• Home
  • blog
  • Thermalright XP-90 CPU Heatsink

Thermalright XP-90 CPU Heatsink

blog image

Thermalright’s XP-90 CPU heatsink, which features four heatpipes, thin aluminum fins and a nickel-plated copper base, was released at the same time as the XP-120. It is essentially an XP-120 shrunk to work with a 92mm fan for those who cannot accommodate a 120mm fan. Amazingly light, a little cheaper, and much easier to install: How does it perform with a quiet fan like the Nexus 92? SPCR puts the XP-90 through its paces and brings you the full monty, complete with high resolution sound files.

October 24, 2004 by Mike Chin with Ralf Hutter and Edward
Ng

Product
Thermalright
XP-90
Heat sink
Manufacturer
Thermalright
MSRP
US$39.99

Thermalright’s XP-120 recently hit the market with the sort of splash
you’d hope to never see at an Olympic diving competition. Large yet light, with five integrated heatpipes and designed for use with a 120mm fan, the XP-120 broke all previous CPU cooling records in our recent review. The smaller XP-90 was released at the
same time with less fanfare, and the new XP models have been dubbed Light ‘n’ Easy, an obvious reference to the low weight and ease of cooling offered. New HS developments are almost always tied to increases in CPU thermal output. The stimulus here was obviously the Intel Prescott core P4s, many of which far exceed 100W heat dissipation.

Thermalright broke with their own tradition with the XP series in using aluminum fins bonded to a nickel-plated
copper base. The last time they used this combination successfully was nearly three years ago with the AX478. (The more recent ALX800 was not so successful and did not use nickel plating on the base.) Heatpipes did not figure in that model, however. The last of their top models (SP94, SP97) featured all-copper construction, thin fins and heatpipes. They’ve been heavy, reaching ~600g without a fan. In contrast, the XP-90 rings in at
a mere 360 grams, yet, remarkably, this is only 10 grams shy of the XP-120’s
weight!

The change in design strategy is clear:

  • Increased cooling surface area with low airflow resistance
  • More efficient distribution of heat using multiple heatpipes
  • Copper in the base where it counts
  • Large diameter fans to allow high CFM without high RPM

These changes have the effect of increasing cooling efficiency to deal with those burning hot Prescotts without stressing the motherboard with high mass or increased fan noise. It’s a pretty smart approach.


Thermalright’s XP-90
is big but considerably smaller than…

…their
XP-120.

As the photos above show, instead of the XP-120’s five heatpipes, there are four heatpipes in the XP-90. The bottom portion of the XP-90 is very similar to the XP-120 and fits tidily into the socket 478 heatsink retention frame. The main difference is the top half of the fins: It is big enough on the XP-90 to fit a 92mm fan; the XP-120 takes a 120mm fan. Length and width are 116 x 96 mm for the XP-90 compared to 110 x 125 mm for the XP-120.

The XP-90 comes in Thermaright’s usual sturdy brown carton, whose contents includes a replacement retention frame for K8 boards that allows the standard P4 clips on the HS to be used, just as in the XP-120.

The kit includes the following:

1) La piece de resistance
2) K8 mounting bracket; for P4, uses stock bracket
3) Clips for mounting 92mm fans
4) Clips for mounting 80mm fans
5) Screws for installing K8 mounting bracket
6) Vibration dampening strips
7) Syringe of thermal interface material
8) Handy-dandy installation guide

SPECIFICATIONS

Compatibility:

AMD: Athlon64 FX 3200+ (socket 939/940) /Athlon64 3200+ (socket 754) and
above
Intel: Pentium-4 socket 478 3.2 GHz and above

Dimensions: L116 x W96 x H75 (mm) Fin only, without fan
Weight: 360g (heat sink only)
Recommended Fan: Panaflo FBL09A12M

FEATURES

  • Multiple heat pipes for exemplary heat distribution efficiency through the
    aluminum fins
  • Aluminum fins soldered to nickel-plated copper base for effective contact
    and heat transfer
  • Light weight and easy to install
  • Compatible across multiple platforms (Socket 478/754/939/940)

The base
of XP-90 is a quarter-inch thick slab of copper, very well machined and nickel-plated
to an almost mirror-like sheen. The aluminum fins appear to be directly soldered to the base. The fins have a very steep vertical rise
from the base to clear any tall mainboard
components near the socket. The heatpipes run from the base to the sides of the top portion of the fins.


