by Sergey Lepilov
08/11/2008 | 04:53 PM
At first our today’s review was supposed to be a comparison of cooling solutions from the mainstream price range. However, once we completed the tests of our major participants and finished working on their design and functionality description, it turned out that they do not actually fall within the boundaries of the mainstream price range. The prices of the Noctua NH-U9B, Thermalright Ultima-90I and Zalman CNPS9300 AT coolers that we are going to talk about today exceed $45. This observation not only changed our choice of the opponent for results comparison, but also affected the preliminary conclusions we have already made for ourselves. Nevertheless, we believe this article will be of particular interest not only to extreme overclocking fans, but also to those users who are currently looking for a new CPU cooler.
Let’s meet our testing participants in person.
Updated version of the Noctua NH-U9 cooler that we have already tested before arrived in a relatively small box designed in Noctua’s traditional color scheme. The package contains a lot of information on all aspects of this cooling system, including its technical specifications and key features:
There is a small cut-out window in the front of the box, where you can see part of the cooler fan and heatsink. The accessories are packed into a smaller section of the box. Among them you can find the following items:
The retention kit for each socket type is sealed in an individual plastic bag with a corresponding sticker. Besides, Noctua NH-U9B comes with the following accessories:
The cooler tower-heatsink consists of 38 aluminum plates ~0.4mm thin spaced out at approximately 3mm. They are pierced with four copper heatpipes 6mm in diameter:
The modified cooler is not very big: it measures 125 x 96 x 95mm and weighs 460g.
The sides of the heatsink array are partially closed with bent plate edges:
The heatsink plates have wavy edges in order to lower the airflow resistance and noise. There is an embossed Noctua logo at the top plate:
The copper heatpipes are set in grooves in the cooler base and then go nonlinearly through the heatsink array:
These two things should improve the cooling efficiency. In the first case, heatpipes have bigger contact area with the cooler base; and in the second case the heat should be distributed more evenly over the heatsink. Non-linear heatpipes positioning also creates additional airflow resistance.
The copper plate in the base of the cooler is barely finished at all. You can not just see, but also clearly feel the machine marks on the base:
Moreover, the base of the cooler sample we received in our lab turned out bent inwards. As you can see, only its sides actually touched the CPU heat-spreader or the test glass surface:
After researching a few forum threads, we discovered that we were not the only ones who experienced problems with uneven cooler base of Noctua NH-U9B. Well, looks like Thermalright is not the only company after all that has uneven cooler bases :)
With a bent base like that the cooler couldn’t work efficiently enough even in the nominal CPU mode. So, I had to use a few types of sand paper to shave the dents off until I managed to obtain a more or less acceptable thermal compound imprint, efficient enough heat dissipation and the opportunity to overclock my test CPU even with Noctua NH-U9B:
I didn’t polish the base surface. To be fair, I have to say that out of four Noctua cooler we have tested so far in our lbs, this is the first one with an uneven base like that.
Unlike its predecessor, Noctua NH-U9B is equipped with a new NF-B9-1600 fan measuring 92 x 92 x 25mm that is actually a smaller modification of NF-P12, we have already discussed in our roundup:
According to the official specifications, the fan rotation speed makes ~1600RPM (±10%) creating 64.3m3/min airflow and generating 17.6dBA of noise. If you use the bundled adapters with built-in resistors, you will be able to lower the fan rotation speed to ~1300RPM and ~1000RPM. The airflow and noise level will also drop down respectively.
You can attach two fans to the cooler heatsink: one for air intake and another for air exhaust. The accessories bundle includes four silicon strips with one sticky side:
… and four wire clips:
With only one fan installed, the cooler looks like this:
And this is what Noctua NH-U9B looks like with two fans inside a system case:
The patented SecuFirm retention mechanism allows installing Noctua NH-U9B facing in any direction, no matter what type of CPU and socket there is. The distance from the mainboard PCB to the lower plate of the cooler heatsink is 32mm.
The modified new cooler from Noctua is priced at $45.
Thermalright packaging didn’t change at all: it is exactly the same as for all the other cooling solutions we reviewed before. The same brown cardboard with two things written down on it: the company name on the top and the cooler model name on the side of the box:
Inside this box there is a polyurethane foam casing that holds the cooler and a small box with accessories:
Like all other cooling systems from this vendor, Thermalright Ultima-90I ships without a fan. Here I would like to say that there are two points of view on this aspect. On the one hand, without the fan the cooler is obviously less expensive and the user is free to choose any fan he wants (or use the cooler in passive mode without any fan at all). But on the other hand, there are a few types of fans with different characteristics. For example, you can differentiate the fans depending on the air pressure or airflow used, which is also pretty important for heatsinks of different cooler types. So, the cooler manufacturer should actually know best of all what type of fans with what rotation speed would be optimal for their solution. With Thermalright, this choice is solely in users’ hands, and in most cases they have no opportunity to try a few different types of fans to find the best one. Anyway, every overclocker can actually decide on the most suitable approach for himself, especially since there are a lot of really good cooling systems in the market these days (according to my estimates, every month there appear about 2-3 new high-quality cooling solutions). So let’s get back to discussing Thermalright Ultima-90I.
