PCB Design and Specifications
All GeForce GTX 280/260 cards are copies of the reference samples, so there is no need to describe their design again. We already did this in our review of the theoretical aspects of Nvidia’s new architecture. Of course, the Zotac card is no different than others.
The single difference from the reference GeForce GTX 280 is the stickers on the cooler’s casing. The card is 27 centimeters long. ATI’s new flagship solutions are just as long as that, but the reasons differ: Nvidia’s card is long because of the 512-bit memory bus while ATI’s, because of the two GPUs. Anyway, you may have problems installing such long cards into a small system case.
Being a copy of the reference card, the Zotac has a seven-phase power circuit governed by a popular Volterra VT1165MF controller. One 6-pin and one 8-pin connectors are used for external power supply. You have to connect an 8-pin power cable or the card won’t start up, reporting a problem with a red LED. It is rather easy to bypass this protection, but we wouldn’t recommend you to do so. Even without the factory overclocking the load on the 8-pin connector is higher than the maximum allowable load for the 6-pin connector (75 watts). It is going to be even higher on the Zotac card which claims to be the fastest GeForce GTX 280 with air cooling. You will see that shortly. The memory power circuit is based on a Richtek RT9259A and receives power from the appropriate section of the PCI Express x16 slot.
The GeForce GTX 280 features a 512-bit memory bus, so there are two times more seats for GDDR3 memory chips in comparison with the traditional 256-bit designs. There are 8 chips on each side of the PCB (Hynix H5RS5223CFR-N2C, 512Mb, 16Mb x 32, 2.05V). They are rated for a frequency of 1200 (2400) MHz. The memory frequency is increased relative to the reference GeForce GTX 280: 1150 (2300) MHz as opposed to 1100 (2200) MHz. The memory bandwidth is thus increased from 140.8 to 147.2GBps. That’s not much. This can hardly affect the performance of the Zotac GeForce 280 AMP! in games, especially as the available memory bandwidth is hardly utilized fully by G200-based solutions.
The A2-revision G200 core is in its full configuration here: 240 unified shader processors, 80 texture processors and 32 raster back-ends. The GPU clock rates are increased relative to the reference card. The main GPU domain is clocked at 700MHz rather than at 602MHz while the shader domain frequency is increased from 1296 to 1400MHz. As opposed to the memory subsystem, the overclocked GPU makes a promise of a considerable performance growth in games that make wide use of the GPU’s computing resources. We estimate this performance growth at 5% to 12% but the gaming tests will show the precise numbers. Due to the complexity of the G200 chip, its display controllers are located in a separate chip called NVIO2. Nvidia used the same solution earlier with the G80 chip.
The graphics card offers two dual-link DVI-I ports and a universal 7-pin port for analog video output (YPbPr, S-Video, Composite). This sample of the GeForce GTX 280 doesn’t support DisplayPort, which would require a special translator chip. The HDMI interface is supported by means of a DVI-I → HDMI adapter (included into the box). You can enable audio-over-HDMI by connecting to the sound card’s S/PDIF output (the G200 chip doesn’t have an integrated audio core) using the included cable. The two MIO connectors covered with a rubber cap allow to join together up to three GeForce GTX 280 AMP! cards into a triple-SLI subsystem. Theoretically, such a subsystem is expected to deliver record-breaking performance. We’ll check out in an upcoming review if it really does. What we are sure of already, a triple-SLI subsystem is going to have a terrific power draw. We guess it would make a 1000W power supply a necessity.