Review: 4 Socket-A Heavyweight Heatsinks

Cooling
Viewing page 6 of 6 pages. Previous 1 2 3 4 5 6

TEST RESULTS

Each heatsink was cleaned and installed in turn on the test system as per the manufacturers' and Arctic Silver's instructions. The system was allowed to cool between tests for ~5 minutes with a large AC domestic room cooling fan running at full speed, the airflow directed over the entire test setup. The test platform and procedure is described in detail in our HS Test Methodology article, and so will not be covered here. (Do read the piece for a fuller understanding of our approach!) The reference Panaflo fan was simply placed on top of each heatsink, blowing down.

The complete set of tests were repeated on three consecutive days to check to the consistency of the test system. The variance for the diode reading with Prime95 after 20 minutes was within 2° C. That is, the highest and lowest readings were within 2° C of each other.

Interestingly (and this is noted with some relief), the variances were quite consistent between the heatsinks in each test set. In other words, compared to the tests run on day 1, all the day 2 tests were 2° C higher, and all the day 3 temperature were 1° C higher. It is not possible to determine the precise cause of this variance, which may be cumulative. It is 5% in the worst case. The numbers shown here are the averaged results.

There was some concern about whether the use of a 80 mm fan instead of the 92 mm size supplied with the Zalman 6000Cu might hurt its performance. A 92mm Silencer fan (from PC Power & Cooling), rated for 23 dBA and 28 CFM, was substituted for the Panaflo on the 6000Cu at several test points. There was no appreciable change in CPU temperature. Due to the many slots in each fin, the airflow from the Panaflo can be felt all around the HS, despite the fan being much smaller than the top area of the fins.

The commentary below refers mostly to the diode temperatures with Prime95, as core temperature under long term 100% load is the most critical data.

Note that 20+ minutes of Prime95 by itself is far more stress than most user would subject their system to in normal circumstances.

Reference Panaflo fan

Reference Panaflo FBA08A12L1A 80mm fan: 24 CFM, 21 dBA @ 1 m, 12V

  • All temperatures in degrees Celsius.
  • Diode: reading from XP1600+ CPU internal thermal diode
  • Socket: reading from in-socket thermistor on ABIT KT7A-R motherboard, latest BIOS
  • Board: reading from DigiDoc thermistor placed ~ 4 inches over northbridge chip on motherboard
  • The room temperature during all the testing: 22° Celsius., measured with DigiDoc thermistor ~6 inches above CPU HS and fan.

Results at 12V: Much as Expected

Model
° C at Idle, 10 mins
° C with Prime95, 20+ mins
Diode
Socket
Board
Diode
Socket
Board
SLK800
33
31
24
41
35
24
AX7
38
34
24
48
40
24
6000CU
43
36
25
56
45
28
MC462A
38
33
25
48
40
26

The Panaflo is rated for 24 CFM airflow in free air at 12V. The Thermalright SLK800 comes out on top by a substantial margin, especially at 100% CPU load. Second place is shared by the AX7 and Swiftech MC462A, which post identical numbers, a surprising 7° C behind the leader. The Zalman comes in last, a whopping 15° C behind the leader and 8° C behind the second placers when under load.

Results at 7V: Tighter grouping

Model
° C at Idle, 10 mins
° C with Prime95, 20+ mins
Diode
Socket
Board
Diode
Socket
Board
SLK800
40
36
24
50
44
25
AX7
44
38
25
55
47
26
6000CU
49
43
27
63
52
29
MC462A
44
39
25
55
48
26

At 7V, the Panaflo can be expected to blow about 14 CFM (according to calculations). It is very quiet at this voltage. Rankings are little changed, but the gaps between the contestants have narrowed by a few degrees.

Results at 5V: An Upset!

Model
° C at Idle, 10 mins
° C with Prime95, 20+ mins
Diode
Socket
Board
Diode
Socket
Board
SLK800
49
42
25
70
58
27
AX7
50
45
26
63
56
27
6000CU
59
51
28
79*
64
30
MC462A
51
47
26
70
57
27

*Stopped before 20 minutes -- see text below.

At 5V, the Panaflo is pretty much inaudible but still blows about 10 CFM (according to calculations, not by measurement). Interestingly, rankings have changed dramatically. At idle, the Thermalrights and the Swiftech are tightly placed in a 2° C window. This is within the margin of error for the test system; they might be considered about equal. But with the CPU under load, the AX7 is now clearly top dog, and the SLK800 and the Swiftech share second place 7° C behind! *The Zalman probably hit thermal overload; the temperature was still creeping up when the plug was pulled on after 18 minutes to avoid CPU damage.

High Power Fan Check

Finally, a high power fan that is likely to be used by performance enthusiasts was brought in for a kind of reference check. The fan is a Sanyo-Denki 80 x 32 mm model 109P0812A201 rated at 53 CFM, 45 dBA, 4600 rpm, 0.56 amp. Quieter than the 80 CFM Delta that is sometimes marketed with high power HS, but loud enough to deafen most ordinary folks. It is the official lower noise fan that came with the Swiftech MC462A. Measured 1 cm from the frame (like the Panaflo), the noise reading is 85 dBA(!), and the vibration it produces is enough to make the whole test platform hum and buzz when simply placed on top of the HS. Here's what it does with these HS:

Sanyo-Denki fan @ 12V
Diode reading in ° C
Idle
Max
SLK800
31
37
AX7
37
44
6000CU
39
47
MC462A
36
41

The SLK800 comes out on top as expected. The Swiftech rises with the high airflow, confirming the often repeated statement that it is optimized for performance with high airflow. The AX7 falls a few degrees behind, and the Zalman brings up the rear with a perfectly respectable 47° C. Now, none of the contenders exceeds 50° C, which is remarkable considering this is the in-die core temperature. The price is a gradual decline into deafness.