Thermalright Ultima-90I measures 115 x 55 x 139mm and weighs 460g. It has a classical tower design: two rows of six copper heatpipes piercing the heatsink array of 48 aluminum plates:
The plates are ~0.25mm thick and the gap between them is 1.5mm:
Our regular readers should already be familiar with this heatsink design, because we have already discussed two higher-end cooling solutions from Thermalright Ultra-120 series with identical heatsink design concept. If we compare our today’s hero against those two, we have to say that the heatsink of Thermalright Ultima-90I is 21mm shorter, 17mm narrower and has 8.4mm smaller depth. Besides, Thermalright Ultima-90I weighs 285g less than Ultra-120 and 330g less than Ultra-120 eXtreme. By the way, 6 copper heatpipes with 6mm diameter make Thermalright Ultima-90I more similar to the latter, Ultra-120 eXtreme. The entire heatsink is nickel-plated and all its parts are soldered together.
The sides of the heatsink are open:
The top plate has no embossed logo, like the top models. Everything is very strict, without any decorations that could exceed the “budget”:
There are special grooves in the copper base of the cooler for each of the heatpipes, which increases the contact surface and improves heat conductivity:
Unfortunately, the base of the cooler is of Thermalright’s typical quality: you can not only see the machine marks with a naked eye but even feel them to the touch:
The evenness of the base is also traditionally “good”, i.e. its middle part stands out:
As a result, the thermal compound imprint on the CPU heat-spreader turned out uneven:
As you may have guessed, the bent cooler base contacted the processor heat-spreader only in the center, however, despite this fact, Thermalright Ultima-90I turned out pretty efficient, unlike Noctua NH-U9B. Looks like even a small contact spot like that is enough to ensure proper cooling of a 45nm processor. However, I am sure I am not the only one who would like to know when Thermalright will finally start making even bases for their cooling solutions? Not to mention the polishing of the base, of course.
To ensure that the fan sits “softly” on the cooler heatsink, this solution is bundled with two rubber strips that should be stuck to both sides of the heatsink before fan installation:
Although the heatsink is completely symmetrical and could accommodate two fans, there is only one set of fan retentions and vibration absorbing strips included.
Thermalright Ultima-90I can work with one or two 92 x 92 x 25mm fans. This is what it will look like:
However, the remarkable thing is that Thermalright engineers also ensured that you will be able to install one 120 x 120 x 25mm fan that is why they included two long wire clips:
Here we have to say, though, that about 30% of this fan’s potential will go to waste, because it will send the airflow above the cooler heatsink. Unfortunately, the uniquely shaped wire clips will not allow you to lower the fan, so that it could at least cool the components in the area around the processor socket. Moreover, the heatsinks installed on the voltage regulator components may also be in the way in this case.
Thermalright Ultima-90I is installed onto a mainboard in exactly the same way as Thermalright SI-128 SE. the cooler is pressed with a metal plate that is attached with spring screws to the spindles inserted into PCB retention holes:
I have to say that this is a very reliable retention mechanism ensuring pretty secure contact between the cooler base and the processor heat-spreader. I only have to add that the distance from the mainboard PCB to the lowest heatsink plate measures 42mm.
This is what Thermalright Ultima-90I looks like inside a system case (with a Noctua NH-U9B fan attached):
You can turn the cooler facing any of the four directions on LGA775 mainboards, because the retention holes there are symmetrical. However, mainboards for AMD K8 and K10 processors only allow two choices, which is actually traditional for CPU sockets of this type.
The cooler we have just discussed is priced at $60, which is comparable to the price on the best air coolers out there these days. Our tests will show if this price point is too high or absolutely justified. However, before we move over to the benchmarks, let’s discuss the third testing participant.
A small cardboard box of the new Zalman cooler has a semi-open front, so that the user could check out the cooler even before purchasing it:
Besides that, the package is decorated with a few cooler photographs, bears a complete list of supported processor sockets and very brief list of technical specifications.
The cooler itself sits securely in a clear plastic casing inside the cardboard box:
This casing protects the cooler against all sorts of transportation damages very well. The bundled accessories are stored in a small section of the box. Among them are the following items:
The design of the new Zalman CNPS9300 AT is not that new at all. In fact, they borrowed the heatsink design from the pretty old Zalman CNPS9500 and CNPS9700 cooler models. The array of wavy plates sits on heatpipes and is cooled with a fan:
Zalman CNPS9300 AT is smaller than its elder brothers. It measures 61.35 x 108 x 132.5mm. It also has fewer heatpipes (two instead of three) and weighs less: only 407g. Other than that, it is exactly the same as CNPS9500/9700.
Zalman CNPS9300 AT uses only two copper heatpipes 6mm in diameter. They originate from the copper base, loop forming an unfinished 8 and return back into the base. This is what this heatpipe web looks like:
The heatsink copper plates are soldered to the heatpipes piercing them:
Zalman has always soldered all heatsink components together, it is their distinguishing feature. In the base the heatpipes lie in special grooves and are also soldered to them:
The whole thing is cooled with a 92mm fan installed in a plastic frame on one of the sides of the heatsink:
Zalman CNPS9300 AT uses a fan with Superflo bearing and 50,000 hours MTBF:
The fan supports PWM rotation speed control feature. Its rotation speed can be adjusted from ~1500RPM to ~2400RPM at 20-30dBA noise. Too bad that even at the slowest rotation speed the fan motor generates very annoying crackling sound that makes the noise from the system case sound very uncomfortable. T maximum rotation speed the sound of airflow created by this fan joins the crackle.
While the fan on Zalman CNPS9300 AT is definitely not the best it could be, the base quality is impeccable: it is perfectly even and polished:
The mirror-quality of the base provides a very good thermal compound imprint (it got slightly smeared when we removed the cooler from the CPU):
Zalman CNPS9300 AT is compatible with all contemporary processor sockets. Despite its relatively small size, the manufacturer still provided a special scheme that will help you determine if the cooler is going to fit into your system case:
As you can see from the pictures above, this small cooler will fit even in compact system cases with components packed pretty tightly. Since we didn’t have a case like that at our disposal at the time of tests, here is a picture of Zalman CNPS9300 AT installed into ASUS ASCOT 6AR2-B:
The installation is fairly simple and doesn’t differ from the installation procedure for Zalman CNPS9700 or 9500 coolers we tested before. The cooler is pressed firmly against the CPU, it doesn’t shift or rotate on top of the processor heat-spreader. LGA775 frame has four retention holes and the retention plate may be inserted in-between the heatpipes, so you will be able to turn Zalman CNPS9300 AT facing any way you like on this platform. However, Socket 754/939/940/AM2(+) mainboards allow the cooler to face only one of two ways, because the retention clip should catch on to the hooks on a standard plastic frame, which will in the end determine the direction of the cooler’s airflow.
In conclusion I would like to offer you a few shots showing the new Zalman CNPS9300 AT side by side with CNPS9700 NT, the largest cooler of the same topology:
As you can see, Zalman CNPS9300 AT is 10mm shorter, 16mm narrower and ~28mm smaller in depth than CNPS9700. Besides, the new cooler weighs nearly half as much (407g against 813g).
Zalman cooling systems have never been cheap, unfortunately. New Zalman CNPS9300 AT features MSRP of $49.95, which is $20 less than MSRP for CNPS9500 and $30 less than MSRP for CNPS9700. However, if we compare it against the competitors’ solutions, the price will definitely strike as way too high for a cooler of this class and efficiency.
The technical specifications and recommended retail pricing for all coolers we discussed today are given in the table below:
The today’s coolers and their competitor were tested in two modes: in an open testbed when the mainboard sits horizontally on the desk and the cooler is installed vertically, and in a closed testbed with the mainboard in vertical position.
Our testbed was identical for all coolers and featured the following configuration:
All tests were performed under Windows XP Professional Edition SP2. SpeedFan 4.34 was used to monitor the temperature of the CPU, reading it directly from the CPU core sensor:
The mainboard’s automatic fan speed management feature was disabled for the time of the tests in the mainboard BIOS. The CPU thermal throttling was controlled with the RightMark CPU Clock Utility version 2.35.0:
The CPU was heated up with OCCT (OverClock Checking Tool) version 2.0.0a in a 23-minute test with maximum CPU utilization, during which the system remained idle in the first 1 and last 4 minutes of the test:
I performed at least two cycles of tests and waited for approximately 20 minutes for the temperature inside the system case to stabilize during each test cycle. The stabilization period in an open testbed took about half the time. Despite the stabilization period, the result of the second test cycle was usually 0.5-1°C higher. The maximum temperature of the hottest CPU core of the four in the two test cycles was considered the final result (if the difference was no bigger than 1°C – otherwise the test was performed at least once again).
The ambient temperature was checked next to the system case with an electronic thermometer that allows monitoring the temperature changes over the past 6 hours. During our test session room temperatures varied between 26.0 ~ 26.5°C. It is used as a staring point on the temperature diagrams. Note that the fan rotation speeds as shown in the diagrams are the average readings reported by SpeedFan, and not the official claimed fan specifications.
I suggest comparing our today’s testing participants against Thermalright SI-128 SE. It is priced at $50, which is comparable to the pricing of our today’s cooling solutions. Besides, it is of pretty similar dimensions. We tested this cooler only in one mode with Scythe SlipStream 120 fan working at ~870RPM. Thermalright Ultima-90I was also additionally tested with the same fan besides the tests with 92-mm Noctua fan(s) at ~1110RPM and ~1730RPM in the primary test session. I decided to skip the intermediate speed of 1360RPM, so you won’t find any results for it in the charts below.
Now let’s check out the results.
Using the “weakest” cooling system with the fan in quiet mode we managed to overclock our 45nm quad-core processor to 3.75GHz inside a closed system case. The processor Vcore was increased to 1.4875V in the mainboard BIOS:
The monitoring programs detected a slightly lower voltage setting than in the mainboard BIOS, around 1.45~1.47V:
The diagram below reveals the results obtained for our today’s testing participants in this case:
Let’s analyze the obtained results in order of appearance on the diagram (the same order as the coolers appearance in our today’s review). At first a few words about Noctua NH-U9B. This compact cooler equipped with one 92-mm fan working at ~1110RPM turned out not very efficient for an overclocked CPU, especially in a closed system case and pretty high room temperature. However thanks to the opportunity to use a second fan for air exhaust and to increase their rotation speed, Noctua NH-U9B improved its cooling efficiency by 9ºC in an open testbed and by 11ºC in a closed system case. Therefore we can say that the efficiency of this Austrian cooler depends a lot on the airflow going through the heatsink array. I would also like to add that 92mm fan at its maximum speed of ~1730RPM (according to monitoring utilities) doesn’t generate too much noise: it never goes past 35dBA (measured with a noise-meter at 1m distance). If the rotation speed is lower, you can barely hear the fan at all. The motor doesn’t crackle, too.
Our second testing participant, Thermalright Ultima-90I is actually the best of the three coolers discussed today. This cooler efficiency depends even more on the fan(s) rotation speed, because this heatsink array is denser than that of Noctua cooler. In the end, Thermalright Ultima-90I equipped with two 92-mm fans for air intake and exhaust rotating at ~1730RPM doesn’t fall too far behind Thermalright SI-128 SE, although the latter is working in a quieter mode in our today’s test session. As for the results Thermalright Ultima-90I demonstrated with a 120 x 120 x 25mm fan, there is nothing surprising about them. if we set both, 92- and 120-mm fans, at the same noise level using our noise-meter device (at ~900RPM and ~1450RPM respectively), Thermalright Ultima-90I will be identically efficient in both cases. That is why we decided not to overload our charts with additional results obtained with a 120-mm fan.
And finally, a few words about Zalman CNPS9300 AT. Only inside a closed system case this cooler performs as efficiently as Noctua NH-U9B with one fan working at ~1730RPM. IN an open testbed Zalman CNPS9300 AT falls pretty far behind its today’ rivals. It only manages to outperform their quietest operational modes. Besides, we also detected crackling of the fan motor, which you can hear even at the minimal rotation speed of ~1530RPM.
To tell the truth, I could have made a much more optimistic conclusion to our today’s review, if it hadn’t been for the recommended pricing of the coolers tested. The thing is that all three cooling solutions discussed today do not belong to the mainstream price segment, because the lowest price starts only at $45, while the most expensive one is priced at $60 and comes even without a fan! With prices that high and with the cooling efficiency they can offer, Noctua NH-U9B, Thermalright Ultima-90I and Zalman CNPS9300 AT will most likely be uncompetitive. In my opinion, they should sell for at least half the price to become more or less successful in the today’s market, because in this case they will be opposing totally different competitors.
You will probably say that they are very compact and that’s their strength. Well, I have to remind you that Thermalright SI-128 SE, which we have also tested today, measures 116.5mm with a fan. Only Noctua NH-U9B is shorter than that, while Thermalright Ultima-90I and Zalman CNPS9300 AT are definitely taller. However, SI-128 SE is far more efficient than they are and costs $50. Anyway, if you are satisfied with the cooling efficiency these solutions have to offer, the choice is yours.