AMD’s low-power Van Gogh APUs have once again been spotted in the Linux Kernel. The SKU in question is a 4-core chip with SMT bringing the total thread count up to 8. We’re looking at a single CCX design based on TSMC’s 7nm process, running at a base clock of 2.4GHz and a boost of 3.4GHz. On the GPU side, we’ve got an RDNA 2 graphics core with a feature level on par with the Navi 21 “Big Navi” GPUs.
We already know that VG will be compatible with LPDDR5 memory along with LPDDR4X. This particular SKU is paired with a 256-bit bus which is likely an error. Remember that LPDDR5 (like LPDDR4 and 4X) will feature two memory channels, each being 16-bits with a wider 16n/6 burst length/prefetch, so 256-bit is highly improbable. We’re likely looking at a 64-bit configuration featuring two DIMMs and four overall channels.
|CPU Arch||Zen 2|
|CPU Core / Thread||4/8|
|CPU Base Clock||1.4GHz|
|GPU Arch||RDNA 2 (gfx1033)|
|Media Engine||VCN 3.0|
AMD’s Van Gogh should be highly scalable with its applications ranging from ultra-low power notebooks to embedded and possibly even edge AI. The fact that the integrated Navi 2 GPU supports mixed and low-precision compute seems to support that notion. Van Gogh is likely to succeed the Ryzen Embedded R1000 & R2000 lineups “Raven 2” which also include a quad-core Zen+ based APU with a 3CU Vega GPU, fabbed on the older 14nm node.