ASUS GTX550 Ti DirectCU TOP ReviewRHKCommander959 - March 15, 2011
» Discuss this article (4)
DirectCU means that the GPU core is in direct contact with copper (element symbol Cu) heat pipes. The GTX 550 Ti has two 8mm heat pipes to wick heat away from the core to the fins where the heat is dissipated to air that is then circulated back into the case. With proper case ventilation, temperatures won't be a problem. The TOP nomenclature means that the card is a level above normal ASUS cards with the core, in this case, overclocked to 975MHz on the GTX 550 Ti, with the best edition dubbed ULTIMATE having a clock speed of 1015MHz.
The GTX 550 Ti GF116 core is built on the Taiwan Semiconductor Manufacturing Companies' 40nm node, carrying 1.17 billion transistors with it and 192 CUDA cores, 32 texture units, and 24 ROP units. The memory bus is 192-bit from three 64-bit controllers and the architecture supports mixed memory densities, thus allowing this card to have 1GB of GDDR5 memory whereas prior generations were limited to either 768MB or 1536MB of memory. The cards are capable of two dual link DVI-I ports, but ASUS chose to swap one for a D-sub VGA output for direct compatibility with older monitors. Personally I'd rather see two DVI ports with an adapter for VGA output, but most people looking at this card won't need two DVI ports anyway. The top of the card shows the translucent smokey gray fan impeller centered on the card with a black plastic cover over the fin array. Rotating to the back shows that the card has a stiffening bar to help keep the PCB from flexing. Many small components are on the back along with empty BGAs where more memory ICs could be installed. All six memory ICs that add up to 1GB are on the other side, underneath the heat sink. ASUS included blue covers for the SLI and PCI Express slots to help protect the gold pins from corrosion.
The 6-pin PCI Express power connection is mounted on the side and only one is needed as the card barely consumes more than the output of just the slot (75W), with reference cards having a TDP of 116 watts. The stiffening bar has a unique shape to it, where it starts thick at the back to help guide airflow and thin the rest of the way with curves to dodge the heatpipes and leave space for a SLI connector. The other side shows the 4-pin PWM fan header near the video outputs. A few memory ICs can also be spotted, which are exposed, but do receive some airflow from the fan, GDDR5 is efficient enough to not need direct cooling at rated speeds. The heat sink assembly clears the components with room to spare.
The card supports HDMI, D-Sub VGA, and DVI-I outputs. The HDMI port is bolted to the slot bracket to strengthen it since those cables simply slide in and out, whereas the VGA and DVI ports lock in and provide structure through the two hex head mounts. The back of the card shows the exhaust vents cut into the fan shroud and some capacitors and chokes. The card takes up two slots of space. The fan is able to take in air through two low spots on the shroud as well as pushing air from all around.
The single SLI port is located in the standard location at the edge of the card near the outputs. The power connection is at the opposite corner of the card exiting the side, this makes it easier to install the power cable, especially in cramped systems.
The base of the heat sink is quite rough, enough so that it could be used as an Emery board. Thermal performance is still terrific with this heat sink and lapping the base would be simple after removing the four mounts. The heat sink design is intended for use on other cards as well, that is why there are four alternative holes drilled and tapped for the mounting hardware. The thermal paste used was quite solid and didn't appear to adhere completely with the core — I had to scrape it off before cleaning the base plate. Swapping the thermal “paste” to Shin Etsu dropped the operating temperature by a few degrees Celsius. Six memory ICs are soldered in and add up to 1024MB of memory, which doesn't divide evenly by 6. NVIDIA used two 256MB ICs and four 128MB ICs to get the total amount since the new architecture allows for mixed densities. All ICs used are Hynix modules, while the GPU die was made at TSMC. The ASUS GPU Guard feature can be seen at the corners of the GPU core — glue helps bond the chip to the PCB and also protects the area from stress cracks or damaged traces caused by heat cycling and heat sink flex.
The card is equipped with ASUS Super Alloy Power (SAP) components that use a specially concocted in-house alloy developed by ASUS. These components consist of a Super Hybrid Engine IC, MOSFETs, chokes, and capacitors for improved efficiency in several regard. For instance, the card should operate without emitting any sort of squeal or buzz since these parts are higher quality — the chokes have concrete cores so that the coil cannot vibrate and thus create noise, which is a common issue on many AMD and NVIDIA cards that employ cheaper, hollow-casing chokes. The Super Hybrid Engine is a chip that throttles the graphics card for optimal power consumption by swapping between high- and low-power profiles. The components increase the card's thermal efficiency with a claimed 35 °C reduction in temperatures, which aids in overclocking. Headroom has also been increased with sturdier and higher voltage-capable MOSFETs and capacitors, with ASUS claiming a 15% performance boost from the SAP components. The capacitors have a much higher expected lifespan in comparison to typical capacitors, rated for a 2.5x MTTF over traditional styles. The photo shows all the components: at the top center is a big black square, this is the Super Hybrid Engine; next the MOSFETs, they are the small black squares with four pins coming off one side and a large metal base coming out the other side; the chokes are the larger gray solid squares that say SAP; and the cylinders are the capacitors. The SAP photo also illustrates each component in the same order that I listed them.
Next up are the detailed specifications and features listing.