Be able to get hold of a GeForce RTX 3080 last week was very complicated. Those who have succeeded have sometimes encountered problems after a few hours of use, a priori rather with cards manufactured by partners of Nvidia. The first lines of inquiry suggested that it was the pilots who were at fault: a poorly scaled algorithm and the card was put into error.
But it looks like the problem may well be material. However, Nvidia is still investigating and no findings have been made yet. For their part, very specialized sites are looking for answers and might have found the beginning of a trail.
A history of capacitor category
Thus, according to Igor’sLAB, it appears that manufacturers do not all use the same capacitors to regulate the power supply to the components, including the GPU, on the card.
To put it simply, Nvidia and a brand like EVGA (for example) use a mix of two very good components (capacitors) and have them paired in some way. The best possible to ensure good performance and a correct price level if we are to believe the statements made by one of the spokespersons of EVGA on the brand’s forum.
In the case of Asus, only one type of these components is used – the best possible – and the layout has been completely revised and optimized which would guarantee as good stability as the Nvidia or EVGA models. MSI would have worked, for its part, on another solution, but which also seems to work according to our colleague.
By observing the cards of Zotac and other partners of Nvidia, not cited by the site, it is capacitors of the same nature but not at all made of the same wood which are soldered to the PCB.
On the one hand, they are less expensive, and, on the other hand, they would have difficulty ensuring a certain cleanliness of the electrical signal, a cleanliness however requested by the RTX 3080.
On paper, they are supposed to provide the same services but in the heat of the moment, they could well fail, especially when the GPU Boost of the card reaches the bottom of 2 GHz, the threshold beyond which the first concerns appear among the players.
If it is indeed the components which are in question, then it will be necessary that the brands which use them make sure to limit the increases in frequency of the GPUs. They will have to push firmware updates that will reduce frequencies, while limiting power consumption. However, players could ask for compensation, especially if the operating frequencies are below those officially announced by the brand on the market.
As for taking back all the cards already marketed, it is impossible. Shake up the production lines as well. They are already having enough trouble supplying the market with RTX 30 series GPUs. It will be easier and less expensive to reprint packaging.