AMD Zen 5 architecture deep-dive: How the Ryzen 9000 processors up the "power efficiency" game

Note: This feature was first published on 17 July 2024.

When AMD officially unveiled the Ryzen 9000 series processors last month at Computex 2024, I was surprised not because of the announcement itself – it was already an open secret that AMD will be launching new desktop processors – but that it will be launched as soon as this month, July. Much sooner than many of us had anticipated.

At the recent Zen 5 Tech Day held in Los Angeles, AMD took a deep dive with selected members of the media including HardwareZone into their upcoming Zen 5 architecture – the fifth generation of AMD’s Zen CPU architecture that powers its next-gen processors for desktops, mobiles and servers, including the Ryzen 9000 series and the mobile-focused Ryzen AI 300 processors.

The new Zen 5 architecture, which I got to see in action promises to be a significant leap forward. According to AMD's chief technology officer, Mark Papermaster, Zen 5 is a foundational shift, much like the original Zen architecture. This isn't just marketing fluff – there are some substantial changes under the hood.

One of the standout improvements is the new branch prediction system that features dual decode pipes. AMD says this enhancement boosts the throughput for the branch prediction engine, reducing latency and increasing accuracy. AMD has also expanded the number of dispatch and execution engines to an 8-wide design, allowing the processor to handle more instructions per clock cycle.

To support this performance boost, AMD has ramped up bandwidth throughout the processor. The L1 cache’s maximum bandwidth has doubled, including the FPU (floating-point unit) in each core. Additionally, the core now boasts six arithmetic logic units (ALUs), compared to four previously. These enhancements contribute to the claimed 16 percent increase in instructions per clock (IPC). For those of us who get into the weeds of CPU architecture, this is exciting stuff.

A particularly intriguing change is the new 512-bit data path for AVX instructions. Previously, AMD supported AVX-512 with two 256-bit paths, but now it's streamlined into a single 512-bit path. One of the ongoing debates between AMD and Intel processors revolves around AVX-512 support. In layman term, this translates into more efficient processing power. What’s interesting is Intel has notably omitted AVX-512 on its Raptor Lake and Meteor Lake processors, and it's doubtful we'll see it on the upcoming Arrow Lake desktops either.

Speaking of Intel: AMD didn't shy away from showcasing performance comparisons with Intel’s 14th-gen CPUs at the 2-day event. The 16-core flagship Ryzen 9 9950X, for instance, showed a 15 percent performance increase over the Ryzen 9 7950X and a 13 percent gain over Intel’s Core i9-14900K in Cinebench R23. AMD claims an even higher improvement of 21 percent over the i9-14900K in Cinebench R24. These numbers are promising, but I’ll leave my final judgement until I test the new CPUs myself.

Lower down the stack, the 6-core Ryzen 5 9600X in the mid-range category caught my attention. AMD suggests a performance boost of up to 94 percent in Handbrake compared to Intel’s equivalent, the Core i5-14600K. Very bold claims, and it’ll be interesting to see how that processor fares in our benchmarks. AMD also shared that performance gains for the 9600X compared to the i5-14600K in games range from five percent in Hitman 3 to 29 percent in Horizon Zero Dawn.

Even the 8-core Ryzen 7 9700X and the 12-core Ryzen 9 9900X show notable productivity gains, between 10 percent and 20 percent faster than Intel’s counterparts. Gaming performance improvements are also consistent, especially in titles like Horizon Zero Dawn.

But arguably there's something far more impressive about Zen 5 processors: their energy efficiency.

AMD continues its partnership with TSMC for manufacturing, but with the Zen 5 series, they're moving from the 5nm and 6nm processes used in Zen 4 to the newer 3nm and 4nm processes. In Zen 4, the CPU cores were crafted on the 5nm process, while the I/O die, which included an AMD Radeon IGP (integrated graphics processor), a PCI Express 5.0 controller, and a memory controller, was made on the 6nm process. At the event, AMD wouldn’t confirm that the Zen 5 processor cores will be made using the 3nm process and in turn, the I/O die the 4nm process, but it’s not too farfetched to assume that it will be similar to Zen 4’s setup since the I/O die architecture remains unchanged. It’s also a tried and proven design, after all.

The shift to these advanced manufacturing processes, however, has allowed AMD to reduce power consumption in the Ryzen 9000 series. According to charts shared by AMD, thermal resistance sees a 15 percent drop that also helps to reduce temperature by up to 7-degree Celsius. Crucially, this is achieved while maintaining the same TDP (thermal design power). It’s not clear if this improved thermal resistance is brought about due to changes in the manufacturing process alone or if other improvements are at play as well (perhaps AMD made a better heat spreader too?), but Ryzen 9000-series processors that have similar clock speeds to their direct Ryzen 7000 predecessors have lower TDPs and so should run cooler and more efficient overall.

If the new processors are indeed more efficient, they will be easier to cool and easier to fit into smaller casings with smaller AIO coolers too. I think this should be the direction of PCs, with flagship high-performance CPU and GPU processors being housed in small-form factor casings.

Beyond raw performance and power efficiency, AMD is also introducing several platform enhancements with the Ryzen 9000 series. A new BIOS update will enable memory speeds of up to 8,000 MT/s on AM5 motherboards. Additionally, the new Curve Shaper tool adds another layer of control over AMD’s Curve Optimizer, which can help overclockers quickly dial in an undervolt or overclock on the Ryzen 9000 CPUs.

So, where does that leave us? With all these performance promises and architectural improvements, the Ryzen 9000 series looks set to be AMD’s most power efficient CPUs yet. But without knowing the pricing, which AMD did not disclose at the event, it's hard to say just how competitive these new CPUs will be compared to existing ones in the market. But we won’t have to wait for long: AMD is shipping Ryzen 9000 series processors on 31 July.

We’ll also be receiving our review samples very soon and will put them through their paces in our benchmarks. Whatever information that AMD has shared with us certainly looks promising, and the previous two generations of Ryzen processors – the 5000 and 7000 series – have proven to be formidable choices for PC enthusiasts and gamers, and the 9000 series looks very likely to continue that trend.

Stay tuned.

Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.