29-09-2020, 04:11 PM
The ongoing problems with GeForce RTX 30 series launch are reaching new grounds. From a very limited stock, sponsored videos ahead of independent reviews to actual hardware issues with the very few cards that were already delivered – this launch will definitely not going to be positively remembered.
We recently reported on a problem that had arisen in the last few days and it seems that we have only touched upon an issue that is rapidly escalating. To make things worse, NVIDIA has still not officially acknowledged the issue.
The problem
The issue is likely related to the capacitors installed at the back of the GPU. In this generation (GeForce RTX 30), this is actually very easy to see as the most manufacturers did not cover this area with a backplate. There are six necessary bottom capacitors that are responsible for filtering NVVDD/MSVDD GPU voltages. The better the filtering the less likely the card is to encounter issues at high frequencies (ie factory-overclocking).
The problem was first reported when custom boards were reaching 2.0 GHz+ clock speeds during gaming. This resulted in a sudden crash to desktop issues without any warning. The worse the voltage filtering and the higher the clock speed of the custom design was, the more likely users were to encounter this problem.
The issue is mostly related (at least from what is being reported now) to the choice of capacitors. AIBs need to follow NVIDIA guidelines to use either POSCAPs (Conductive Polymer Tantalum Solid Capacitors) – marked red on the picture below – or MLCCs (Multilayer Ceramic Chip Capacitor) – marked as green on a picture below. A combination of both is also possible and actively used on many cards (including Founders Edition). The under-specified guidelines could also be a culprit of the problem, because AIBs rely on those documents to the letter, especially when proper testing is not possible before launch.
It is worth noting that Buildzoid (video attached below) claims that these are not POSCAPs (in fact none of the RTX 30 cards use them). These are called SP-Caps.
We recently reported on a problem that had arisen in the last few days and it seems that we have only touched upon an issue that is rapidly escalating. To make things worse, NVIDIA has still not officially acknowledged the issue.
The problem
The issue is likely related to the capacitors installed at the back of the GPU. In this generation (GeForce RTX 30), this is actually very easy to see as the most manufacturers did not cover this area with a backplate. There are six necessary bottom capacitors that are responsible for filtering NVVDD/MSVDD GPU voltages. The better the filtering the less likely the card is to encounter issues at high frequencies (ie factory-overclocking).
The problem was first reported when custom boards were reaching 2.0 GHz+ clock speeds during gaming. This resulted in a sudden crash to desktop issues without any warning. The worse the voltage filtering and the higher the clock speed of the custom design was, the more likely users were to encounter this problem.
The issue is mostly related (at least from what is being reported now) to the choice of capacitors. AIBs need to follow NVIDIA guidelines to use either POSCAPs (Conductive Polymer Tantalum Solid Capacitors) – marked red on the picture below – or MLCCs (Multilayer Ceramic Chip Capacitor) – marked as green on a picture below. A combination of both is also possible and actively used on many cards (including Founders Edition). The under-specified guidelines could also be a culprit of the problem, because AIBs rely on those documents to the letter, especially when proper testing is not possible before launch.
It is worth noting that Buildzoid (video attached below) claims that these are not POSCAPs (in fact none of the RTX 30 cards use them). These are called SP-Caps.
![[+]](https://leakedbb.com/images/collapse_collapsed.png)