AMD Strix Point Halo “55W” Ryzen APU Spotted, Strix Point “28W” Benchmark Leaks Out

wccftech.com:

Two new leaks of AMD's next-generation Strix Point APU family have emerged and they cover both 55W "Halo" & 28W "Standard" Ryzen chips.

AMD Strix Point APUs Leak Out: 55W Halo Variant For FP11 Platforms & 28W Standard Variant For FP8 Platforms

The latest leaks have been discovered by Harukaze5719 who managed to spot a brand new shipping manifesto that contains the next-gen AMD Strix Point "Halo" APUs. The shipment includes microprocessors with the OPN ID of "100-000001243" which is listed as a Strix "Halo" SKU. The chip has a listed TDP of 55W and is designed for the FP11 platform. Based on other details, the chip is a "A0" stepping.

Related Story AMD Pushes Out Zen 5 Patches At Linux, New CPU Models Added To Kernel

Based on what we know about AMD's Strix Point APU, the lineup will come in two flavors, a classic monolithic design featuring up to 12 CPU cores based on the Zen 5 core architecture, 16 RDNA 3+ iGPU cores, and an XDNA 2 "Ryzen AI" NPU. Then there's the Strix Point "Halo" lineup which will be the chiplet offerings with up to 16 CPU cores based on the Zen 5 core architecture, up to 40 RDNA 3+ iGPU cores and a much faster NPU rated at 3x the AI TOPs of Hawk Point. This is the same Halo SK that's been listed in the shipping manifesto.

Following are some of the main features to expect from Strix Point (1) APUs:

• Zen 5 (4nm) Chiplet Design
• Up To 16 Cores
• 40 RDNA 3+ Compute Units
• XDNA 2 Engine Integrated
• 2H 2024 Launch (Expected)

Besides the Strix Point Halo APU, another standard Strix Point APU has also been spotted within the GravityMark GPU benchmark database. This is the standard monolithic chip known as Strix Point (1). This CPU has a very similar OPN ID as the one that leaked out a few hours ago with it being "AMD Eng Sample: 100-000000994-19_Y". Based on what we can tell, this is a 12-core & 24-thread chip that features 24 MB of L3 cache and should also feature the highest-end iGPU config of up to 16 RDNA 3+ compute units.

Now this is a GPU specific