Contact surface, smooth like butter.


Installed: Clearance was not an issue with the test bed.


Fan mounts via clips that notch to the frame, like
on the XP-120.

XP-90 offers two sets of clips for mounting
either an 80mm or a 92mm fan. Thermalright assures us that cooling is the same with the fan blowing up or down. The
vibration dampening strips were utilized during the test. Like the XP-120,
the clips included with XP-90 hook to the lower
flange of the fan, rather than the upper flange.
No fan was included, as is the norm for a Thermalright HS.

Even though the HS clips to the stock retention bracket
in a conventional way, it is still best to lay a tower case system on its side. It takes
a lot of force to get the clips engaged. This can be done without removing the mainboard from the case — although for K8, the stock retention bracket must be replaced, which means motherboard removal.
Start by engaging the clips under
the high portion of the fins, and then hook the clips near the heatpipes. This is much easier to do than on the XP-120, whose large fins completely block easy access to all four clips.

HS TEST PLATFORM & PROCEDURE


Ye Olde Test Setup

The P4 HS test platform is an open system not enclosed in a case.

Intel P4-2.4C Northwood core – Thermal Design Power is 66.2W; Maximum Power is 75W.
Intel 875PBZ motherboard – Intel 875P Chipset; on-die CPU thermal diode
monitoring
ATI Radeon 7500 passively cooled video card (AGP)
Mushkin PC3200 Level II – 2 x 512MB DDRAM
Seagate 80GB Barracuda IV hard drive
Seasonic SuperSilencer 300W (rev A1) PSU modded with 5V Panaflo M1A

Arctic Silver Ceramique Thermal Compound
Two-level metal platform with rubber damping feet. Motherboard on top; other
components below.
CPUBurn processor stress software
Motherboard Monitor 5.3.4.0 software to track CPU temperature and fan
speed

FANS USED FOR TESTING

Nexus 92

This is the quietest 92mm fan I have heard or measured by a significant margin. Little is known about its manufacturing origin, save that it is made in China for Nexus (in the Netherlands). While the slow speed and the smooth bearing makes for a very quiet fan, it also does not move much air, especially considering its size.

Measured Data
12V
: 1400 rpm : 25 cfm : 20 dBA/1m
9V : 900 rpm : 15 cfm : 18 dBA/1m
7V : 700 rpm : 10 cfm : 17 dBA/1m
5V : 500 rpm : 7 cfm : 15 dBA/1m

MP3: Nexus 92 @ 12V

MP3: Nexus 92 @ 9V

MP3: Nexus 92 @ 7V

MP3: Nexus 92 @ 5V

Panaflo 80L

SPCR’s reference cooling fan since our launch, the Panaflo hypro-bearing 80mm low speed fan has been used in more of our HS reviews than any other fan. Sample variances are bigger now than when we first encountered them, with Panasonic’s factory in China producing fans that seem to have more bearing noise. It remains a favorite for many reasons: Low cost and availability (in the US and Canada), quiet smooth sound, and high airflow/noise ratio.

Measured Data
12V
: 1900 rpm : 29 cfm : 24 dBA/1m
9V : 1500 rpm : 20 cfm : 19 dBA/1m
7V : 1200 rpm : 14 cfm : 18 dBA/1m
5V : 800 rpm : 9 cfm : 16 dBA/1m

MP3: Panaflo 80L on HS @ 12V

MP3: Panaflo 80L on HS @ 9V

MP3: Panaflo 80L on HS @ 7V

Panaflo 92L

The 92mm Panaflo low speed fan above is not the one used in our testing. It is the only one on hand in the SPCR lab. It was accidentally dropped during testing, causing the obvious damage above. Unfortunately, this occurred before any noise / airflow measurements were done on this bigger version of the Panaflo 80L. New samples are on their way to me now, and measurements and sound files for these will be posted here ASAP.

