Ryzen 9000 Series – Everyone Should Relax
Ryzen 9000 Series – Everyone Should Relax
It was definitely a disappointing start since not all promised features arrived as expected. However, we did see some positive changes in performance—just a small improvement—but we used less power, received new data, and got valuable feedback from key platforms. I prefer a launch that falls short but opens the door for real progress over one that fails completely and harms our reputation or morale.
You're right, I did mention the launch was a joke and a complete failure. I wasn't trying to defend AMD's approach. Just to clarify, the new CPUs aren't as bad as the reviewers portray. It's unusual since most problems stem from the extremely low default TDP. It's clear even AMD isn't certain about the perfect balance between power and performance in their products, especially with the 9700X.
I never tried putting words in your mouth, just shared my thoughts. Another idea I had was to build software focused on power efficiency, similar to how fans are designed—so they run more efficiently or at least make it easier to adjust. In my opinion, defaults and overclocking don’t work well; we need tools that let you set the power draw to whatever voltage suits your task best.
It's a solid concept and you're not alone in feeling uncertain. Just because we read online doesn't always give us the full picture, so don't stress too much. People usually don't think as clearly from screens.
The biggest problem that AMD tends to have is that they release slides and performance numbers that simply dont exist in their launch events, or that they are so misleading that they should be held liable for it (see 13700k vs 5900xt comparison slides using a 6600xt gpu) AMD has acted like its a huge new changes and it would make a huge difference, except it really has not made much of an impact yet. Yes the 9700x is quite efficient, but so is the 7700, and the 7700x in Eco mode can also be pretty much similar in efficiency and performance. Are the CPUS bad? Absolutely not, but when you dont keep these companies honest and making sure they actually put in the effort to get actual performance increases with a new release, you get the stagnation we have had to deal with for Intel and NVidia, and now AMD. These CPUS are simply marginal improvements for all that work and R&D that they cost to make, that really does not instill much confidence in AMD for the future cpus, or that matter GPUS as they have shown that they are not capable of making huge strides for their GPU architectures, just a few steps are a time. We are still waiting for AMD GPUS to have their Zen 3 moment.
I only glanced at the long passage in OP. From my view, Zen 5 stands out as the top x86 microarchitecture currently available, which isn’t too surprising given its newer design aimed at higher performance. I think AMD invested in improving it for enterprise tasks to strengthen their position in the market. The increase in AVX-512 resources suggests they’ve surpassed Intel in that segment. The chiplets used in consumer CPUs are identical to those found in EPYCs and similar devices, but consumers are benefiting from this shift. APUs intended for desktops without server roles don’t gain from these improvements.
What confuses me the most is how predictable it was for engineers and leaders at AMD to reduce their CPU performance by around 20% while setting such a minimal power cap—something that seems irrelevant to the industry’s priorities.
I understand this partially. A 20% improvement at the same power use is quite notable after two years. We should also consider that shrinking transistors is becoming increasingly challenging without major obstacles. I once forecasted this technological slowdown nearly ten years ago, when TSMC engineers shared insights about the difficulties they anticipated with 5nm technology. Today we see similar trends from Intel, AMD, and Nvidia, showing smaller gains over time. It’s clear that progress can’t match what it was a decade or even a few years back. Still, I fully agree with you about AMD’s claims regarding performance improvements—those numbers seemed unreal.
The issue lies in mixing logic with executive decisions. When engineers are involved, it's a different matter, but with executives it often leads to problems. A blend of the dunning Krueger effect and understanding what the people want works best. Case in point: case and point copilot, most fan control tools come from gaming companies, and if executives are part of the mix, things usually go wrong.
It's basically an explanation. Understanding the situation needs some context, but I get your point. Sure, AMD seems to be shifting focus toward data centers, but using a consumer-grade CPU as an example doesn't really add up. It's confusing, especially for even the most dedicated hobbyists who might quickly spot the problem. How could someone in R&D make such a mistake?