Even at first glance you can tell the Zalman Reserator1 is something completely different on the PC cooling landscape. Zalman’s first water-cooling system is the first fanless kit to be widely available commercially. SPCR reviewer Russ Kinder says it’s almost totally silent. Click here for his long-awaited review of the Zalman Reserator1.
Aug 3, 2004 by Russ
Water cooling system
|$249.95 at Sharkacorp (who kindly provided the review sample)|
Image courtesy of Zalman
From the first time you see it, it is obvious that the Zalman Reserator1 represents
something completely different on the PC cooling landscape. The first impression
is accurate: The Reserator1 is completely different:
For one, it is Zalman’s first complete water-cooling system, previously they’ve
sold their waterblocks only as individual components.
Secondly, it is the
first totally passive water-cooling kit to be widely available commercially.
Thirdly, its almost totally silent.
The concept is elegant in its simplicity: Use water as a medium for moving
the heat of the CPU a much further distance than what is possible with a conventional
air-cooler. Once freed of the confines of the area directly around the CPU,
the options for then transferring the heat energy from the water to the air
are greatly expanded.
Conventional air-coolers, as well as most water-cooled systems, use a fan to
force air across the heatsink/radiator fins to accomplish this transfer. The
Reserator1 replaces the forced air movement with natural convection. Convective
airflow is nowhere near as efficient as forced air; the volume of air passing
over any given part of the heatsink is simply much lower. Zalman balances the
reduced efficiency by greatly increasing the available surface area for heat
transfer, and by configuring the shape to take maximum advantage of the convective
The system consists of two essential primary components: The radiator/reservoir/pump
unit, and the CPU waterblock, and comes complete with all the assorted accessories
Image courtesy of Zalman
We’ll start with the small bits:
Image courtesy of Zalman
When up and running, the activity in the flow tube is the
only indication that the user has that the pump is running, so its inclusion
in the kit was a smart choice. There have been some reports of flow tube
cracking during assembly or use, but our test sample performed as expected,
with no troubles. At the time of publishing, Zalman is offering a recall
on the plastic flow tubes, and is exchanging them free of charge for one
with stainless steel fittings.
The depth of the accessories package is further indication
of the Reserator1 being targeted towards the novice watercooler, rather
than the expert. The manual that is included is very thorough and well written;
it could walk someone completely new to water-cooling through the entire
assembly process, from start to finish.
Now on to the big parts of the kit.
The true heart of the entire system, the Reserator combines the pump, the
radiator, and a reservoir into a single unit. It is essentially a deeply-finned
water-filled tube of extruded aluminum, with a pump submerged at its base.
All of the aluminum has been anodized a deep, iridescent blue.
Reserator is a compound word derived from ‘Reservoir’ and ‘Radiator’
– it acts as a reservoir while radiating heat. This product works
well with natural convection and integrates a water pump inside
Radiation Area 1.274 m2 Weight 6.5 kg Dimensions 150 (dia.) x 592 (h) mm Material Anodized Aluminum Coolant Capacity 2.5L max. Water Pump 5W, Qmax 300L/h Maximum lift 0.5m
As a design element, this thing is, dare I say, pretty. Obviously aesthetics
are probably not top on the top of the list of requirements while selecting
PC cooling components, but there is something to be said for design quality.
And since you’ll be looking at this unit for the entire time it’s in use,
having it look good is at least a nice fringe benefit.
A good angle for understanding the finned nature of the Reserator. There
are 44 fins in all, each 2.5cm deep, and tapering from 2mm’s thick at their
base to 1mm at their tips.
Image courtesy of Zalman
Moving down to the base, you can see the water line connections, labeled to
Disassembling the unit reveals the pump. The pump is attached to the base
of the Reserator by means of a pair of stainless steel screws and strap.
Instructions are included in the manual for replacing the pump with an external
one, should the user desire.
