AMD Blitzes Computex With AI, From Mobile To Desktops And Data Centers
How does the Ryzen AI 300 Series compare in terms of performance with AMD's previous generation mobile CPUs?
The Ryzen AI 300 Series mobile processors from AMD represent a significant leap in performance compared to the company's previous generation mobile CPUs. The new series is built on the Zen 5 and Zen 5c CPU architectures, which offer an average of 16% improvement in instructions per clock (IPC) over the Zen 4 architecture. This translates to better single-threaded and multi-threaded performance.
In terms of AI performance, the Ryzen AI 300 Series processors feature a new XDNA 2 NPU that offers 50 trillion operations per second (TOPS), a significant increase from the 10 to 16 TOPS offered by the previous generation's NPUs3. This makes the Ryzen AI 300 Series the most powerful laptop SoC for AI workloads, surpassing both Qualcomm's Snapdragon X Elite and Intel's upcoming Lunar Lake chips.
The new processors also come with an upgraded RDNA 3.5 integrated graphics engine, which offers up to 16 compute units, an increase over the previous generation's maximum of 12 compute units. This results in better graphics performance for thin-and-light laptop designs.
The Ryzen AI 300 Series processors also have up to 12 high-performance cores and support for 24 threads, along with up to 50% more on-chip L3 cache memory compared to the previous generation Zen 4-based chips.
Overall, the Ryzen AI 300 Series mobile processors offer significant improvements in performance, AI processing throughput, and graphics performance compared to AMD's previous generation mobile CPUs4.
What are the key features of the newly announced Ryzen AI 300 Series processors by AMD?
The newly announced Ryzen AI 300 Series processors by AMD come with several key features:
-
50 TOPS NPU: The processors are equipped with a new 50 TOPS (Tera Operations Per Second) rated Neural Processing Unit (NPU), which is a significant upgrade from the previous generation. This enables the chips to handle AI-related tasks such as running generative AI models locally.
-
Zen 5 and Zen 5c Architecture: These mobile CPUs are built on AMD's new Zen 5 and Zen 5c CPU architectures. The Zen 5c architecture is a variant of Zen 5 that is optimized for small size, similar to Intel's E-cores.
-
High-Performance CPU Cores: The Ryzen AI 300 Series processors come with 10 or 12 high-performance CPU cores, supporting 20 and 24 threads, respectively. This is a significant upgrade from the previous generation, which had a maximum of eight CPU cores.
-
Increased On-Chip L3 Cache Memory: The new processors come with up to 50% more on-chip L3 cache memory compared to AMD's previous Zen 4-based chips.
-
Support for Copilot+ AI PCs: The Ryzen AI 300 Series processors surpass the requirements for Copilot+ AI PCs, offering a 3X performance lift over AMD's previous generation mobile CPUs.
-
Improved Graphics: The new processors come with upgraded RDNA 3.5 integrated graphics with up to 16 Compute Units (CUs). This is comparable to a desktop RX 6500 XT and is an increase over the previous generation's maximum of 12 CUs.
These features make the Ryzen AI 300 Series processors a powerful choice for AI-related tasks, content creation, and productivity in thin and light laptop designs.
What advancements does the Zen 5-based Epyc data center CPUs, code named Turin, bring in terms of core density and performance?
The Zen 5-based Epyc data center CPUs, code named Turin, bring significant advancements in terms of core density and performance. Turin processors will feature up to 192 brand-new Zen 5 CPU cores, which translates to 384 threads with SMT, in a single SP5 socket. This marks a substantial increase in core density compared to the previous generation.
The Turin chips are designed to be drop-in compatible with existing SP5 platforms, which will facilitate their faster ramp and adoption by cloud giants and server makers. In addition, these next-generation EPYC CPUs are expected to deliver more than 96 cores and a more versatile memory subsystem.
AMD's Zen 5 architecture is built on a 3nm process, allowing for a higher transistor count and enabling the company to maximize the number of CPU cores placed on a single chip. The Turin chip on display is thought to be a Zen 5c configuration, which uses CCDs designed around AMD's compact Zen 5 core configuration. This approach trades off per-core performance in favor of allowing more CPU cores overall, with lower clockspeed limits and less cache memory throughout the chip.
The 192-core variant of Turin may use twelve CCDs with sixteen Zen 5C "dense" cores each, offering a significant increase in core count compared to the previous generation4. Overall, the Zen 5-based Epyc data center CPUs promise to deliver impressive performance and efficiency increases, positioning AMD well to capture an even larger share of both first and third-party workloads in the data center market.