AMD has set its sights on releasing its next-generation Zen CPUs, known as the Ryzen 8000 series, in the coming year. However, the latest rumors surrounding these chips suggest that AMD is prioritizing multi-threaded performance over single-threaded performance. Although this claim may seem broad, it does make sense given AMD’s focus on server chips as the main revenue generator, rather than desktop CPUs for gamers.
AMD typically designs a single CPU core that is adapted for both server and client PC platforms. While the CPU core chiplets remain consistent across platforms, the additional application-specific chiplets, such as IO and memory controllers, are tailored to meet the specific requirements of each platform. Due to the greater financial rewards from server chip sales, AMD has allegedly adjusted the architecture of Zen 5 to favor server performance over client PC performance.
This shift in focus has resulted in a narrower gap between the maximum single-core clock speed and the maximum multi or all-core clock speed for Zen 5 compared to previous Zen chips. As a consequence, the single-thread frequencies may have been impacted, although not significantly. Engineering samples of the Ryzen 9 8950X, according to RedGamingTech, are reportedly achieving clock speeds of 5.6GHz, only 100MHz lower than the existing Ryzen 9 7950X. However, it’s crucial to note that these are engineering samples and not final silicon, so there’s room for potential improvement.
In addition to clock speeds, RedGamingTech also provides insight into Zen 5’s instructions per clock (IPC) performance. IPC measures the performance of a single software thread per clock cycle, regardless of the clock speed. The speculated IPC improvements from an internal slide allegedly showed a 10 to 15% increase. However, RedGamingTech suggests that these figures were based on a narrow measure of SpecInt, and the overall IPC gains across multiple integer and floating-point tasks may reach 20% or higher.
Assuming Zen 5 achieves similar peak clock speeds as Zen 4, it is expected to deliver approximately 20% more single-thread performance. However, the multi-thread performance gains are anticipated to be more significant, considering the narrowing frequency gap between single-threaded and multi-threaded loads. It can be argued that the emphasis is on boosting multi-thread performance rather than sacrificing single-thread performance. Nevertheless, if single-thread performance had been the priority, it could potentially have been even more impressive.
RedGamingTech also provides some interesting insights into the rumored Strix Halo or Sarlack, an APU that AMD is reportedly developing. This APU is said to feature a 16-core CPU and 40 RDNA 3.5-spec GPU compute units, along with a 256-bit shared memory bus. Notably, the shared memory bus is a novel addition for a PC-based APU, as it is more commonly found in console APUs. RedGamingTech claims that Sarlack can deliver comparable performance to an RX 6750 XT paired with a 16-core Zen 5 CPU, all within a combined power consumption of 75 to 125W. This achievement is noteworthy, considering that AMD’s highest-performance 16-core desktop CPUs alone consume 170W.
It’s essential to remember that all this information regarding Zen 5 and the rumored Sarlack is still speculative. As the release of new CPUs or GPUs draws closer, the quality of information tends to improve. Therefore, while these latest rumors provide intriguing possibilities, they are far from confirmed. As the development cycle progresses, it is likely that more concrete details will emerge regarding AMD’s plans for the Ryzen 8000 series.
Leave a Reply