As water-cooling pumps go, the compact Eheim unit mounted inside the Reserator
is a relative lightweight. A comparison of the 80gph and 18â head of the
Reserator’s Type 300 pump to the commonly used Type 1048 pump’s 158gph and 59â
head is a pretty powerful indicator that the Reserator is going to be a low
flow system, in any configuration.
Much is made of flow-rate in the world of water-cooling. Put
simply, it is the water-cooling equivalent of air-cooling’s airflow, measured
in Cubic Feet per Minute. Like CFM, flow-rate is an important indicator
of cooling performance, but it is not the only indicator. As we’ve seen with
conventional heatsink testing, designs can be optimized to perform under low
flow. How well the Reserator1 performs will largely depend on how well Zalman’s
engineers designed the system to accommodate the flow rate dictated by their
choice in pumps.
GOLD CPU WATERBLOCK
64mm Diameter x 31mm Tall
Gold plated copper & anodized aluminum
$49.95 at Sharkacorp
(if purchased separately from the Reserator1 kit)
Image courtesy of Zalman
The color coordination continues to the waterblock. Zalman includes their ZM-WB2 Gold waterblock as the other
primary component in the system.
More information from Zalman:
1. Pure copper base ensures excellent heat dissipation, and anodized
aluminum top prevents corrosion.
2. Intel Pentium 4 (Socket 478), AMD Athlon/Duron/Athlon XP (Socket
462), Athlon 64 (Socket 754) compatible design for board compatibility.
3. Three types of compression fitting are offered for use with
1) CPU block- 1 pc.
2) Hand screw – 2 pcs.
3) Thermal grease – 1 PC
4) Clip-1 PC
5) User’s manual
6) Fittings-6 PCs (3 sets) *
Components for Intel Pentium 4 (Socket 478):
7) Two (2) Clip Supports for Socket 478
Components for AMD Athlon / Athlon XP (Socket 462):
8) One (1) A-Type(blue) Clip Support for AMD Socket 462
9) One (1) B-Type(white) Clip Support for AMD Socket 462
10) Four Bolts – For Fastening Clip Supports
11) One Set of Washers
Components for AMD Athlon 64 (Socket 754):
12) Two Nipples
13) One Backplate
When included in the Reserator system, the ZM-WB2 includes only 1 set of
fittings. Probably a good thing, having only the fittings that fit the included
tubing will certainly avoid confusion during assembly.
The ZM-WB2 employs classic old-school waterblock technology; no jets,
no slots, no fancy impingements, just a good sized chunk of copper shaped to
allow maximum surface area contact between the block and the water flowing through
it, all encased in an aluminum shell with a single inlet and outlet. All in
all, a design that should perform well in a low flow situation.
Image courtesy of Zalman
The base surface is mirror-like, probably the best CPU mating surface I’ve
ever seen, in terms of finish at least. (Yes, it is even better than the famous
Swiftechs) It should be; unlike typical heatsinks where the mating surface is
polished bare metal, the ZM-WB2’s base is gold plated. Any imperfections in
the copper below would be hidden by the plating process. Zalman’s reasoning
for breaking out the precious metals goes beyond vanity. By plating the copper
with gold they alleviate some of the corrosion risk inherent with having mixed
metals in a water-cooling loop. The plated copper with the anodized aluminum
should greatly reduce the corrosion potential.
Zalman goes so far as to specifically recommend against adding any corrosion
inhibiting additives to the distilled water of the loop. This seems like
a bit of wishful thinking. In reality there will be exposed copper and aluminum
in the system, even if in tiny amounts, and the plating will do nothing
to stop the inevitable growth of living things inside the system. A bottle
of corrosion inhibitor/algaecide is cheap insurance.
