Get the most out of the overclocking reserve with the Nvidia scanner
The Founders Edition cards are already overclocked beyond Nvidia's basic specification, but they aren't trimmed to their individual "cotz limit" in the GPU lottery. Ex works, as always, applies: In order to define a clear frequency range and a typical maximum GPU boost rate, all delivered GPUs must be able to process at least the same clock rates safely. From then on, however, each GeForce RTX will face a different limit before becoming unstable. This very GPU-dependent upper limit even changes depending on the load. Some enthusiasts even consider it a matter of honour to reach exactly this threshold, often using all the possibilities of overclocking tools very time-consuming. Incidentally, this also seems very familiar to me.
Nvidia is now doing everything it can to prevent individual overclocking with an API/DLL package that partners such as EVGA and MSI can integrate into their utilities. Instead of jumping back and forth, testing part of the frequency/voltage curve at once and adjusting it to the stability of different workloads, Nvidia's scanner performs a special arithmetic routine in its own process. Although Nvidia says that the metric usually encounters mathematical errors before it crashes, the fact that it is included means that the algorithm can easily recover itself when such a crash occurs.
This gives the tuning software the ability to increase the voltage and try the same frequency again. As soon as the scanner reaches its maximum voltage setting and encounters a last error, a new frequency/voltage curve is calculated based on the known results. From start to finish, the process reportedly takes less than 20 minutes.
Interestingly, the functionality of the Nvidia Scanner is not limited to geforce RTX video card owners. The company says it will (later) support older boards (although it is not known how far this backward compatibility will extend). By the way, I took a few photos during the demo, which tactfully go a little further than what is shown in the slide:
- 1 - Einführung und Vorstellung
- 2 - TU102 + GeForce RTX 2080 Ti
- 3 - TU104 + GeForce RTX 2080
- 4 - TU106 + GeForce RTX 2070
- 5 - Performance-Anstieg für bestehende Anwendungen
- 6 - Tensor-Kerne und DLSS
- 7 - Ray Tracing in Echtzeit
- 8 - NVLink: als Brücke wohin?
- 9 - RTX-OPs: wir rechnen nach
- 10 - Shading-Verbesserungen
- 11 - Anschlüsse und Video
- 12 - 1-Klick-Übertaktung
- 13 - Tschüss, Gebläselüfter!
- 14 - Zusammenfassung und Fazit
Kommentieren