AMD Radeon Vega 7/8 (Ryzen 5000 mobile APU, dual-channel DDR4-3200, 64GB)
Reference: List of AMD Ryzen processors on Wikipedia
- Class
- APU
- Memory
- 64 GB DDR4-3200
- Runs (Q4_K_M)
- —
- Bandwidth
- 51.2 GB/s
- TDP
- 35 W
- Released
- 2021-01-12
Sourcing and disambiguation notes
Keyed by chip class plus memory configuration, matching ryzen-ai-max-395-128gb and radeon-780m-ddr5-5600. Covers the Zen 3 mobile APUs -- Cezanne and Lucienne, Ryzen 5 5600H through Ryzen 7 5800HS -- carrying a Radeon Vega 7 or Vega 8 iGPU on dual-channel DDR4-3200. 51.2 GB/s is arithmetic from that memory: 3200 MT/s across a 128-bit dual-channel bus, never inferred from a measurement. The Vega 7 and Vega 8 differ by one compute unit, which is a real difference and is why this entry is worth questioning -- but at this bandwidth both parts are memory-starved rather than compute-starved, and the two authors behind its benchmark record agree to 1.7% on PREFILL, the metric a CU difference would show up in first. That is the empirical case for one entry rather than two. TDP is recorded as 35W because both authors state a 35W package cap; these chips are specified 35-54W and a machine at the top of that range may prompt-process faster. This is the dataset's LOWEST-bandwidth APU by a wide margin -- the others are 89.6 and 256 GB/s -- which is the point of carrying it. Memory capacity is in the id because it decides which models fit; these APUs commonly ship at 16-64GB and the record's authors ran 64.
Specification sources
- https://www.amd.com/en/products/processors/laptop/ryzen/5000-series/amd-ryzen-5-5600h.html — vendor, 2026-08-26
Measurements
Decode is token generation — the speed you feel while an answer streams. Prefill is prompt processing — the wait before it starts. Why bandwidth predicts decode speed.
No records yet for this hardware. Know of a published benchmark on it? Submit the link.
What verified, single-source and estimated mean, and the same rows with every filter and sort in the benchmarks explorer.
Speed over time
What it can run
Fit is arithmetic, not a measurement — how it is computed. Against 64 GB; models that fit are listed largest first.
No modeled quant fits in 64 GB.