Suffice it to say for the time being that Panaflo 92L is a much different beast than the 80L. All the samples we have encountered are noisier, buzzier and move a lot more air. The manufacturer specs it at:


12V : 2100 rpm : 42.7 cfm : 27 dBA/1m.

SPCR MP3s: HOW TO LISTEN & COMPARE

The recordings above were made with a high resolution studio quality digital recording system. The microphone is 3″ from the edge of the fan frame at a 45° angle, facing the intake side of the fan to avoid direct wind noise. The ambient noise during all recordings is 20 dBA or lower (typically 15~16 dBA).

A quick and simple way to use these recordings for valid listening comparisons is to play the quietest recording (Panaflo 80L at 7V) on only one speaker (or headphones) and set the volume so it is just barely audible a meter away. You must also turn off any special sound effects, and set equalizer / tone controls to neutral or flat. Don’t touch the volume setting afterwards, and use the same one speaker when you listen to any of the other files; that will be reasonably close to the actual recorded sound levels.

For full details on how to calibrate your sound system to get the most valid listening comparison, please see the yellow text box entitled Listen to the Fans on page 3 of the article SPCR’s Test / Sound Lab: A Short Tour.

You may also be interested in this MP3 file, which was made with no noise sources in the recording room: MP3 – Room Ambient, 15 dBA. The measured ambient was 15 dBA. Don’t be surprised if you cannot hear any difference between this file and the quietest fan recordings.

A CAUTION about Measured Fan Data

Airflow, measured in Cubic Feet per Minute, is a difficult parameter to nail down for all kinds of technical reasons. Plus there is much more variation between different samples of the same model than any manufacturer would like to admit. For these reasons, please allow a ±10% range for the measured data above. Note, too, that the SPL measurements are increasingly questionable the farther below 20 dBA/1m we go, as the noise source gets closer to the ambient noise level.

The heatsink was cleaned and installed on the test system as per the manufacturer’s
and Arctic Silver’s instructions. Prime95 was then run for 8 hours to verify
system stability and cure the Ceramique. The system was then shut down and not
restarted until the next morning when the actual testing was done. The system
was allowed to cool between tests for 30 minutes. Each test was run for 30 minutes
even though all temperatures generally stabilized within 15 to 20 minutes.

The tests were run three times on consecutive mornings to check to
the consistency of the results. All results were within 1-2°C of each other
and the average readings are included in the charts.

Ambient temperature was measured at 71°F (21°C) over the entire series
of tests. No tests were run unless the ambient temperature was at that reference
level.

* All temperatures in degrees Celsius.
* Diode: Reading from P4-2.4C CPU diode via Motherboard Monitor.
* Temp Rise refers to the difference between ambient temperature and
the diode reading at maximum load.
* °C/W refers to the °C of temperature rise per watt of heat
dissipated by the CPU.

FANS AT 12V

Thermalright XP-90: Fan at 12 Volts
Fan
Airflow Direction
Idle
Load
°C rise
°C/W MP
°C/W TDP
Panaflo 92mm L1A
up
26°C
37°C
16
0.21
0.24
down
23°C
35°C
14
0.19
0.21
Nexus 92mm
up
26°C
42°C
21
0.28
0.32
down
26°C
43°C
22
0.29
0.33
Panaflo 80mm L1A
down
27°C
42°C
21
0.28
0.32

The following notes apply to all the temperature tables:
°C rise
refers to the rise in
temperature over the ambient at load.
°C/W – TDP calculations:
Intel’s TDP of 66W was used.
°C/WMP calculations: CPU
Heat& CPUMSR Projects’
estimate of 75W was used.

With the Panaflo 92L fan, the cooling performance of the XP-90 is in champion territory. It doesn’t quite match the performance of the larger XP-120 with a larger 120mm fan, but it comes so very close. That it manages to do this with a fan that is so much smaller is a clear testament to the cooling efficiency of the design. However, the noise of the Panaflo 92L at 12V leaves much to be desired; it is hardly what most SPCR enthusiasts would accept for their own PCs.

Cooling performance with the slower and quieter Nexus 92mm and Panaflo 80L fans is still very good. The performance is virtually identical despite the different sizes. The only explanation we can think of is that the larger diameter of the Nexus 92mm fan is offset by its lower CFM. Note that the measured CFM on the Panaflo 80L is significantly higher than for the Nexus 92, and higher than what Panasonic cites. It is also 4 dBA/1m louder than the Nexus 92 at this voltage.

