Review: Heavyweight P4 Heatsink Roundup

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

TEST RESULTS

Key Components in P4 HS Test platform

The P4 HS test platform is an open system not enclosed in a case. The test platform and procedure is described in detail in our HS Test Methodology article.

  • Intel P4-1.8A Northwood core - Maximum power is ~50W. 67° C rated maximum junction temp. We'll pull the plug if any HS lets the temp go much above 60° C. But truth be known, I've pushed the P4-1.8 on the test platform all the way to 80° C without harm or forced shutdown.
  • Intel D845PEBT2 motherboard - Intel 845PE Chipset; on-die CPU thermal diode monitoring
  • Panaflo FBA08A12L1A 80mm DC fan
  • Any VGA card (AGP)
  • 256 MB DDRAM - PC2100 generic
  • Any hard drive (in Smart Drive from Silicon Acoustics)
  • DigiDoc5 w/ thermal sensors
  • Any Good PSU
  • Zalman Multi-Connector (ZM-MC1)
  • Arctic Silver 3 Thermal Compound
  • Two-level plywood platform with foam damping feet. Motherboard on top; most other components below. Eases heatsink changes and setup.
  • CPUBurn processor stress software
  • Intel Active Monitor software to show CPU temperature

The reference Panaflo fan was placed on top of each heatsink, blowing down, and tests run on each HS at 12, 7 and 5 volts feed to the fan. Panaflo fan summary:

Panaflo FBA08A12L1A Speed, Airflow and Noise
VDC
RPM
dBA @ 1cm
dBA @ 1m*
CFM
Comment
12V
1900
48~50
21 (manufacturer's spec)
24
Plainly audible from several feet away. Some low frequency vibration can be heard and felt; direct mounting to a chassis accentuates the low frequency noise. Most audible aspect of sound is centered around 5-10 KHz; whine and whoosh are best descriptors. Much quieter than most commonly used fans.
7V

1235

 

35~38
9 (calculated)
14 (calculated)
All aspects of noise greatly diminished. From a few feet away, it is inaudible to many people. Qualitatively, it is a hum rather than a whoosh or whine. No high pitched noise at all. Enough vibration remains that direct chassis mounting produces a bit of low frequency accentuation.
5V
760
32~35
2 (calculated)
10 (calculated)
Essentially inaudible in free air except from under one foot distance. Someone with higher hearing sensitivity might hear it from 2 feet, but it would only be a whisper, or rather a very quiet buzz. The vibration level is low enough that direct chassis mounting produces no significant boost in low frequency noise.

*dBA @ 1 meter sound pressure level is the most common industrial measurement for noise. The Panaflo is rated by the manufacturer at 21 dBA. Sound level meter systems capable of measuring accurately below 20 dBA are priced well over US$10,000. Furthermore, the ambient noise level must be lower than the measured source by at least several dBA. These factors make low noise measurments at one meter at the lower voltages is extremely challenging. Thus our fallback to the nearfield measurement technique, 1 cm from edge of fan frame, not in the airflow path.

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 domestic room fan, the airflow directed over the entire test setup. The ambient temperature was 20C throughout.

The stock Intel HS was tested in 2 configurations --

  1. as is
  2. modified for use with the Panaflo fan.

I wondered whether the use of a 80 mm fan instead of the standard 92 mm size supplied with the Zalman 6500 heatsinks might hurt their performance, because the fin area of the 6500s is so spread out. But the standard Zalman fan is too loud even at 5V, perhaps more annoying than the Panaflo even at 12V. A 92mm Silencer fan (from PC Power & Cooling), rated for 23 dBA and 28 CFM, was substituted for the Panaflo on the 6500s at 12V and at 7V. It gave results identical to the Panaflo..

Note that 20+ minutes of CPUBurn at high priority is far more stress than most user would subject their system to in normal circumstances.

  • All temperatures in degrees Celsius.
  • Diode: reading from P4-1.8 CPU diode via Intel Hardware Monitor.
  • Temp Rise refers to the difference between ambient temperature and the diode reading. The ambient temperature during all testing: 20° Celsius., measured with DigiDoc thermistor ~6 inches above CPU HS and fan.
  • ° C/W refers to the number of degrees of temperature rise per watt of heat dissipated by the CPU. Intel rates its P4-1.8A at 49.6W. With speed throttling and other tricks, the actual power dissipation of the P4 is actually something of a mystery, so for now, we'll round off that number to 50W. If the power dissipation is correct, ° C/W can be used to accurately predict the performance of the HS with processors of different wattage.

