Overclocking, Consumption and Thermals
Geforce GTX 580 packs plenty of OC potential, with many partners jumping on the OC ship. While the reference GTX 480 runs at 772MHz for the GPU and 1002MHz for the memory, our factory overclocked Gainward GTX 580 1536MB Golden Sample runs at 805MHz for the GPU and 1050MHz for the memory.
This will be enough for about 4% better gaming results, but it must be said that there’s additional room for overclocking. In fact, our further overclocking resulted in 11% better performance when compared to the reference card.
Overclocking GTX 580 cards will depend on core voltage and overclocking results will vary from card to card. Gainward team leaves for sure room for additional overclocking on this card, and the good thing about Golden Sample cards is that GPU voltage has been increased by default, which will help with overclocking results.
We recently tested a reference GTX 580 (1025mV default voltage) and managed to overclock the GPU to 855MHz, but we didn’t mess with voltages or change the fan speed from AUTO mode.
Gainward pushed the default voltage to 1062mV, resulting in increased stability at higher clocks. Despite the higher than reference voltage, we didn’t manage to push Gainward’s GTX 580 GS beyond 845MHz.
Further overclocking isn’t possible without manual voltage changes. We managed to hit 875MHz for the GPU and 1110MHz for the memory (4440MHz effectively) by pushing the voltage to 1138mV.
The fan isn’t too loud during intensive gaming and we can finally say that we’re pleased with the noise levels. Speeding up the fan didn’t contribute to overclocking headroom.
Higher factory clocks at the same time mean that the card will consume more compared to the reference card.
FurMark test is not quite favored by Nvidia’s driver and you’ll have to manually disable throttling on the card or else it will downclock on its own. Compared to our system with a reference GTX 580, our GTX 580 Golden Sample consumption measurements show about 12W higher consumption. Note however that after testing three GTX 580 cards, we realized that our reference GTX 580 consumes higher than average.
We downclocked and lowered the voltages on the other two GTX 580 cards we received from Nvidia’s partners and measured far lower consumption than with the reference card provided by Nvidia.
The following results were recorded after we disabled Nvidia’s throttling mechanism.