AMD's new Zen 5-based Ryzen 9000 CPU issues: leaker says it needs a 'reboot'

AMD's new Zen 5-based Ryzen 9000 series CPUs are officially here... but it's not the party we expected. There are some glaring issues and questions about Granite Ridge, where in some cases the new Zen 5 chips are SLOWER than Zen 4 chips.

In a new video from leaker Moore's Law is Dead, who compiled a bunch of very good questions I'd like to see answers from AMD on, as well as pulling some quotes from the tech media that have reviewed AMD's new Ryzen 9000 series "Granite Ridge" processors.

The weird 'software issues' that come with Zen 5, with MLID pointing out that Tom's Hardware said: "AMD needs to fix its drivers. The situation with the chipset drivers making irreversible changes to the operating system is nonsensical and problematic for end users and reviewers alike. AMD has known about this issue for 16 months, and despite endless public complaints from end users and the press, it hasn't been addressed".

In specific benchmarks like SPEC2017, AnandTech's review of AMD's new Ryzen 9 9950X and Ryzen 9 9900X does indeed show what AMD claimed: an 11% integer performance uplift over the 7950X, and a bigger 24% improvement in floating point performance over the 7950X. That's not bad, but you and I aren't running SPEC2017 on the daily, are we?

• Read more: AMD Ryzen 7 9700X CPU Review
• Read more: AMD Ryzen 5 9600X CPU Review

Cinebench 2024 multi-core results are impressive to see with the Ryzen 9 9950X with PBO (Precision Based Overclocking) that has a multi-score of 2340, compared to the stock 9950X with 2208, and up from the 2061 on the 7950X (PBO) and 2047 with the 7950X at stock. That's a 14% uplift from the 7950X (PBO) up to the 9950X (PBO). Not bad.

But... as MLID points out, according to Tom's Hardware's data, the new Ryzen 9 9950X with PBO enabled is only 7% faster than Intel's flagship Core i9-14900K... but AMD itself said in its own slides before launch that the 9950X is 21% faster than the Core i9-14900K in Cinebench 2024.

AnandTech went a little deeper, pointing out in its review of the 9950X and 9900X that the 9950X has an average latency of 180ns between the CCDs, which is "twice the cost of the previous generation". AT continued, saying: "Making all this the more confusing, Granite Ridge (desktop Ryzen 9000) reuses the same IOD and Infinity Fabric configuration as Raphael (Ryzen 7000) -- all AMD has done is swap out the Zen 4 CCDs for Zen 5 CCDs. So by all expectations, we should not be seeing significantly higher inter-CCD latency here".

• Read more: AMD delays Ryzen 9000 series 'Zen 5' launch, 'QA issues' with first Zen 5 chips
• Read more: AMD Ryzen 9 9950X already OC'd to 6.6GHz on LN2, hits 55K in Cinebench R23

Further diving into the issues with Zen 5, MLID points out the issues that Ryzen 9000 series CPUs have with Windows, core parking, and more. But Phoronix tested on Linux, where the new Zen 5-based Ryzen 9000 series CPUs really stand out, like truly. Top of the charts, beating the pants off Intel all day long... on Linux.

MLID continued, saying: AMD launched the new generation CPUs without finished software in Windows, or maybe even fully working microcode... and when I asked that question to the first person I was able to get a hold of, this person basically told me... yes... well, they don't specifically in the case of Zen 5, what they are seeing here looks a lot like the microcode just isn't finished. This is normally what you'd see out of early engineering sample benchmarks before it was completely done".

MLID reached out to a source at AMD, who told him: "So, the first thing that people should know is that the Zen 2 team was the first team to be in charge of the Zen 5 architecture, and they insisted on using their Zen 2 codebase instead of starting with what was built by the Zen 4 team because they were"more familiar with it".

The source continued, saying: "second, Zen 5 changed ownership between team several times throughout its development. This lead to many delays and a lot of unnecessary rework. And finally, I overheard that marketing didn't communicate details about Core Parking to Reviewers until 5 days after they received test samples, and just in general marketing was dropping the ball on communication".

"Do you see the picture I am painting? I didn't work on Zen 5, but from what I can gather this project made key mistakes from the start, was plagued with problems throughout its development, and ultimately it looks like they just rushed the product out the door before the software was finished".