As listed in the spec’s, the ZM-WB2 has excellent CPU compatibility: Socket
478, 754 (and by thus 940 and 939 as well), and Socket A. The only catch for
Socket A is the requirement of the 4 socket mounting holes. Motherboard compatibility
should be nearly universal since the waterblock is actually smaller than a stock
For testing the ZM-WB2 was installed on a Socket A board. Installation was
straightforward and simple. The retention mechanism uses a center pressure point
from the spring clamp, which proved to be a double-edged sword: On one hand,
its a foolproof way to ensure that the force is applied only to the center of
the CPU die, a real consideration when installed on XP’s. But the single center
point, combined with the round shape of the ZM-WB2, allows the entire block
to spin around that center axis. There is no chance of it coming loose, as it
will only rotate 30Â° or so until stopped by the fitting, but it did raise
fears of a smeared TIM application.
ZALMAN ZM-GWB1 VGA WATERBLOCK
Included in our testing is the new ZM-GWB1 VGA waterblock.
The VGA cooler is available as an optional add-on to the Reserator1 kit,
or as a stand-alone item to be combined with non-Reserator water-cooling
With the huge recent increases in the heat production coming from GPU’s, the
same logic that supports water-cooling the CPU also applies to the GPU. The
Wattage of the newest VGA cards, such as the 100+ Watt Nvidia 6800 Ultra’s,
actually exceeds that of the majority of CPUs found in users PCs. If it makes
sense to move the CPU’s heat outside the case, then the same should be true
of the equally-hot GPU.
More info from Zalman:
1. Pure aluminum base ensures optimal weight and heat dissipation.
2. Anodized product surface prevents corrosion.
3. Two types of water blocks provide support for virtually all
types of ATI and NVIDIA VGA card. This product can be reused with
upgraded video cards. (Note: This product is not compatible with
VGA cards that do not have holes around the VGA chip)
4. RAM heatsinks provide effective heat dissipation.
1. Small – 1EA, Large Type – 1EA
2. RAM Heatsink – 8EA
3. Tube Clamp
4. Thermal grease
5. etc (O-ring – 4EA, Nipple – 2EA, Bolt & Nut – 1EA)
Dimensions Small Block 35.6 x 66 x 27.5 mm Large Block 35.6 x 76 x 27.5 mm Weight Small Block 60g Large Block 75g Waterblock Material anodized aluminum Ram-sinks anodized aluminum 8pcs, 5g/p
VGA WATER BLOCK DESIGN
Apparently Zalman got the bulk rate on the blue anodizing.
From a design standpoint, the ZM-GWB1 is about as elegantly simple as a
waterblock could be. It is an anodized aluminum block milled straight through
with a single 8mm hole. The design seems an excellent choice for inclusion
in a low-flow system like the Reserator1. The straight, unobstructed water
path is produces the least amount of flow restriction possible. To the water
flowing through the system, the ZM-GWB1 is virtually the same as a few extra
inches of tubing.
Note 1″ EAR fan mount atop ZM-GWB1 base.
The mating surface is relatively smooth, on par with Zalman’s heatpipe
Aluminum ramsinks are included with the
8 non-blue-anodized aluminum ramsinks are included with the
VGA cooler. They come with pre-applied thermal double-sided tape, and installation
to the VGA RAM is a simple peel-and-stick affair.
Installation of the waterblock is simple for anyone familiar with Zalman’s
heatpipe VGA coolers and northbridge coolers, as it uses exactly the same mounting
system. As with the heatpipe coolers, two different sizes of component are included
to ensure compatibility with both Ati and Nvidia cards. That’s a particularly
nice touch, since it allows easy future upgrading from one card maker to the
other, without worries about whether you’ll have to buy a new waterblock to
fit your new card. (And the extra waterblock could probably be pressed into
service as a northbridge cooler; the mounting system and general size are the
For testing the ZM-GWB1 was installed on a Sapphire 9500 VGA card. While
no longer anywhere near the top of the AGP heat or performance heap, it
does have a couple of points in its favor for our testing purposes:
Installation onto the 9500 went smoothly: Just matter of selecting the
proper block, assembling the washers and thumbscrews, and sticking on the
ram-sinks. Overall, a much simpler process than installing Zalman’s heatpipe coolers (HP series).