    For example, if you use the 6500Cu with Panaflo at 12V on a P4-2.8 rated for 68W:

  • 0.28° C/W x 68W = 19° C

So 19° C is the rise over ambient temperature you can expect under high stress.

  • To estimate temperatures inside a PC case, add ~10° C to the room temperature. The case temp rise can be less or more; this varies on so many factors that it is difficult to be more specific.

Results at 12V

Model
° C - Idle, 10 mins
° C - CPUBurn, 20+ mins
° C/W
Diode
Temp rise
Diode
Temp rise
Intel HSF (52 dBA)
26
6
43
23
0.46
Intel + Panaflo
26
6
42
22
0.44
6500AlCu
24
4
37
17
0.34
6500Cu
23
3
34
14
0.28*
AX478
24
4
36
16
0.32
OCZ Eliminator
24
4
39
19
0.38

Intel thermal control fan: At what temp?

Puzzled by the constant speed of the Intel fan, I took extreme measures to explore. I turned up the heat by bending a gooseneck lamp with a 100W incandescent bulb over the HSF.

A thermal sensor was inserted into the center of the HS fins near the fan as shown above, and a second sensor placed beneath the HS, touching the edge of the CPU heat spreader. The relationship between the temperatures of the CPU diode, the heat spreader edge and the top of the HS fins is as roughly follows:

  1. °C CPU diode = reference temp
  2. °C at heat spreader edge = reference temp + 3°C
  3. °C at top of fins = reference temp + 7°C

CPU temp at 100% utilization was 42° C. With the lamp, CPU temp rose to 47° C before the fan finally began to speed up. It hit 3700 at 50° C, then stabilized with the bulb almost touching the fan center. At that point, the fan noise measured 59 dBA.

Inside a PC case where the temperature easily can be 10° C higher than the test room ambient, this fan could become a real screamer at 12V compared to a Panaflo, but it's still relatively mild compared to... well, lots of fans popular with overclockers.

The Intel HSF does a pretty good job of keeping the P4-1.8 cool. Surprisingly, its thermal fan remained at 3100 RPM throughout the test.

The fan noise measures about the same dBA as the Panaflo at 12V (measured at 1"), but is definitely more audible and annoying. The higher speed of the Intel fan, which has a smaller blade diameter (~60mm), gives it a higher pitched whine.

With the Panaflo at 12V, the Intel HS turns in slightly better performance as with the stock fan, but the noise is easier on the ear.

The results for the Zalman 6500AlCu are as expected: the ° C/W is 0.02 worse than Zalman's claim with their stock fan, which provides 50% more airflow than the Panaflo. It is very good performance.

The performance of the Zalman 6500Cu at 12V is the best of the group. Given its mere 14° C rise over ambient under long term CPU stress, this HSF combination should safely cool the P4-1.8 in a real world system even at nearly 40° C room ambient.

*Note that Zalman specifies 0.28 ° C/W for the 6500CU with their 92mm fan, rated for 36 CFM, 50% higher than the Panaflo. The Panaflo should not give the same ° C/W result. Practical experience suggests to me that with the Panaflo, the ° C/W of the 6500Cu should be around 0.02 higher than with the stock Zalman fan. It's difficult to imagine that any company in this competitive marketplace would deliberately underrate their product, so I am a bit dubious about our result. Yet, the test came out the same on 3 separate occasions, and the result for the 6500AlCu is exactly where it should be, so the test platform does not seem the culprit here.

The Thermaright AX478 also turns in an impressive performance, coming in between the two Zalmans. The actual advantage over the 6500AlCu is so small that you might as well consider them equal at this airflow level.

The Eliminator came in 4th place behind the 6500AlCu, possibly handicapped by the shortness of its fins. It still represents good performance, considerably better than the Intel P4 HS with Panaflo or stock fan.