DATA ANALYSIS

The test system's AMD XP1600+ (Palomino core) is rated to produce 62.8W maximum at the default 1.75V Vcore. All monitoring programs show the test motherboard's Vcore, set to default, to be 1.81~1.84V, slightly higher than normal. This puts the power up a bit; let's call it 65W. This assumes Prime95 working in Windows XP causes the XP1600+ to dissipate 65W.

HS cooling performance is best summarized by ° C/W or how many degrees the object temperature rises for each watt of heat it dissipates. A specification of 1° C/W means temperature rises by 1° C for each watt of heat -- this is very bad performance for a CPU HS, BTW.

Using 22° C (measured 6 inches above the fan intake) as the room ambient temp, the ° C/W calculated for each HS at each fan voltage (with Prime95) is shown below.

Temp rise is probably the simplest, most useful data here. If you're considering using one of these HS with a similar CPU, just add the appropriate temp rise number to your case ambient temp and you'll get a reasonable prediction of performance in your setup. Temp rise refers to the temperature rise in ° C over ambient temperature - the difference between diode temp readings and the room ambient.

Model
Panaflo 12V

Panaflo 7V

Panaflo 5V
Sanyo-D 12V
Temp rise
° C/W
Temp rise
° C/W
Temp rise
° C/W
Temp rise
° C/W
SLK800
19
0.29
26
0.40
48
0.74
15
0.23
AX7
26
0.40
33
0.49
41
0.63
22
0.34
6000CU
34
0.52
41
0.63
58
0.88
25
0.38
MC462A
26
0.40
33
0.49
48
0.74
19
0.29

Note that the high airflow (53 CFM) Sanyo-Denki has the most dramatic impact on the 6000CU: an improvement of 9° C (compared to the result with the 24 CFM Panaflo). The Sanyo-Denki fan improves the Thermalright models only by 4° C and the high-airflow optimized Swiftech by 7° C.

CONCLUSIONS

Real world conditions are tougher than those in our lab. When a system is installed in an enclosed case, the ambient temperature is sure to rise at least 5° C, sometimes 10° C or more, depending on particulars. Many quiet PC enthusiasts run systems with no case fan or just one low airflow fan. In consideration of these factors, low noise enthusiasts are urged to add 10° C to the results above when trying to guesstimate what their temperatures would be. Note too, that the results here are specific to the XP1600+ Palomino.

It is best to regard the test results not in an absolute way, but rather, as comparative guidelines. The result obtained with any of these heatsinks will vary greatly on the particulars for each system.

#1 - Thermalright SLK800: There is little question that the SLK800 is a topnotch performer with almost any level of airflow. Its overall design is ingenious and well executed, and extremely user friendly. Its small footprint makes it feasible to install even on motherboards with little room around the socket. We are in agreement with the overclocking and performance community: this is one hell of a HS.

#2 - Thermalright AX7: Beaten out from second spot by the Swiftech only in the extreme airflow contest (which we don't really care about here), the AX7 is an excellent performer and a tremendous value. Like the SLK800, its clip is user-friendly and easy to use, even though its larger size makes it difficult on tight motherboards. Its performance with the 5V Panaflo is an amazing victory -- but remember that the actual temperature reached was in excess of 60° C. If you add 10° C for an approximation of in-case temperature, use of a 5V Panaflo with this HS -- or any other in this round up -- on an XP1600+ or higher power processor is not a safe recommendation.

#3 - Swiftech MC462A: There are good reasons why this model was the kingpin for so long. It retains its edge well against newer competitors sporting more recent technologies. Sheer size and mass, combined with manufacturing excellence and a great mounting system combine for good cooling performance even with low airflow. The performance of its descendants is something I look forward to examining.

#4 - Zalman 6000Cu: Saved from humiliation with its fine high-airflow performance, the 6000Cu suffers in comparison with the other contenders. The soft 2-lug clip is the weakest part of its design, and is probably responsible for the poor showing here. An afterthought following the tests: Does this particular Zalman clip provide enough pressure between the CPU die and the HS base? Weeks of preliminary test setup experimentation meant all of the HS in this survey were mounted and removed from the motherboard many times. It is possible that the Zalman clip suffered more deformation and metal fatigue than the clips on the Thermalright models, causing a reduction in contact pressure. A follow-up on this suspect clip is promised.

This review has naturally resulted in a few small changes of the rankings in our Recommended Quiet Heatsinks, which you'll want to review before making any buying decisions.

Socket-A Heatsink Roundup Intro - p.1

Thermalright SLK800 - p.2

Thermalright AX7 - p.3

Zalman CNPS6000cu - p.4

Swiftech MC462A - p.5

* * * * *

Discuss this article in our Forums.



Previous 1 2 3 4 5 6

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