by Sergey Lepilov
04/26/2009 | 03:00 PM
Today AMD Radeon HD 48x0 graphics cards seem to be in much greater need for efficient and quiet cooling than Nvidia GeForce GTX 260-285. The reference coolers of the AMD solutions are much noisier; besides, the PCB of these graphics cards heats much more than the PCB of the GTX solutions. Nevertheless, the new extreme cooler from the Swiss Arctic Cooling Company is designed for Nvidia GeForce GTX 260-285, which is clearly indicated by its model name: Arctic Cooling Accelero Xtreme GTX 280. However, Arctic Cooling also has a VGA cooler for the dual-processor Radeon HD 4870 X2, which suggests that they may soon launch an Accelero Xtreme for Radeon HD 4870/90 as well. But today we are going to talk about a new cooler for GeForce GTX 260-285 graphics cards, will compare its cooling efficiency and acoustic performance against that of a reference cooling solution.
As usual, Arctic Cooling product is sealed in a clear plastic casing molded following the cooler shape:
As a result, you can see the front of the cooler. The back side is covered with paper insert with the technical specifications and two comparative diagrams for the cooling efficiency and noise. Besides the cooler, there are also aluminum heatsinks for the voltage regulator components and memory chips, sticky thermal pads, a set of retention screws and an installation manual:
Moreover, there is an additional heatsink plate for the voltage regulator elements and a plate for the back of the card, which we are going to dwell on later in this review.
Arctic Cooling Accelero Xtreme GTX 280 is priced at $55 and comes with 6-year warranty.
The new cooler measures 287 x 104 x 54mm and weighs 680g, which is close to the contemporary CPU cooler dimensions rather than VGA ones. Accelero Xtreme GTX 280 is built on five copper heatpipes 6mm in diameter going through the copper base. They hold two aluminum heatsink arrays cooled with three 92-mm fans:
The cooler is 54mm tall, which means that when you install it into your system, it will block not only the next PCI slot, but also the following slot:
Moreover, to ensure easy airflow to the cooler fans it is recommended to keep the third PCI-E slot empty as well.
The heatpipes inside the heatsink are all lined up in a little curve:
I don’t understand completely why they did it this way, because the curve is too small to lower the airflow resistance, for instance. As for more even heat distribution over the heatsink, this curve will hardly improve it. So, let’s consider it a small alignment production issue.
The fans are attached to a plastic frame that can be removed from the heatsink completely. This is what it looks like without the fans:
There is an aluminum heatsink right above the cooler base and the heatpipes coming out of it. This heatsink is soldered to the heatpipes, and the heatpipes – to the copper cooler base plate:
There is Arctic MX-2 thermal interface already pre-applied onto the center of the base plate. It is one of the best thermal compounds today:
However, as you can see from the imprint on the GPU heat-spreader, there is not enough thermal paste to cover the entire contact surface completely:
Of course, the GPU die is smaller than the imprint, however, we applied thermal compound to the entire cooler base to ensure proper contact between the GPU and the cooler (it is a little smaller than the GPU heat-spreader).
The base could have been finished a little better, because you can not only see the machine marks but also feel them by touch:
Three fans cooling the heatsink rotate with the same speed. They are PWM controlled in the interval from 1000 to 200RPM.
Arctic Cooling measures the acoustic in Sones and the maximum noise from this solution is claimed to be maximum 0.5 Sone. The overall airflow is 81CFM. The fans use a dynamic fluid bearing with unknown MTBF.
However, if the manufacturer promises that Accelero Xtreme GTX 280 will serve you well for 6 years, the fan bearings should last least that long.
The fans are connected to the standard GeForce GTX 260-285 connector with a single four-pin cable:
And the last part of the Arctic Cooling system discussed today is the aluminum backplate for the reverse side of the graphics card PCB…
… and an aluminum heatsink plate for the PCB power components:
You have to remove the protective plastic films from both of them before installation.
Accelero Xtreme GTX 280 is designed for GeForce GTX 260 and GTX 280 graphics accelerators. However, we are pretty certain that this cooler will also fit the GeForce GTX 285 and GTX 275 graphics accelerators. The only question is if the heatsink plates will fit the power elements. We will be able to answer this question only when these graphics cards arrive into our cooling lab.
Today we are going to install and test Accelero Xtreme GTX 280 on ZOTAC GeForce GTX 260 AMP2! Edition 896MB graphics accelerator:
When we removed the reference cooler, we had to clean and degrease the NVIO chip, memory chips and power elements. The installation manual (1.46MB file) suggests that you should use a soft eraser for that purpose. After that remove the protective film from the thermal pads and stick them onto all components:
The aluminum heatsinks will go right on top of the pads:
Both these heatsinks stay in place very securely (the heatsink covering NVIO and two memory chips is attached with two screws from the back of the PCB). The memory chips on the front of the PCB do not need any thermal pads, because they will be covered with a heatsink that already has sticky thermal interface applied to it:
However, these heatsinks do not hold too well even on clean and degreased chips. The same is true for the aluminum backplate. It is a real pity that Arctic Cooling engineers didn’t think through this particular issue. However, when I accidentally put the memory heatsink onto the aluminum backplate, it took me a while to rip it off.
