Posts Tagged “adrian”

SLI scaling analysis with GTX 780

In this SLI scaling analysis, I have selected a number of different variables to test, which all come into play with SLI scaling in games. The cards used to test are Nvidia GeForce GTX 780's (yes they are not the most powerful cards, but still up to the task, and this is about scalability, not brute power). These are tested in single, dual and triple card configurations, in a variety of different games. Settings were chosen which work well with a dual card configuration, designed to maximise all available graphics settings whilst still being playable. This allows the cards to be pushed extremely hard, which also allows the different variables to show their effects on scaling, while producing a more accurate result. My definition of playable is not an arbitrary number of frames, but the subjective user experience of the game feeling smooth enough to be an enjoyable experience. The following variables have been chosen to assess their impact on fps, perceived smoothness i.e. with microstuttering etc and to discover any bottlenecks which may impact on 2-way and 3-way SLI, for instance: is a quad core hyperthreaded processor capable of keeping up with 3 demanding cards, I've read many reviews that have often made assumptions about where the bottleneck (if any) is occuring, the data presented here will hopefully shed some more light on these factors. |--- | **What we're testing** |:- | Frame rates; minimum, average and maximum, of course the minimum and average fps is more important to playability. | Each graphics cards maximum temperatures during play. | Each GPU's load. | Each GPU's core clock rate. | The amount of gpu memory used. | The CPU's load and frequency (frequency is shown as a % of the CPU clock rate, which is at 4.4Ghz)Read more...