The Supernode Revolution - Knowledge Comic Storyboard
Character Reference: characters/characters.png
Cover
Filename: 00-cover-supernode.png Core Message: The future of AI computing is built on supernodes
Visual Design:
Title: "THE SUPERNODE REVOLUTION" in bold retro typography
Main visual: Massive AL128 rack in art deco industrial style, glowing with data streams
Color scheme: Deep teal, warm orange, copper accents, cream background
Subtitle: "Inside Alibaba Cloud's Panjiu AL128 AI Infrastructure"
Visual Prompt: Retro 1960s-70s technical illustration style cover. Bold sans-serif title "THE SUPERNODE REVOLUTION" at top. Center: massive server rack rendered as art deco industrial marvel, 128 illuminated GPU panels arranged in grid, copper pipes and brass fittings, glowing orange and teal data streams flowing between nodes. Dr. Lin (professional male, 50s, salt-pepper hair, thick retro glasses, white lab coat) standing beside rack for scale. Background: aged cream paper texture with faint blueprint grid. Subtitle at bottom. Warm orange (#FF6B35), deep teal (#006666), copper (#B87333) color palette. Vintage aged paper texture throughout.
Page 1 / 6
Filename: 01-page-problem.png Layout: standard Core Message: AI models are growing faster than traditional servers can handle
Panel Layout
Panel Count: 3 Layout Type: vertical stack
Panel 1 (Size: 1/3 page, Position: Top)
Scene: Abstract visualization of growing AI models Image Description:
Camera angle: Wide shot
Visual: Mountain graph showing exponential growth - model sizes from millions to trillions of parameters
Style: Retro infographic with vintage data visualization
Text overlay: "Model Parameters: 1B → 10T+"
Panel 2 (Size: 1/3 page, Position: Middle)
Scene: Traditional server struggling Image Description:
Camera angle: Side view
Visual: Old-style server rack looking small and overwhelmed, heat lines, warning lights
Character: Traditional CPU-centric architecture shown as dated 1970s mainframe
Color tone: Muted, struggling
Panel 3 (Size: 1/3 page, Position: Bottom)
Scene: Dr. Lin introduces the solution Image Description:
Camera angle: Medium shot
Dr. Lin standing confidently, blueprints in hand
Speech bubble: "The answer? 128 GPUs working as one supernode."
Background: Hint of AL128 rack silhouette
Visual Prompt: Retro technical illustration comic page, 3 vertical panels. PANEL 1: Vintage infographic showing exponential growth curve, "MODEL SIZE" label, numbers climbing from millions to trillions, retro graph style with grid lines. PANEL 2: Old 1970s-style mainframe computer looking small and inadequate, heat waves rising, warning lights blinking, vintage tech aesthetic. PANEL 3: Dr. Lin (50s male, salt-pepper hair, retro glasses, white lab coat, teal turtleneck) standing confidently holding blueprints, speech bubble "The answer? 128 GPUs working as one supernode", silhouette of massive server rack behind him. Aged cream paper texture, vintage color palette: teal (#006666), orange (#FF6B35), copper accents.
Page 2 / 6
Filename: 02-page-supernode.png Layout: standard Core Message: Meet the AL128 - 128 GPUs in one rack
Panel Layout
Panel Count: 3
Panel 1 (Size: 1/2 page, Position: Top - Splash)
Scene: Full reveal of AL128 supernode Image Description:
Camera angle: Hero shot, slight low angle
Visual: Massive AL128 rack in full glory, art deco industrial design
Details: 128 illuminated GPU panels, copper cooling pipes, orthogonal interconnect architecture
Lighting: Dramatic, GPUs glowing orange and teal
Label callouts: "128 GPUs", "350kW Power", "500kW Cooling"
Panel 2 (Size: 1/4 page, Position: Bottom Left)
Scene: Modular architecture diagram Image Description:
Cutaway view showing decoupled components
GPU nodes on left, CPU nodes on right, power separate
Vintage blueprint style with labels
Panel 3 (Size: 1/4 page, Position: Bottom Right)
Scene: Flexible configuration Image Description:
Visual: Different GPU-to-CPU ratios shown (8, 16, 32 per OS)
Retro dial/gauge aesthetic showing flexibility
Visual Prompt: Retro technical illustration comic page. TOP HALF (splash panel): Massive AL128 supernode rack, art deco industrial design, 128 illuminated GPU panels in grid pattern, copper pipes for cooling, brass fittings, orthogonal architecture visible. Hero shot low angle. Orange and teal glowing nodes. Callout labels: "128 GPUs", "350kW", "2kW liquid cooling per GPU". BOTTOM LEFT: Blueprint-style cutaway showing modular design - GPU nodes separate from CPU nodes separate from power nodes, vintage technical drawing style. BOTTOM RIGHT: Retro gauge/dial showing flexibility "8/16/32 GPUs per OS" configurations. Aged paper texture, teal and orange color scheme, copper metallic accents.
