I'm sorry but I'd score this system MUCH higher (%85-90) than NBC did (%76).
- The lack of a NPU mostly kills WIN11 "AI" nonsense that nobody wants or likes
- The APU MAY not be the latest model, however performs very well
- The 680M iGPU may fall below a 780M iGPU, however only an 'issue' if playing games without a eGPU (OCuLink/USB4).
- Oh, did we mention it HAS an OCuLink port? A very desirable option
So as far as price/performance go, how much would one save by purchasing THIS system over a higher cost system with a newer APU? Seriously?
In a world where many content creators are whining about the price of PC parts, here's a complete system (no monitor/KB/Mouse) that is expandable for fairly reasonable $$$.
QuoteQuad-Channel
It's classical dual-channel, or 128-bit (2 * 64-bit per (SO)DIMM channel), to be precise. Can you stop with the misinformation? (Strix Halo would be quad-channel / 256-bit, if measured using the (SO)DIMM' 64-bit per channel).
Theoretical maximum: 76800 MB/s = 4800 MT/s * 128-bit / 1000 / 8.
Practical / what benchmarks measure: Theoretical maximum * 0.65 to 0.85.
Btw, AIDA64 Memory Read of 45811 MB/s is unusually slow for dual-channel.
If it only had 12 GB more RAM, then it could fit gpt-oss-120b or a glm-4.5-air quant with some context and 96 GB RAM could fit almost full context.