Results at 7V

Model
° C - Idle, 10 mins
° C - CPUBurn, 20+ mins
° C/W
Diode
Temp rise
Diode
Temp rise
Intel HSF (42 dBA)
27
7
47
27
0.54
Intel + Panaflo
28
8
49
29
0.58
6500AlCu
26
6
43
23
0.46
6500Cu
25
5
39
19
0.38
AX478
25
5
42
22
0.44
OCZ Eliminator
25
4
46
26
0.52

At 7V, the Intel HSF still does very well. The fan spins at 1570 RPM, and the noise level is within a dBA or two of the Panaflo. Quiet, but still with higher pitched hummy / whiny quality than the Panaflo; this makes it much more audible, at least to me (and two other listeners). The stock fan actually beats the Panaflo by 2° C. Hmm. Interesting.

6500AlCu still cools admirably. It would be perfectly safe to run the P4-1.8 in a PC case with any case temperature up to ~44C. But the margin between it and the stock Intel HSF has dropped a degree.

The 6500Cu remains clearly on top by a somewhat larger margin than at 12V. Impressive performance. Again, the AX478 falls between the Zalmans. This time it is closer to the AlCu than the Cu. The Eliminator can only just better the stock Intel HSF at this airflow despite its higher mass and all-copper construction.

Results at 5V

Model
° C - Idle, 10 mins
° C - CPUBurn, 20+ mins
° C/W
Diode
Temp rise
Diode
Temp rise
Intel HSF (40 dBA)
28
8
49
29
0.58
Intel + Panaflo
29
9
58
38
0.76
6500AlCu
27
7
49
29
0.58
6500Cu
27
7
47
27
0.54
AX478
27
7
49
29
0.58
OCZ Eliminator
28
8
56
36
0.72

The stock Intel HSF is a real surprise here. With its tiny fan spinning at 1360 RPM, I was betting CPU temperature would soar, but no, it's temperature went up only by 2° C, less than any of the others in the step between 7V and 5V. It matched the performance of the Zalman 6500AlCu and the AX478 for a 3-way tie! The Intel HS with the Panaflo was the one that soared instead, climbing all the up to 58° C, along wth the Eliminator, which could only manage 56° C.

I have to admit I was so surprised at the result that I spent way too much time repeating the tests with pretty much identical results.

Even with the fan running at 5V, all of these heatsinks except for the Intel + Panaflo and the Eliminator are still suitable for use with the P4-1.8A in typical systems in reasonable temperature (with case temperature up to ~36° C.)

A final test was run on the Eliminator with no fan at all. Idle temp was fine at 32° C, but within 10 minutes at 100% CPU utilization, the temp climbed to 80° C before I stopped the test. It is not really suitable for fanless operation except in a very well cooled case, but that is likely true of all of these heatsinks.

RANKINGS

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

#1 - Zalman 6500Cu-B: Its well-executed, oversize, open-fin all-copper design provides great cooling performance with very little airflow. A 92mm or 120mm fan that has more airflow than the Panaflo (but equal or less noise) would be a nice match for this HS. The only possible caveat is the weight. A kilogram is a lot of weight hanging off a vertically mounted motherboard. One wonders whether the nice plastic clips will apply enough tension on such a mass, especially over a long period of time. Gravity is pretty constant. Those seeking higher security will probably be happier with the 6500Cu-A model, which has through-the-motherboard bolt mounting. It would be nice if they offered a version without fan -- for those who want to use a different fan.

#2 - Thermalright AX478: Like its twin AX7 in SPCR's first HS roundup, the AX478 squeaks into second place. The efficiency and low cost of this model is admirable. Very nice finish, design and performance -- plus you pay only for the HS and use any 80mm fan of your choice.

#3 - Zalman 6500AlCu-B: Only marginally bettered by the AX478, the aluminum-copper hybrid is also a good performer. Its low 400 gram weight is within Intel's maximum weight, which eliminates any concern about the clip. One wonders if the number of copper fins couldn't be increased (at least to Intel's 450 gram limit) to improve performance without adding cost significantly. Perhaps that would make the Cu a harder sell... Again, it would be nice if they offered a version without fan at a lower price.

#4 - OCZ Eliminator: Even hampered by its short fins, the OCZ does fairly well at 12V and 7V, providing slightly better performance than the stock P4 HSF. But even with its fine build quality and the benefit of the back support plate for the motherboard, it is difficult to recommend this heatsink at the asking price, which BTW, is the only one I could find on the web. At a discounted price, it would be worth considering, especially for machines that must with the rigors of travel or shipping.

