Originally posted by Developer12
View Post
Doing real design work is not that relevant especially considering the pace of evolution in the field. Media encoding/decoding is maybe a single exception. I'll repeat myself: So you have to an engineer to know what the engineers are doing?
The IC industry has changed a lot over the past ~decade. Highlevel CPU designs can change and adapt quickly, thanks to the investment into programmatic designs and sophisticated tooling. Take a look at the difference between Zen(+) and Zen2... The designs are worlds apart. Again, yes a manager typically won't design low level components, however this is where I disagree: A good manager would know and understand the advantages and disadvantages of many different designs and design configurations.
You don't even need an engineering background to obtain valuable trade secrets, high-school math/science and some experience is enough. A low level trade secret like branch prediction training is probably not something a manager (without an engineering background) would completely understand. Additionally you can't just copy the results directly from one design to another... On the other hand high-level data-centrality (e.g. Infinity Cache) practical R&D costs an insane amount of resources. I won't not be surprised if it costs more than something as challenging as branch predictors (which is a lot cheaper to validate). Data-centrality is a beast that spans over multiple fabs and packaging tech. Almost anyone can understand the results once you have found what works well.
Comment