Page 3 / 6
Filename: 03-page-scaleup.png Layout: standard Core Message: Layer 1 - ScaleUp interconnection within supernode
Panel Layout
Panel Count: 3
Panel 1 (Size: 1/3 page, Position: Top)
Scene: Visualization of single-stage switching Image Description:
Camera angle: Isometric view
Visual: 64 GPUs connected to 8 ALink switches in single stage
Style: Retro circuit diagram, glowing connection lines
Data: "14-28 Tbit/s per GPU"
Panel 2 (Size: 1/3 page, Position: Middle)
Scene: ALink protocol advantage Image Description:
Comparison: Ethernet vs ALink
ALink shown as sleek lightning bolts, Ethernet as slower waves
Speed indicator: "Ultra-low latency"
Panel 3 (Size: 1/3 page, Position: Bottom)
Scene: Dr. Lin explaining the key insight Image Description:
Dr. Lin pointing at holographic topology diagram
Speech bubble: "Single-stage switching means instant communication between any two GPUs."
Emphasis on "non-Ethernet" advantage
Visual Prompt: Retro technical illustration comic page. PANEL 1: Isometric view of ScaleUp topology - 64 GPU nodes (as glowing orange cubes) connected to 8 ALink switch nodes (as brass junction boxes) via single-stage mesh, all connections glowing teal. Label: "14-28 Tbit/s per GPU". Vintage circuit diagram style. PANEL 2: Side-by-side comparison - ALink protocol as sleek lightning bolts vs Ethernet as slower wave patterns. ALink clearly faster. "ULTRA-LOW LATENCY" label. PANEL 3: Dr. Lin gesturing at holographic topology diagram floating above his hand, speech bubble "Single-stage switching = instant GPU-to-GPU communication". Retro hologram effect with scan lines. Aged paper texture, teal/orange/copper palette.
Page 4 / 6
Filename: 04-page-layers.png Layout: standard Core Message: Three-layer interconnect architecture
Panel Layout
Panel Count: 3
Panel 1 (Size: 1/3 page, Position: Top)
Scene: Layer 2 - ScaleOut between supernodes Image Description:
Visual: Multiple supernode racks connected via high-speed NICs
Style: City skyline metaphor - supernodes as buildings, connections as highways
Data: "400-800 Gbit/s per GPU to other supernodes"
Panel 2 (Size: 1/3 page, Position: Middle)
Scene: Layer 3 - Data Center Network Image Description:
Supernodes connected to cloud storage, databases, internet
Visual: Retro telephone switchboard aesthetic for DCN
Labels for different data types: storage, KV cache, user data
Panel 3 (Size: 1/3 page, Position: Bottom)
Scene: Complete three-layer diagram Image Description:
Elegant summary showing all three layers as nested circles
ScaleUp (inner, fastest), ScaleOut (middle), DCN (outer)
Vintage Venn diagram style with glow effects
Visual Prompt: Retro technical illustration comic page. PANEL 1: ScaleOut layer visualization - multiple AL128 supernode racks as art deco towers connected by glowing highway lines, "400-800 Gbit/s" speed labels, city skyline metaphor. PANEL 2: DCN layer - supernodes connected to cloud icons, database symbols, internet globe via retro telephone switchboard-style connection panel. Labels: "Storage", "KV Cache", "User Data". Vintage telecom aesthetic. PANEL 3: Complete three-layer summary as nested glowing circles - innermost "ScaleUp" (teal), middle "ScaleOut" (orange), outer "DCN" (copper). Bandwidth values labeled on each ring. Vintage Venn diagram style. Aged cream paper texture throughout.
