AMD Radeon 780M (Ryzen 7 7840U / 7840HS, dual-channel DDR5-5600)
Reference: AMD Radeon 700M series on Wikipedia
- Class
- APU
- Memory
- 32 GB DDR5-5600
- Runs (Q4_K_M)
- —
- Bandwidth
- 89.6 GB/s
- Released
- 2023-01-04
Sourcing and disambiguation notes
Keyed by chip plus memory configuration rather than by product, matching ryzen-ai-max-395-128gb and core-ultra-7-258v. It covers the Ryzen 7 7840U and 7840HS, which AMD gives identical memory support (DDR5-5600 or LPDDR5x-7500) and the same Radeon 780M with 12 compute units; they differ only in TDP, 28W against 35-54W. Decode is bandwidth-bound, so that difference does not reach the tokens-per-second figure, though it does affect prompt processing. 89.6 GB/s is arithmetic from the stated memory -- 5600 MT/s across a 128-bit dual-channel bus -- not inferred from any measurement. THE MEMORY MUST BE STATED BY THE AUTHOR for a run to belong here: the same silicon on LPDDR5x-7500 is 120 GB/s, a third faster, and several otherwise-usable runs on sibling AMD APUs have been refused for exactly this reason. Form factor is recorded as mini-pc because that is the commonest enclosure for this chip, but the entry also covers laptops such as the Framework 13; a run's chassis is noted on the record rather than here. No price: this chip ships in many machines at very different prices and no single representative listing was verified.
Specification sources
- https://www.amd.com/en/products/processors/laptop/ryzen/7000-series/amd-ryzen-7-7840u.html — vendor, 2026-08-26
- https://www.amd.com/en/products/processors/laptop/ryzen/7000-series/amd-ryzen-7-7840hs.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 32 GB; models that fit are listed largest first.
No modeled quant fits in 32 GB.