When all heatsinks are successfully installed, the cooler is fastened to the PCB and GPU with four screws with washers:
This is what our super-cooler looks like when installed onto a GeForce GTX 260:
And this is the same cooler inside the system case:
At least two mainboard slots next to the graphics card should be left empty. As you understand, this cooler will not suit for SLI configurations.
We tested the new cooling system from Arctic Cooling inside a closed system case. Our testbed was identical for all coolers throughout the test session and featured the following configuration:
To increase the total system heat dissipation and make it the testing conditions a little harder for the participating cooling solutions we overclocked our quad-core processor to 4.01GHz and increased its Vcore to 1.36875V:
The system RAM was working at 1524MHz with 8-8-8-18 timings and 1.55V voltage. The graphics card is already pre-overclocked to frequencies higher than those of the reference GeForce GTX 260. Unfortunately, only the graphics memory could be overclocked further. The GPU frequency refused to go up even when we raised the voltage from 1.06V to 1.12V. The resulting operational frequencies for our graphics card equaled 648/1404/2430MHz. Before we started the tests we cleaned the cooling system of our reference GeForce GTX 260 (the card has been working in our testbed for about 3 months by then) and replaced the thermal interface with a fresh layer of Arctic Cooling MX-2.
The testing programs were installed under Windows Vista Ultimate Edition x86 SP1. We used DirectX libraries (from March 2009) and GeForce 185.68 drivers. We used FurMark version 1.6.5 to warm up the graphics cards. It was launched for 15-minute runs in stability test window mode in 1280x1024 resolution:
Judging by the load it creates for the graphics cards, it could be compared to Linpack. In other words, the graphics cards are very unlikely to run so hot in games.
We used the irreplaceable RivaTuner v2.24 (created by A. Nikolaichuk aka Unwinder) to monitor the frequencies and temperatures of our GeForce GTX 260. So, the complete screenshot during the test session looks as follows:
The tests were run at least twice. 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 at 23.5-24°C.
The noise level of each cooler was measured after 1:00AM in a closed room about 20sq.m big using CENTER-321 electronic noise meter. The measurements were taken at 1m and 3m distance from the closed system case. During the acoustics tests all five 120-mm case fans were slowed down to ~520 RPM. In this mode the background noise from the system case measured at 1m distance didn’t exceed ~33.3 dBA. When the system was completely powered off, our noise meter detected 30.8 dBA (the lowest on the charts is 30 dBA). The subjectively comfortable noise level is around 34.5~35 dBA.
First let’s check out the noise measurements. Note that on the upcoming cooling efficiency diagram the results will be sorted out in order of noisiness. This way we will be able to base our verdict on the combination of both characteristics.
Click to enlarge
No doubt, Accelero Xtreme GTX 280 is quieter than the reference GeForce GTX 260 cooler and the difference is dramatic. Both testing participants run equal in terms of noise levels until about 35% of their capacity and then “take alternate routes”. It is interesting that even at 50% capacity (around 1560RPM) Accelero Xtreme GTX 280 is moderately noisy, while GeForce GTX 260 cooler at the same 50% (1790RPM) brings evident discomfort to the relatively quiet system. At maximum fan rotation speeds the difference in noise reaches 11.1dBA!
The results on the diagram include not only the GPU temperatures, but also the Ambient temperature and the temperature of the hottest voltage regulator component of the three (VRM). The lower the cooler is on the diagram, the higher is the noise level from it:
Against the background of steady “±2-4°C” in our CPU cooler reviews, Accelero Xtreme GTX 280 demonstrates outstanding advantage over the reference GeForce GTX 260 cooling system: 30°C! And this remarkable result is achieved at a much lower noise level. Excellent performance!
Unfortunately, not everything was so unbelievably great. The “bad egg” in this case was the VRM heatsink that proved unable to transfer and dissipate the heat efficiently enough. While for Radeon HD 4870/4890 VRM temperature of 113°C seems more or less typical, then for GeForce GTX 260/280 it is absolutely not. Accelero Xtreme GTX 280 with the fans working at their maximum speed keeps this temperature below 100°C, but above the reading from the reference GeForce GTX 260 cooler. I believe it was the high VRM temperature that wouldn’t let me overclock the card any further with Accelero Xtreme GTX 280 cooler.
It is such a pity that they overlooked the VRM cooling issue. Otherwise, Accelero Xtreme GTX 280 could be an ideal solution for GeForce GTX 280/260 graphics cards. Impressive efficiency advantage over the reference cooler in GPU and ambient temperature at a significantly lower level of generated noise make Arctic Cooling product a very attractive choice. At the same time, inefficient VRM heatsink plate puts a fat question mark at the end of this sentence. Therefore, we would strongly advise that all potential Accelero Xtreme GTX 280 owners should take care of the VRM cooling on their own and keep in mind that the new Arctic Cooling solution is incompatible with SLI configurations and removes only part of the hot airflow outside the system case. Other than that, it is an excellent solution.