#5 - Stock Intel: Intel's P4 HSF is a sleeper in quiet mode, providing good cooling performance even down to 5V, where it mostly keeps up with the big boys and becomes virtually inaudible as well. For a HS that comes "free" with the CPU, it is outstanding value. Spend $6-8 for a Zalman fanmate and play to your heart's content.

The decreased performance with the Panaflo fan at 7V and 5V suggests several things:

  1. Despite its smaller size, the Intel fan is actually moving more air at those voltages. The fan speed only dropped by 200 RPM between 7V and 5V, which means it is pushing almost as much air at 5V as at 7V.
  2. The big rise in temp with the Panaflo at 5V (compared with the others) shows that the heatsink's intrinsic cooling power without its integrated fan is no match for the aftermarket competition here.
  3. The HS and fan are highly optimized to work together symbiotically, with the fan concentrating all of its airflow where it counts most.
  4. The super pressure between CPU and stock HSF is a real benefit; perhaps it forces better contact between heat spreader and core, minimizing the effect of that extra thermal resistance.

FINAL WORDS

In many way, there are no losers here, only winners. With a Panaflo at 12V, both 6500 models and the AX478 should be able to tackle even the hottest P4s adequately. At the quiet 7V level, all five are still serious performers.

If we accept Intel's 67° C thermal limit and assume that the maximum case temperature should be 35° C, then we can predict the maximum CPU heat dissipation in quiet operation with a little simple math:

  • Maximum allowable heat rise = 67°C - 35 C = 32°C
  • So, °C/W x Max CPU Watts = 32°C
  • Solve for CPU Watts = 32° C divided by °C/W

CAUTION: Please note that these calculations are just that; real life never turns out quite as you calculate. Actual results will be affected by a host of other factors specific to each PC system. Also, thermal behavior is not always linear, so the °C/W with heat sources much different from the P4-1.8A may change. There is evidence to suggest that the 533 MHz P4s have a steeper thermal curve than the 400 MHz processors.

A. Anticipated maximum safe CPU heat dissipation with 7V* (very quiet)

HS
°C/W
Max CPU W
Fastest CPU**
Intel
0.54
59
P4-2.26G
6500AlCu
0.46
70
P4-2.8G
6500Cu
0.38
84
P4-3.06G
AX478
0.44
73
P4-2.8G
OCZ Eliminator
0.52
61
P4-2.53G

Remember that the Intel fan whines more at this voltage than the Panaflo. All the HS can ramp up fairly high, but at this voltage even the lowest performer can take on a P4-2.26G, and the Zalman 6500Cu can handle a P4-3.06G. According to calculations.

B. Anticipated maximum safe CPU heat dissipation with 5V* (virtually silent)

HS
°C/W
Max CPU W
Fastest CPU**
Intel
0.58
55
P4-2.2G
6500AlCu
0.58
55
P4-2.2G
6500Cu
0.54
59
P4-2.26G
AX478
0.58
55
P4-2.2G
OCZ Eliminator
0.72
44
P4-1.6A

* Intel with HS its own fan; all other HS with Panaflo 80mm "L". ** Based on Intel's documentation on P4 thermals.

Naturally the performance of every heasink drops, but remember that at this voltage level these heatsinks will make no noise that will be audible outside a decent PC case.

Now, one other factor that comes into play here is that Intel recommended maximum die temp changes from P4 to P4. Some 2.8 variants, for example, are rated for a maximum of 75° C, while the 2.53 is 71-72° C. The 3.08 drops again to 69° C. Have fun with your calculators!

The point is that these HS will ramp up to different performance levels depending on airflow. It's up to you to choose which performance / airflow / noise level is right for you. If you seek the highest cooling potential, then the 6500CU is the obvious choice, but if your Siberian basement office requires that you wear a down coat to work in all year round, the stock Intel HSF at 5V will probably do fine even for >3G processors.

Much thanks to Zalman USA, Thermalright and OCZ Technology for the review samples.

< Intro and Intel P4 HSF

< Zalman 6500B-Cu and 6500B-AlCu

< Thermalright AX478

< OCZ Technology Eliminator

* * * * *

Discuss this article in our Forums.



Previous 1 2 3 4 5

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