The Nvidia RTX 30-series capacitor controversy
Some alternative thoughts.
If you’ve been reading up on RTX 30-series cards, you may have come across some alarmist reports of crashes, usually in situations where the card is clocked above 2,000MHz. Some attribute this to poorly validated designs involving the choice of capacitors located on the rear of the PCB behind the GPU. EVGA believes this is the case, while other manufacturers believe that capacitors are not the cause.
I don’t pretend to be an armchair electrician but sadly many youtubers and tech press seem to think they are. Some in the media provide great analysis and do a lot to explain what the issue ‘might’ be, however, devices like these are complicated, with most of the pertinent information hidden in low level schematics, specifications and IP that the media and end users will never see. Having said I’m not an expert in semiconductor design, allow me the indulgence of providing some thoughts.
All RTX 30-series cards feature an advertised boost clock, but for some reason, all the cards we’ve tested to date clock themselves above this rating. Is this aggressive self-overclocking more of a contributing factor? FYI none of the cards we tested presented any issue at all.
If a card comes with an advertised boost clock of 1,900MHz, why does it consistently clock above this? Why not have the boost advertised at 1,900MHz and let the user increase the power limit and overclock? Does any RTX 3080 ever operate at its base clock? Consider this: the i9-10900K features a maximum boost clock of 5.3GHz. If it tried to clock itself to 5.5GHz, would you be surprised if it crashed? Quite rightly, you’d be questioning why the motherboard is bugging like that. Forget about blaming the VRM or some other component, you’d assume it’s a BIOS bug.
The manufacturers claim they are following Nvidia’s design specifications. A cynic might say Nvidia wanted first wave reviews to look more impressive by allowing the cards to clock to the very edge of stability and beyond specification to artificially get benchmark wins over a possible AMD competitor. Then release a driver to lower performance and bring stability. I don’t think that’s the case – at least I hope not! Tin foil hat, off.
Of course, all this is pure speculation. We only have Nvidia and its partners’ statements to go by. The tech media use hedging language like ‘could be causing’ or ‘likely contributing’ but none of them really know the reason because they don’t have access to the kind of low level internal specifications or information that’s needed to accurately comment.
Reports of instability seem to have decreased with the release of a new driver. If this means keeping boost clocks nearer their stated specifications then that’s what should happen. If you want to overclock, increase power limits or overvolt your card to the edge of stability, then you should be free to. Can these GPUs consistently and reliably clock to 2,000MHz+? If not, why not? Is it the capacitors? Immature early driver? First wave Samsung 8nm quirks? Stressed multi rail power supplies? Time and widespread usage will reveal more but we’d expect things to evolve and mature quickly so we can get back to gaming without worrying about what’s going on under the hood.
“The manufacturers claim they are following Nvidia’s design specifications. A cynic might say Nvidia wanted first wave reviews to look more impressive by allowing the cards to clock to the very edge of stability and beyond specification to artificially get benchmark wins over a possible AMD competitor.”