Re­li­able com­put­ing ac­cel­er­a­tion for AI- and DSP-​Applications

Today AI ap­pli­ca­tions are dri­ving tech­nol­ogy de­vel­op­ment. In order to ful­fill per­for­mance and en­ergy re­quire­ments, in­no­v­a­tive so­lu­tions are needed. One im­por­tant pos­si­bil­ity is the use of in-​​memory-computing ar­chi­tec­tures for AI and gen­eral DSP ap­pli­ca­tions. IHP has con­sid­er­able know-​​how and ad­vanced tech­nol­ogy de­vel­op­ment for sys­tems based on RRAM tech­nol­ogy. We are in­ves­ti­gat­ing the use of RRAM cells and cor­re­spond­ing ar­rays in ac­cel­er­at­ing rel­e­vant math­e­matic op­er­a­tions such as ma­trix mul­ti­pli­ca­tion, em­u­lat­ing neural struc­tures, and per­form­ing sig­nal pro­cess­ing al­go­rithms in ana­log do­main. In order to ef­fec­tively per­form these tasks, sig­nif­i­cant focus is on the mod­el­ling of the cells and ar­rays and re­li­a­bil­ity man­age­ment.