitch of the motor hum. The hum is very slightly rougher, but not enough that it will be clearly audible once it is installed in a case. Most users should be happy with the noise level at 7V, and the performance is still good enough for most AMD-based systems, although a system with poor airflow may have problems.

5V / 11 CFM: At 5V, the fan is around the ambient noise level in our lab, which means it should be more or less silent once it is installed in a case. A trace of motor hum can still be heard if you put your ear up to it, but it is almost as quiet as the light buzz from the bearings. Performance at this level is beginning to suffer a bit; the CPU is now 7°C warmer than the temperature with the fan at 12V. Most users will not see much acoustic benefit undervolting past 7V as it is likely to be below the ambient noise level — or the louder components in their system.

Scythe Katana with Nexus Reference Fan
Fan Voltage
Airflow
Load Temp
°C Rise
°C/W MP
°C/W TDP
Noise
12V
20 CFM
51°C
27
0.34
0.39
23 dBA/1m
9V
14 CFM
53°C
29
0.37
0.42
20 dBA/1m
7V
11 CFM
57°C
33
0.42
0.48
<19 dBA/1m
5V
7 CFM
66°C
42
0.53
0.61
<19 dBA/1m

Airflow: Measured in Cubic Feet per Minute mounted on the HS
Load Temp:
CPUBurn for ~20 mins.
°C Rise: Temperature rise above ambient at load.
°C/W MP / TDP: Temperature rise per Watt, based on CPU's Maximum Power (79W) or Thermal Design Power (69W) rating (lower is better)
Noise: SPL measured in dBA/1m distance with high accuracy B & K SLM

Our reference Nexus fan is even slower and quieter than the stock fan, so the results are somewhat higher temperatures and lower noise. However, when the two fans are compared at the same airflow, the noise levels are very similar, and it is difficult to say which fan is quieter at a given speed. Because the Nexus is a slower fan, its can potentially be even quieter than the stock fan, but very few users are likely to notice the difference since even the stock fan will be below ambient most of the time.

One thing that is noticeable is that cooling performance drops off rapidly when airflow drops below ~11 CFM, which means that running the Nexus fan at 5V is impractical for most users. The load temperature jumps 9°C when the fan is dropped to 5V from 7V, whereas the drop to 7V from 9V caused only a 4°C rise. The performance-for-noise "sweet spot" is right around the point when the stock fan is running at 5-7V: The noise level is near ambient, but the cooling performance is still not bad.



Previous 1 2 3 4 5 Next

Cooling - Article Index
Help support this site, buy from one of our affiliate retailers!
Search: