by Sergey Lepilov
08/05/2010 | 03:27 PM
Top-performance graphics cards have come to produce so much heat that developers of their reference cooling systems often fail to find an optimal compromise between cooling performance and noise. A very good example is the Nvidia GeForce GTX 480 card with its native cooler. The cooling system of the dual-processor ATI Radeon HD 5970 has the same problems, being only quiet in 2D applications but loud and not very effective in 3D ones. Moreover, it is designed in such a way that the air flow cools one GPU first and then, having heated up, moves to the other GPU, the difference between the temperatures of the two graphics cores amounting to 10° and more as the consequence. The problem of cooling is less urgent for an ordinary user who runs his graphics card at its default frequencies but this difference in temperatures may prevent an overclocker from accelerating his dual-processor HD 5970. The Switzerland-based Arctic Cooling has addressed the issue by releasing a new cooler called Accelero XTREME 5970. Let’s take a look at it.
The cooler comes in a blister pack, a cardboard insert showing its specs, performance comparison diagrams and other information.
Particularly, the manufacturer claims that the Accelero XTREME 5970 is 40°C better than the reference cooler in terms of performance and 7 times better in terms of noisiness even if the reference cooler’s fan rotates at 2000 RPM. That’s impressive, but is this claim really true? We’ll check this out soon.
The following accessories are included with the cooler: a heatsink for power circuit components, a back-panel bracket, a set of thermal pads and an installation guide.
There are no screws and no thermal grease in the box. We will explain why shortly. The recommended price of the Arctic Cooling Accelero XTREME 5970 is rather high at $76.95 but the manufacturer offers a 6-year product warranty.
The Accelero XTREME 5970 follows Arctic Cooling's traditional design guidelines. There are longitudinal heat pipes, a large aluminum heatsink consisting of two halves, and three fans in a plastic frame to blow at the whole arrangement.
The cooler measures 290x104x58 millimeters and weighs 692 grams. Thermal grease (Arctic Cooling MX-2) is already applied to the two copper bases.
To have a closer look at the cooler’s heatsink, we had to remove the plastic casing with fans as well as the aluminum plate with heatsink which was attached to the bases.
The heatsink consists of 119 aluminum fins, 0.4 millimeters thick and 1.9 millimeters apart from each other. The fins are grouped into three sections of similar width.
There are a total of eight heat pipes, 6 millimeters in diameter, so each GPU has four of them. The left and medium sections of the heatsink are pierced by four heat pipes and the right section, by eight pipes in two rows.
Thus, each GPU is cooled by the same number of heat pipes and has the same share of the heatsink for itself. The heat pipes are connected to the copper bases by means of soldering. The fins seem to be press-fitted on the heat pipes.
Three white 92mm fan impellers are installed on the black plastic frame.
Hanging on four pins, each impeller has 11 sharp blades and a small sticker on the rotor.
The fans are connected to the 4-pin plug on the PCB of the Radeon HD 5970 and are PWM-controlled in a speed range of 900 to 2000 RPM. The combined airflow of the fans is specified to be 81 CFM. The level of noise at the maximum speed of 2000 RPM is declared to be no higher than 0.5 sones. For comparison, the manufacturer says that the default blower of the reference Radeon HD 5970 cooler produces 3.5 sones of noise at 2000 RPM, which is 7 times as high. The maximum power draw of one fan is 2 watts. The manufacturer does not specify the service life of the fluid dynamic bearings of the Accelero XTREME 5970 but if one of its fans fails after 5 years and 11 months of cooling your Radeon HD 5970, you will be able to contact Arctic Cooling’s support because the company offers a 6-year warranty for this and other products it manufactures.
Installing the Arctic Cooling Accelero XTREME 5970 on today’s fastest graphics card is likely to make the graphics card warranty void. We don’t think this can repel true enthusiasts, though. First of all, you have to remove the card’s native cooling system.
It is a rather simple operation. You just have to unfasten all the screws (do not forget the screws in its sides) and carefully take the massive cooler off (it will still sit rather tightly on the thermal pads and thermal grease even when the screws are unfastened).
Next you install the VRM heatsink and thermal pads:
You must carefully follow the instructions because the thermal pads differ in thickness depending on what component they are meant for. If you put a too thick pad on some component, it may prevent the other pads from contacting properly with the cooler.
Here is our Radeon HD 5970 with all the thermal pads and the VRM heatsink installed:
Now we only have to peel off the protective film from the pads and install the cooler proper, fastening it with the screws from the back-plate of the native Radeon HD 5970 cooler. The Accelero XTREME 5970 comes with good thermal grease already applied, but there is too much of it and the excess thermal grease comes out at the edges of the GPU dies.
The layer of thermal grease should be much thinner. We guess that it would even be better if the manufacturer didn't apply it at all but instead included some Arctic Cooling MX-2 or MX-3 into the box.
When installed on the graphics card, the Accelero XTREME 5970 looks like this:
If you take a look at the card from above, you can note that the PCB bends at the right edge of the second GPU and after. Why? Because the main heatsink of the Accelero XTREME 5970 presses against the tall VRM heatsink and bends the PCB. We don’t know why this problem has escaped Arctic Cooling’s engineers.
So, we had two ways to do our tests. We could test the bent graphics card or test it in its straight form but without the VRM heatsink. Neither variant was really good, yet we chose the first one and installed the curved thing into our system case.
The graphics card with Accelero XTREME 5970 blocks the two neighboring expansion slots on the mainboard but Arctic Cooling claims this product is compatible with CrossFireX configurations. Well, it is possible to install another Radeon HD 5970 but we can’t predict how this would affect the cooler’s performance. And we don’t have a second dual-processor card to check this out in practice.
We tested the new cooling systems inside a closed system case. Our testbed was configured as follows:
We overclocked our processor to 4.3 GHz and increased its Vcore to 1.3975 V in the mainboard BIOS. The memory voltage was at 1.64 V and its frequency was around 1.47 GHz (7-7-7-14_1T timings).
The graphics card frequency wasn’t increased during the primary test session and stayed at 725/4000 MHz:
The testing programs were installed under Microsoft Windows 7 Ultimate RTM x64. We used DirectX End-User Runtimes libraries (from February 2010) and Catalyst 10.5 graphics card drivers. We used 15 runs of FireFly Forest test from the semi-synthetic 3DMark 2006 suite in 2560x1600 resolution to warm up the graphics cards. We enabled anisotropic filtering 16x:
Besides, we additionally used FurMark version 1.8.0 burn test that was launched with a renamed EXE-file for about 20 minutes in stability test mode in 2560x1600 resolution:
We used MSI Afterburner utility version 1.6.0 Beta 5 to monitor graphics card temperatures and frequencies and GPU-Z version 0.4.3 utility:
The tests were run at least twice for each type of load. The temperature stabilization period between the two test cycles was about 10-12 minutes. The ambient temperature was checked next to the system case or open testbed with an electronic thermometer with 0.1°C precision that allows monitoring the temperature changes over the past 6 hours. During our test session room temperature stayed around 22.4-22.6°C.
The noise level of each cooler was measured after 1:00 AM in a closed room about 20 m2 big using CENTER-321 electronic noise meter. The noise level for each cooler was tested outside the system case when the only noise sources in the lab were the cooler and its fan(s). The noise meter was installed on a tripod and was always at a 200 mm distance from the cooler fan rotor. The mainboard with the graphics card equipped with the tested cooler was placed at the edge of the desk on a sheet of polyurethane foam. The lowest noise reading our noise meter device can register is 29.8 dBA and the subjectively comfortable noise level in these testing conditions was around 35 dBA (do not mix it up with low noise level). The fan(s) rotation speed was adjusted in the entire supported range using the new controller revision by changing the voltage with 0.5 V increment.
The following diagram shows how effective the Accelero XTREME 5970 is:
So, the performance of the new product from Arctic Cooling is impressive indeed. Working at a quiet 1250 RPM, the Accelero XTREME 5970 proves to be more effective than the graphics card's native cooler at 2170 RPM by 33°C and 41°C in terms of the temperature of the first and second GPU, respectively! Besides higher performance, the Accelero is much quieter than the native cooler then. At its maximum speed of 1920 RPM, the new cooler lowers the temperature by 4-5°C more, which is not even necessary since the GPUs are only 60-65°C hot.
The Accelero XTREME 5970 is also good at cooling the power circuit components, being only inferior to the native cooler when the latter is rotating at an unacceptable 4670 RPM.
Now let’s see how effective the coolers are at overclocking.
We only checked out the temperature of the graphics card overclocked to 830/4400 MHz at the maximum speed of the fan(s) because of a dangerously high temperature of the power circuit components. You can see the results in the following screenshots…
…and in the diagram:
So, the Arctic Cooling Accelero XTREME 5970 copes easily with cooling the overclocked graphics card although the power circuit components get as hot as 117°C. But when we replaced the thermal pad of the VRM heatsink with thermal grease, the peak temperature of these components lowered by 19-24°C. We are more worried about the 9°C difference between the GPUs. We guess it must be due to the difference in contact between the cooler’s bases and the GPU dies which is the consequence of the bending of the PCB we noted above. It just shows up more when the graphics card is overclocked and generates more heat. Compared to the reference Radeon HD 5970 cooler, the Accelero XTREME 5970 doesn’t look much better, but we shouldn’t forget about the noise factor.
The following diagram shows how much noise is produced by the reference Radeon HD 5970 cooler and the Arctic Cooling Accelero XTREME 5970.
The new cooler from Arctic Cooling is much quieter than the reference one. We must confess that the difference is huge subjectively.
The Arctic Cooling Accelero XTREME 5970 proves to be many times quieter and much more effective than the reference cooling system of the dual-processor ATI Radeon HD 5970 graphics card. The Accelero XTREME 5970 is not taken aback by the card's extremely high heat dissipation even at overclocked frequencies.
However, the Swiss product is not without downsides, the most important of which is the tall VRM heatsink. The main heatsink presses against the latter and bends the card’s PCB as the result. We also think that instead of applying a generous amount of Arctic Cooling MX-2 thermal grease on the cooler, the manufacturer should provide it in a tube so that the user could apply it himself as necessary. The recommended price of $76.95 is higher in comparison with the Accelero XTREME 4870 X2, which can hardly please potential buyers, either. And finally, this cooler should be used in a well-ventilated system case because otherwise most of the hot air from it will remain inside, frying up the rest of the computer components.
If you can put up with these shortcomings, the Accelero XTREME 5970 is going to be a good choice for you.