The need for deep neural network (DNN) models with higher performance and better functionality leads to the proliferation of very large models. Model training, however, requires intensive computation time and energy. Memristor-based compute-in-memory (CIM) modules can perform vector-matrix multiplication (VMM) in situ and in parallel, and have shown great promises in DNN inference applications. However, CIM-based model training faces challenges due to non-linear weight updates, device variations, and low-precision in analog computing circuits. In this work, we experimentally implement a mixed-precision training scheme to mitigate these effects using a bulk-switching memristor CIM module. Lowprecision CIM modules are used to accelerate the expensive VMM operations, with high precision weight updates accumulated in digital units. Memristor devices are only changed when the accumulated weight update value exceeds a pre-defined threshold. The proposed scheme is implemented with a system-on-chip (SoC) of fully integrated analog CIM modules and digital sub-systems, showing fast convergence of LeNet training to 97.73%. The efficacy of training larger models is evaluated using realistic hardware parameters and shows that that analog CIM modules can enable efficient mix-precision DNN training with accuracy comparable to full-precision software trained models. Additionally, models trained on chip are inherently robust to hardware variations, allowing direct mapping to CIM inference chips without additional re-training.

Consumers spent $10.6B on digital games in October 2020, up 14% year-over-year. Consistent with ongoing trends, console spending grew the fastest, with earnings up 18% over 2019. This was especially impressive growth given that Call of Duty: Modern Warfare launched in October 2019 and many of the biggest titles of the 2020 holiday season were not released until November. Earnings on other platforms rose as well, with mobile up 15% and PC up 10%. The title, which was released on September 28 on mobile, PlayStation 4 and PC, is an unprecedented international success for a game made by a Chinese developer.

Digital games earned $10.7B in September 2020, up 14% year-over-year. Games earned more across all platform types than during the same period in 2019. Mobile revenue was up 9%, PC rose by 8% and console earnings increased 40% as major fall titles began to hit the market. NBA 2K21 was the top-earning title of the month, selling 1.9M digital units. Digital console sales were up 19% over NBA 2K20 at launch, and in-game revenue similarly increased by 8%.

Ghost of Tsushima sold 1.9M digital units in July to become the fastest-selling new PlayStation IP. The game far outpaced the early sales of other new franchises launched on PlayStation 4: Horizon: Zero Dawn sold 1.9M units in its first two months on the market in early 2017, and Days Gone sold 1.3M units in its first two months after launching in late April 2019. Ghost of Tsushima also set the record for the biggest console game ever launched during the month of July. The Last of Us Part II, also a PlayStation exclusive, set the June record only one month prior. A lack of alternative entertainment options due to COVID-19 has contributed to the success of recent PlayStation titles during what is historically a quiet season for new game releases.

Overall earnings were up 3% from May ($10.2B) and 9% over June 2019 ($9.6B). Both PC and console revenue were down from their 2020 peaks set in March. This was offset by growth in mobile revenue, which hit an all-time high. Mobile game spending typically sees a boost in the summer months, and this seasonal growth was likely further bolstered by the limited availability of other entertainment options due to COVID-19. The Last of Us Part 2 sold 2.8M digital units in June, easily setting the record for exclusive PlayStation launches.