Comparison Cards: When 1 GB Is Greater Than 1.8 GB
Standing in for the reference design 1 GB GeForce GTX 285 graphics card is our review sample from XFX. Originally overclocked to 670 MHz on the GPU, 1,508 MHz shaders, and GDDR3-2500, we dropped it to the same 660/1,505/2,400 settings of Gigabyte’s 2 GB card to more accurately evaluate the effects of Gigabyte’s memory increase.
Now that we had “one model down” from the 2 GB GeForce GTX 285, we wanted to go “one model up,” to see exactly where the card fits in the performance market. Since the GeForce GTX 285 is Nvidia’s best GPU, it would take a two-GPU card to beat it. Today’s test system came equipped with another XFX card, its single-PCB GeForce GTX 295.
Equipped with a total of 1,792 MB GDDR3-2016 memory, each GPU gets exactly half that amount. Each GPU uses the same data while sharing the graphics load, making the card act as a single 896 MB graphics unit.
The same issue occurs even in 3-way and Quad SLI systems, where the total amount of useable graphics memory is the same as that of one GPU. Thus, even if we were to put three 2 GB GeForce GTX 285 graphics cards in a system, it would act as a triple-GPU 2 GB graphics unit rather than a 6 GB array. Given this important clarification, splitting 1,792 MB across two graphics processors results in less application memory than a single-GPU 1 GB card.