Page 5 / 6
Filename: 05-page-future.png Layout: standard Core Message: GPU-centric architecture and future directions
Panel Layout
Panel Count: 4
Panel 1 (Size: 1/4 page, Position: Top Left)
Scene: Direction 1 - Break PCIe limits Image Description:
Visual: PCIe barrier shattering, ScaleUp connections breaking through
Metaphor: Chain breaking, freedom realized
Label: "Beyond PCIe"
Panel 2 (Size: 1/4 page, Position: Top Right)
Scene: Direction 2 - Memory semantics Image Description:
GPU accessing memory directly, simplified path
Vintage computer memory diagram with modern twist
Label: "Native GPU Memory Access"
Panel 3 (Size: 1/4 page, Position: Bottom Left)
Scene: Direction 3 - Simplified 2-layer topology Image Description:
3-layer diagram transforming to 2-layer
ScaleUp + High-bandwidth DCN merger
Arrow showing simplification
Panel 4 (Size: 1/4 page, Position: Bottom Right)
Scene: Direction 4 - Optical interconnection Image Description:
Light beams replacing copper cables
Retro-futuristic laser aesthetic
Label: "Light-Speed Future"
Visual Prompt: Retro technical illustration comic page, 4-panel grid. TOP LEFT: PCIe barrier visualized as chain/wall shattering, ScaleUp protocol breaking through as lightning, "BEYOND PCIE" label, liberation metaphor. TOP RIGHT: GPU chip directly accessing memory pool via simple path, old complex paths faded out, "NATIVE GPU MEMORY ACCESS" label, simplified diagram. BOTTOM LEFT: 3-layer diagram (ScaleUp/ScaleOut/DCN) transforming with arrow to 2-layer (ScaleUp + DCN), simplification visual. BOTTOM RIGHT: Optical interconnection - ethereal light beams replacing copper cables, retro-futuristic laser aesthetic, "LIGHT-SPEED FUTURE" label with 256 GPU projection. All panels: vintage technical illustration style, teal/orange/copper palette, aged paper texture.
Page 6 / 6
Filename: 06-page-conclusion.png Layout: standard Core Message: Supernodes are the future of AI computing
Panel Layout
Panel Count: 2
Panel 1 (Size: 2/3 page, Position: Top - Large)
Scene: Vision of the future data center Image Description:
Camera angle: Wide establishing shot
Visual: Data center filled with glowing AL128 supernodes
Style: Retro-futuristic utopia, clean, powerful
CXL memory pools connecting everything
Light beams and data streams flowing
Panel 2 (Size: 1/3 page, Position: Bottom)
Scene: Dr. Lin's closing statement Image Description:
Dr. Lin in confident pose, supernode behind him
Speech bubble: "128 GPUs per supernode. Ultra-low latency. The optimal balance for AI's next decade."
Final call to action: "Embracing open standards. Building the future."
Page Hook: "Learn more at Alibaba Cloud"
Visual Prompt: Retro technical illustration comic page, conclusion. TOP (2/3 page): Futuristic data center vista - rows of glowing AL128 supernode racks in art deco industrial style, CXL memory pools as shared glowing resources connecting racks, light beams representing optical interconnects crisscrossing overhead, retro-futuristic utopia aesthetic. Warm orange and teal glows. BOTTOM (1/3 page): Dr. Lin in heroic confident pose, AL128 supernode behind him glowing, speech bubble: "128 GPUs per supernode. Ultra-low latency. The optimal balance for AI's next decade." Smaller text: "Embracing open standards. Building the future." Alibaba Cloud logo subtly integrated. Vintage color palette, aged paper texture, triumphant composition.
Last updated