The uniqueness of the Reserator1 required the development of a testing methodology
that is unique as well. The resulting testing procedure is notable as much for
what it does not test, as for what does: Simply put, the Reserator1 is designed,
sold, and intended for use primarily as an integrated system. For that reason,
it will be tested as a system, with only limited amounts of attention paid to
the performance of individual components.
For the tests, the Reserator1 was assembled in the same manner, with all the
same components, as it would likely be used in reality. The goal with the
testbed is to recreate the same conditions that the unit will be asked to perform
under, not to necessarily produce the best possible scores.
The test loop consisted of:
For previous Socket A reviews we’ve relied upon processors with <70 Watts of max heat. While perfectly acceptable at the time, the pace of CPU heat output has now passed them by. With current CPU's now breaking the 100 Watt barrier, it was time for a testbed upgrade.
Several criteria were established for the Socket-A Torture Testbed:
The first two conditions were met by some careful rummaging through the
spare parts bin. A combination of an Abit KR7A-133 motherboard, and an unlocked XP-2100
T-bred was selected to serve as the primary components. The motherboard
was socket-modded to over-volt the CPU to 1.85v. Using the FSB and Multiplier
ranges available with that motherboard/CPU combo allows a range of dissipated power
from as low as 62.2 Watts at 1300Mhz, to as high as 105.3 Watts at 2200Mhz, as calculated with CPU Power by Kostik, a software utility that you can find in the SPCR Software Downloads section.
The third condition, accurate temperature readings, was achieved by bypassing
the stock in-socket thermsistor and reading the temperature directly from the
CPU die with an external reader soldered to the CPU pins. This device reports
its readings through the SMBus.
The Maxim MAX6657 thermal monitor chip in action.
Once assembled, the temperature monitor and motherboard/CPU
combo were run through a laborious series of temperature testing to calibrate
the sensor’s readings.
The other testbed components are stock components from around
- Seagate 7200.7 HDD
- 512mb of PC2100 RAM
- Fortron FSP300-60PN âAuroraâ PSU
- Windows XP Pro/SP1
- Kill-a-Watt AC meter for accurate readings of the total AC draw of
To isolate the effects of adding the VGA cooler, the water-cooled 9500
was replaced by a passively cooled SiS 4mb AGP for some of the tests. For
these tests, the 9500 remained in the water-cooling loop, but was not powered.
Look Ma’, no heatsink! Behold the power of 1x AGP/4mb graphics.
- Each test was conducted multiple times, and the average temperature
- Ambient temperature was 23Â°C, unless noted otherwise.
- CPU Load temps were achieved by running CPUBurn for a minimum of 12
hours. (The extended time run was required to have the temperatures
- 3DMark03 was run as a continuous loop to achieve VGA load.
- Idle temps were recorded at the Windows desktop.
- Motherboard monitor was used to record the temperature readings from
the Maxim MAX6657, via the SMBus.
TEST 1: XP2100 at stock, no VGA in cooling loop.
XP2100 @ 1733Mhz, 1.60 volts: 62.1W
System AC Draw
Ancient SiS VGA card used for this test.
This first test was conducted to provide a direct comparison with the previous
SPCR Socket A tests.
If compared to the best air-coolers tested on the basis of temperature,
the Reserator’s results are quite respectable. The 0.37Â°C/W thermal
conductivity is on par with a ThermalRight SP97/L1A combination @ 10 volts,
or a Zalman 7000a @ 5 volts. That’s lofty company to be in league with.
But if the noise level is factored in, the Reserator is simply in a league
of its own.
Acoustically, the Reserator1 is about as close to the holy grail of silent
cooling as is likely possible. The pump, dampened by 2.5 liters of water and
7kg of aluminum, is virtually noiseless. About the only noise from the pump
is a low frequency vibration transferred through the shell of the Reserator.
It is more felt than heard. Setting the Reserator on a hard floor surface exaggerates
the vibration: on carpet it is undetectable until you place your hand on top
of the unit. The biggest noise producer of the entire system is actually the
flow indicator. It produces a slight clicking noise as it rattles around inside
its plastic chamber. Not an obtrusive noise, and you’d have to have very low
ambient noise levels to even notice it.
