Test protocolFor this test, we used the protocol we introduced in our report on graphics cards and thermal characteristics, which allows us to make a detailed analysis of the various choices AMD’s partners have made here.
We measured the temperature of the GPU and various components under actual usage conditions, in an Antec Sonata 3 box and we also took an infrared shot, opening the box for as short a time as possible. We also measured noise levels under actual usage conditions, namely with the cards in a closed casing. We placed the sonometer 60cm from the casing.
We altered the protocol slightly by adding an additional phase because we carried out the temperature test for each card at maximum overclocking. This enabled us to observe the power stage when put under increased stress, but also allowed us to verify overclocking stability further.
When several overclocked variants of the same model are available, we tested just one of them. There’s no need for a detailed analysis of the performance of the different solutions because all the cards would give identical performances at identical clocks. This is therefore dealt with briefly in the report.
We concentrated on the most common models and, for now, we’re able to present results on eight of them, in addition to the reference card. We will of course add models in the future if need be.