Performance in our test platform with the fans blowing down is slightly better than blowing up, contrary to Thermalright’s claim that it is the same.

The 20 dBA/1m of the Nexus 92 may be quiet enough for many people, but naturally SPCR reader want even quieter. And this cooler has plenty of cooling headroom to reduce the fan speed and noise.

FANS AT 7V

Thermalright XP-90: Fan at 7 Volts
Fan
airflow direction
idle
load
°C rise
°C/W MP
°C/W TDP
Panaflo 92mm L1A
up
26°C
41°C
20
0.27
0.30
down
26°C
40°C
19
0.25
0.29
Nexus 92mm
up
27°C
50°C
29
0.39
0.44
down
26°C
46°C
25
0.33
0.38
Panaflo 80mm L1A
down
26°C
47°C
26
0.35
0.40

Performance at 7V remains excellent with the Panaflo 92L, which is actually still better than with the other two fans at 12V. Its noise, though greatly reduced, is still somewhat more intrusive than most SPCR enthusiasts would like, with some clicking and buzzing issues. Both of the other fans will be essentially inaudible inside a typical quiet PC, and they still provide quite good cooling.

FANS AT 5V

Thermalright XP-90: Fan at 5 Volts
Fan
airflow direction
idle
load
°C rise
°C/W MP
°C/W TDP
Panaflo 92mm L1A
up
26°C
44°C
23
0.31
0.35
down
27°C
42°C
21
0.28
0.32
Nexus 92mm
up
27°C
58°C
37
0.49
0.51
down
26°C
52°C
31
0.41
0.50
Panaflo 80mm L1A
down
27°C
52°C
31
0.41
0.50

The Panaflo 92L is still going strong at this voltage, but again the noise is not as good as either the Nexus 92 or Panaflo 80L even at 7V. The Nexus 92 and Panaflo 80L are at about the borderline of thermal prudence at this voltage. Inside a case with ambient temps at least 10°C higher, the CPU temp would be in the low 60s. This is still safe, but it does not leave as much headroom for hot weather or other extreme conditions. Still, it is the best performance we’ve encountered with any heatsink running the Panaflo 80L at 5V.

Finally, a quick comparison against a couple of obvious competitors at similar noise levels.

Thermalright XP-90 Vs. Competitors
HSF
airflow direction
idle
load
°C rise
°C/W MP
°C/W TDP
XP-90 + Panaflo 92L @7V
down
26°C
40°C
19
0.25
0.29
XP-90 + Nexus 92 @12V
down
26°C
43°C
22
0.29
0.33
*Thermalright SP94 + Panaflo 80L @ 9V
down
0.39
**Zalman 7000AlCu @ 5V
down
26°C
45°C
23
0.31
0.35
* The SP4 was tested on a different platform under different conditions; however the °C/W figure should be a valid basis for comparison.
** The Zalman 7000 @ 5V has about the same noise level as a Nexus 92L @ 12V and a Panaflo 80L @ 9V. MP3 of Zalman 7000cu at 5V

CONCLUSIONS

Thermalright has a winner in the XP-90. The only HS that can clearly best it is Thermalright’s own much larger XP-120 with a much larger fan. Yet it will probably fit nicely without problems on the vast majority of motherboards in conventional ATX cases, and it even works well with a Panaflo 80L fan at low speed. It does all this without any serious weight penalty; with a 92mm fan, it should easily fall within the 450 grams maximum HSF mass recommended by Intel.

What’s not to like? About the only thing we can think of is the high pressure on the clips and maybe the absence of a quiet fan packaged with the XP-90 at an attractive combo price. Plus, if your system can accommodate the XP-120, it would still be the preferred performance choice. But these are trivialities we have to scratch around for so as to look not overly enthusiastic. With the XP series, Thermalright has raised the bar for heatpipe-equipped CPU coolers.

Many thanks to JAB-tech and Thermalright for the XP-90 samples.

* * *

Discuss this article in the SPCR Forum.

Leave a Comment

Your email address will not be published. Required fields are marked *