Ryzen Z1's Tiny iGPU

Original Article ↗ Hacker News Discussion ↗

Enthusiasts are questioning why AMD, despite powering consoles like the PS5 with capable system-on-chips, offers relatively weak integrated GPUs in its desktop Ryzen lineup. Commenters point to technical limits such as AM5’s narrow memory bus, the high cost and packaging complexity of on-package or HBM memory, and deliberate product segmentation to protect discrete GPU sales and ecosystem partners. The Ryzen Z1 and Z1 Extreme handheld chips exemplify these trade‑offs, sparking debate over whether future designs like AMD’s rumored Strix Halo—featuring wider memory interfaces and stronger iGPUs—can deliver console-class performance in PCs without undermining flexibility, upgradability, and cost.

Why AMD Doesn’t Push Powerful Desktop iGPUs

  • Many wonder why AMD doesn’t ship desktop CPUs with console-class iGPUs.
  • Explanations offered:
    • Foundry capacity and margins: prioritize higher-margin, specialized SKUs.
    • Memory bandwidth: AM5 is limited to dual-channel (128‑bit), capping practical iGPU performance until a new socket (AM6) with more channels.
    • Thermal and board complexity: more GPU on-die increases cooling and motherboard cost.
    • Product segmentation and fear of cannibalizing discrete GPU and board-partner sales.

Memory Bandwidth, On‑Package RAM, and HBM

  • Shared system DDR5 is seen as the primary bottleneck; dual‑channel DDR5 still far below discrete GPU bandwidth.
  • Some argue you “just need” more channels (3–4 or 256‑bit/quad‑channel designs); Strix Halo rumors fit this.
  • Others propose on‑package LPDDR/HBM/GDDR as Apple‑style wide buses:
    • Pros: huge bandwidth, better efficiency, console‑like performance.
    • Cons: high cost, complex packaging (interposers, larger IO dies), yield issues.
  • Hybrid designs (fast on‑package + slower DIMMs) are debated:
    • Critics: two-speed memory is complex, may need separate controllers, unclear OS support.
    • Supporters: GPUs already juggle fast VRAM vs slower system RAM; could expose fast pool as “VRAM” and slow pool as main RAM.

Market Demand, Modularity, and Upgradeability

  • One view: buyers want separate CPU and GPU to mix/upgrade on different cycles; tying them together complicates roadmaps.
  • Counterpoints:
    • An iGPU doesn’t prevent adding a discrete GPU later.
    • There’s perceived demand for:
      • Console-like, mini‑ITX boxes with strong iGPUs.
      • Quiet, low‑power PCs that still handle esports/medium‑settings gaming.
    • Some think a powerful APU could be cost‑competitive if it truly replaces both CPU and midrange GPU; others say if it’s ~95% of the cost of separate parts, it loses the price advantage.

Z1 / 7840U / Handhelds and APUs

  • Z1 (non‑Extreme) is widely viewed as underpowered and poorly positioned on price; Z1 Extreme/7840U considered the “real” part.
  • Z1 line is seen as branding over existing Phoenix dies, with differentiation by firmware, power limits, and TDP tuning rather than distinct silicon.
  • Debate on binning vs pure power/cooling constraints; TDP figures are called misleading since many chips draw far above their nominal ratings under load or overclock.

iGPU Performance, Use Cases, and Future Direction

  • APUs are criticized for over‑emphasizing CPU cores vs GPU; some want 4–6 CPU cores and a much larger iGPU.
  • Others doubt a big market: “serious” gamers still prefer discrete GPUs; ultra‑strong iGPUs would remain bandwidth‑limited.
  • Handheld PCs and thin‑and‑light laptops with 7840U‑class iGPUs demonstrate strong demand for “good enough” mobile gaming, even when a user owns a more powerful desktop.
  • Some predict discrete GPUs may become less central as wide‑memory APUs (e.g., Strix Halo‑like designs) mature; others think bandwidth and thermal realities will keep dGPUs relevant.