
AMD Puts Hope in Packaging, Memory, Logic and Optical Communications for the Next Decade
At ISSCC 2023 earlier this week, AMD mentioned the way forward for computing within the subsequent decade. CEO Lisa Water head server and confirmed that AMD has carried out admirably on supercomputer, server, and GPU efficiency tendencies over the previous few many years. Nevertheless, the well-crafted plans are most likely extra fascinating, displaying how AMD goals to maintain the pedal within the steel and use superior applied sciences to keep away from diminishing semiconductor course of shrinkage benefits.
Within the efficiency slides above, AMD claims to have efficiently doubled total server efficiency each 2.4 years since 2009. Nevertheless, it doesn’t share any projections concerning this market. AMD is assured to take a more in-depth look into the long run with GPU efficiency tendencies (slide 2 within the gallery above). You may see right here that it claims to double GPU efficiency each 2.2 years since 2006. The chart reveals that this pattern stabilizes till a minimum of 2025.
AMD’s supercomputer efficiency is probably the most profitable by way of developments over time, and the final chart above reveals that for the reason that late 90’s AMD processors have been instrumental in doubling supercomputer efficiency each 1.2 years. As well as, AMD predicts that we’ll attain Zetascale supercomputer efficiency in about ten years. AMD additionally took the time to spotlight the effectivity positive aspects and the extreme wrestle to maintain Moore’s Regulation alive whereas the logic density is diminished.
AMD has generously laid out a few of its key plans to assist it transfer ahead with effectivity and efficiency positive aspects over the subsequent decade. In accordance with AMD, superior packaging shall be a powerful driver of each efficiency and effectivity. We have already seen some outcomes from AMD going this fashion with the usage of chipsets and 3D V-Cache, and that may proceed.
Some superior packaging avenues to discover embrace 3D CPU and GPU silicon integration “for next-level effectivity.” Moreover, AMD believes that “even tighter integration of computing and reminiscence” will allow greater bandwidth at decrease energy. AMD will even goal processing in reminiscence. A slide shared on ISSCC confirmed a processor with an HBM module on it. AMD says that with the ability to run key algorithmic cores in reminiscence will take away vital load and latency from the system.
One other huge goal for effectivity financial savings and subsequently potential efficiency positive aspects is chip I/O and communication. Particularly, utilizing optical communication tightly built-in with the computing die is predicted to supply a big productiveness acquire.
AMD has additionally taken a while to brag in regards to the AI ​​efficiency positive aspects delivered by its processor portfolio over the previous decade. The presentation mentioned some use circumstances for AI computing and highlighted the potential efficiency positive aspects AI can present for simulations.
After all, AMD is not the one one inspecting the advantages of superior packaging, chips, die stacking, in-memory computing, optical computing, and AI acceleration. Nonetheless, it is good to see it has stable plans for fierce competitors and is producing chips for fans sooner and extra environment friendly than ever earlier than.
#AMD #Places #Hope #Packaging #Reminiscence #Logic #Optical #Communications #Decade