The GeForce RTX 3080 was expected to be NVIDIA’s latest and greatest consumer graphics card so much so that it was quickly snatched up by bots when it launched. The consumers who were quick to get one, however, may now be regretting their haste. Reports about the card crashing under various conditions and heavy loads are flooding the Internet and a few tech experts and one graphics card maker have weighed in what could be the cause of this unfortunate mystery.
Like any piece of electronics, graphics cards make use of capacitors and a cluster of them are actually employed to regulate the voltage that passes through the card. When the capacitors fail to properly filter high-frequency garbage, graphics cards tend to become unstable and crash. That is probably why the common denominator among the reports is boosting the GPU clock to 2 GHz or higher.
igor’s LAB investigated several RTX 3080 cards as well as NVIDIA’s specifications and theorized that the kind and mix of capacitors used for this purpose may be the cause of instability. NVIDIA’s specification allows for two types of capacitors to be used, large-area and cheaper POSCAPs and smaller and more expensive MLCCs. Unsurprisingly, different manufacturers, including NVIDIA, used different combinations of those.
The most unstable combination is no combination at all, using POSCAPs for all six capacitor “slots”. Even one slot using MLCCs, which have to group 10 in a bundle, would be enough to stabilize the card as in the case of an MSI Gaming X Trio. NVIDIA’s own Founders Edition card used two blocks but the best of them all is probably ASUS’ TUF RTX 3080 which uses nothing but MLCCs.
The investigation in no way writes off POSCAPs entirely, just that may not be sufficient on their own to handle the GeForce RTX 3080’s higher loads. Unfortunately, since these issues are in the deepest levels of hardware, those who already bought affected cards have pretty much no recourse but to lock their cards down to lower, more stable clock rates.