Leadtek WinFast GTX 280
Like every version of GeForce GTX 280 we have tested in our labs, Leadtek’s is a precise copy of Nvidia’s reference sample. The sticker on the cooler’s casing is the only difference:
The PCB is as long as the Radeon HD 4870 X2 and HD 4850 X2, i.e. 27 centimeters, and the GeForce GTX 280 is not far simpler than the latter cards. Although it has only one GPU on board, it uses a 512-bit memory bus. So, it is very unlikely that some graphics card maker will come up with a nonstandard PCB for GeForce GTX 280.
A seven-phase GPU voltage regulator based on a Volterra VT1165MF controller resides in the right part of the PCB under the cooler’s casing. It has two connectors for external power supply, one of which is of the 8-pin variety. It is impossible to connect the PSU’s 6-pin connector to the latter: the card won’t start up, reporting a power problem by means of a red LED on the mounting bracket. The memory voltage regulator is based on a Richtek RT9259A controller that is powered by the PCI Express x16 slot.
Like the Sapphire HD 4850 X2, the Leadtek WinFast GTX 280 is equipped with 16 memory chips. These are Hynix H5RS5223CFR-N2C chips with a capacity of 512Mb (16Mb x 32), voltage of 2.05V, and rated clock rate of 1200 (2400) MHz. They make up a 1024MB memory bank accessed across a 512-bit memory bus. The memory is clocked at a frequency of 1100 (2200) MHz, providing a bandwidth of 140.8GBps, which is higher than the total bandwidth of the Sapphire HD 4850 X2 (128GBps). Half of the chips are located on the reverse side of the PCB and cooled with a metallic plate that is part of the cooling system. Fiber pads soaked in white thermal grease, typical of all solutions from Nvidia, are used as a thermal interface.
As opposed to the RV770, the die of the G200 processor has a heat-spreading cap protecting it from damage. The Leadtek WinFast GTX 280 comes with a revision A2 core marked as G200-300-A2. The second number seems to indicate a chip series according to the frequency and/or number of operating functional subunits. The core works in its full configuration with 240 unified shader processors, 80 texture-mapping units, and 32 raster back-ends. The card has the same GPU frequencies as the reference sample: 1296MHz for the shader domain and 602MHz for the main domain (including the raster back-ends).
The G200 is manufactured on 65nm tech process. Even without the display controllers it is as large as 576 sq. mm, incorporating 1.4 billion transistors. Therefore the display controllers are implemented as an individual NVIO2-A2 chip located near the DVI connectors. The graphics card is equipped with two dual-link DVI-I ports supporting resolutions up to 2560x1600, one universal analog video output, two SLI connectors (to combine up to three cards in a single multi-GPU subsystem), and an internal S/PDIF header. The latter is necessary to translate digital sound from your sound card into HDMI because the G200 does not incorporate an audio core.
The Leadtek WinFast GTX 280 uses the time-tested cooler that traces its origin back to the cooler of the GeForce 8800 GTX. Leadtek’s card differs from other versions of GeForce GTX 280 with the stickers: a picture of a robot on the cooler and a Leadtek logo on the fan.
The cooling system is based on a large heatsink consisting of thin aluminum plates and connected with heat pipes to the copper sole that contacts with the GPU cap. The metallic casing is part of the heat dissipation system, too. The aluminum base of the cooler has protrusions with thermal pads against those elements that require additional cooling such as power transistors or the NVIO chip. The airflow is created by a blower located at the back of the card. It blows through the heatsink and the hot air is exhausted out of the system case through the slits in the card’s mounting bracket. The cooler boasts high efficiency. The fan works at a reduced speed most of the time, ensuring comfortable noise level.
This cooling system seems to be better than Sapphire’s but its job is simpler, too. It has to cool one GPU only and the peak power consumption of the GPU is lower than the combined consumption of two RV770 chips even when the latter are clocked at reduced frequencies. We’ll check this out in the next section, though.