TEST 2: XP2100 at 2.2 Ghz & 1.85 volts, no VGA in cooling loop.
XP2100 @ 2200Mhz, 1.85 volts = 105.3W
System AC Draw
Ancient SiS VGA card used for this test.
Now we really turn up the heat. 105 Watts of heat puts this CPU up into
the same egg-frying category as the P4EE’s and Prescotts, and substantially
beyond the hottest of any of the AMD CPUs available currently. 12+ hours continuous
CPUBurn is far beyond the sort of stress any typical user will put the system
under. As previous tests have shown, running a CPU loading application such
as Folding at Home will not produce the same levels of CPU heat production
as CPUBurn. Under such conditions, 63Â°C is an impressive result. It’s safe to say that the Reserator1 has the
cooling power to be used with any current CPU available today. (At
least at stock speeds)
TEST 3: XP2100 at 2200Mhz & 1.85volts, with ATI 9500 in cooling loop.
XP2100 @ 2200Mhz, 1.85 volts + Ati 9500
System AC Draw
These results require some interpretation. Comparing the idle temps
first, we see the immediate impact of adding the 9500 to the water-cooling
loop. The idle temp jumps by 4Â°, and AC draw is increased by 25 Watts,
just from adding the idling 9500. However the full load temperature increases only by 1Â°C.
If we assume that my 300 Watt Fortron Aurora PSU
has efficiency similar to the 350 Watt version reviewed
by Mike, we can work backwards to get a rough estimate of the total
Mike’s testing showed that a 90W DC output required an input of 137W AC (66% efficiency). So 135W means the system was drawing 91W DC in idle.
The same review showed that a 150W DC output required an input of 217W AC (69% efficiency); as is the norm for PSUs, efficiency rises with increased output. Since the 189W system AC power draw at load is between 135W (66% efficiency) and 217W (69%), an estimate of 68% seems reasonable: This means the DC output at load is 189 x 0.68 = about
If we subtract the 105W for the CPU from the 129W total DC draw, we get 24
Watts of DC power left to be consumed by the rest of the system, including the
ATI 9500. At idle we saw the 9500-equipped system draw approximately 17 watts
more DC than the SiS VGA system did. A jump from 17 watts to 24 watts for the
9500 at load seems unbelievably small. A check of MBM’s CPU load log after an
extended testing session revealed the source of the apparant error: the CPU
load was often well below 100%. The combination of CPUBurn and 3DMark doesn’t
stress both the CPU and the GPU fully, simultaineously. This definitely complicates
A variety of different applications, benchmarks, and priority settings
were tried in an attempt to fully load both the VGA and the CPU simultaneously
for the 12+ hours needed to stabilize the temperatures, but nothing approached
the 189 Watts of maximum load seen with CPUBurn and 3DMark. Without the CPU being
fully loaded, there is no reliable way to determine what portion of the 129
total DC draw is caused by the CPU, and what portion stems from the VGA. (It
also calls into question the VGA Wattages calculated by other websites who use
the differences between idle and load Wattages as indicators)
The performance numbers really speak for themselves. For the noise level,
the temps are quite simply far beyond what is currently achievable with even
the best of air-cooling. Compared to other high-end water-cooling systems, the
temperature results are decent, but unspectacular. But in reality, comparisons
of the Reserator1 to conventional water-cooling kits are probably unfair. It
is designed for users wanting the maximum in silence, and ease of use, not maximum
Question: Could you select individual components
from a variety of manufacturers to assemble a system that will outperform
Question: Would it be as quiet?
Answer: Doubtful. If it has even a single
fan it would be louder than the Reserator.
The Reserator clearly succeeds in achieving its designers’ goals and represents another major achievement in silent computing for Zalman.
Much thanks to Sharkacorp for the Reserator1 review sample and for their patience.
* * *