Care to explain what that "very simple and obvious" path to success is?
Sure, look at what AMD did in the mid-teens by hiring Jim Keller and having him design the Zen architecture. AMD managed this while teetering on the precipice of insolvency. Yes, they bet the company on this architecture but Lisa Su is a computer engineer and knew that Jim Keller is the best silicon designer in the world. He led the team that invented AMD64 (now known as x86-64) so he definitely deserves that moniker. The Zen architecture has a feature that Core does not (chiplets) and its implementation of SMT is superior.
Then, just to twist the knife, AMD created the multi-generational platform (AM4), inarguably the greatest x86 platform ever created. AMD leveraged economics in their silicon production by using chiplets and essentially created a captive market with AM4. I say a captive market because once you were on AM4, it made no financial sense whatsoever to buy an Intel CPU (for at least 99% of users anyway) so Intel was essentially locked out of that part of the market for the duration of AM4's life.
As good as all of this was for AMD, what really put them over and crushed Intel was the release of EPYC because the data centre is the bread-and-butter of any silicon manufacturer with the consumer market just being the icing on the cake. Intel had literally
nothing that could properly compete with EPYC and to this day, they still don't.
The very simple and obvious path to success is to assess the market by looking at what advantages potential customers aren't getting right now that would be beneficial to them, advantages that they desire. Then simply provide those advantages in a package that is affordable.
The problem has been this:
With AMD so cash-strapped and close to insolvency, they had to become laser-focused on the AMD side of the business because it was the CPU side where they were really lacking. Since, at the time, being the leader in data centre CPUs was the key to survival, their focus on the AMD side was 100% justified. Couple that with the fact that their CPU architecture at the time (FX) was completely outclassed by Intel when looking at performance-per-watt, AMD had to completely re-work their CPU design and philosophy.
At the time, the ATi side of the company wasn't in dire straits because the GCN architecture is one of the greatest GPU architectures that we've ever seen when it comes to longevity and scalability. It was kind of a reverse of today's situation with the ATi side propping up the AMD side. So, with pretty much all of the R&D money going to the AMD side, the ATi side began to stagnate. Then Raj Koduri came along and, with not a lot of R&D funding to work with, his only option was to try and extend GCN's life even further, even though, by then, GCN was essentially obsolete. That's how we got the debacle known as Vega.
What was needed was for a new architecture to replace GCN (the way that RDNA did) but AMD's executives had become so used to the AMD side needing the lion's share of funding that they just assumed that the ATi side would be ok being underfunded for another generation. Clearly, they were wrong, but you know how corporates are, they're convinced that they know everything and they really screwed up. Raj Koduri was much-maligned but I think he did the best he could with what he was given and he wasn't given much. I believe this because he now works with Jim Keller and if Jim Keller believes in him, then I believe that we were wrong to think that he was just incompetent.
In the GPU space, AMD corporates probably don't care much about RDNA because their bread-and-butter is their data centre CDNA, just like nVidia's bread-and-butter is their AI stuff.
The first thing that AMD will need to make the ATi side really competitive with nVidia is a strong desire to do so. Right now, I think that they're just making sure that it continues to survive instead of trying to make it thrive. That's called "corporate logic" and it's an oxymoron.