Silicon Interposer vs Organic Interposer vs Silicon Bridge: Wiring Density, Size and Cost

In 2.5D packaging, several chips such as a processor and HBM stacks sit side by side and are connected by wiring much finer than a package substrate can offer. There are three main ways to provide that connection. A silicon interposer is a slab of silicon with fine wiring made by the wafer interconnect process, placed under all the chips. An organic interposer is built from redistribution layers (RDL) of polymer dielectric and copper wiring. A silicon bridge is a small piece of silicon with fine wiring embedded only under the areas where chips meet. The silicon interposer wins on wiring density, while the organic interposer and silicon bridge win on size and cost. The choice depends on how many connections are needed and how large the package is.

AspectSilicon interposerOrganic (RDL) interposerSilicon bridge
How it is madeMultilayer fine wiring (copper damascene, as in chip interconnects) and TSVs are built on a silicon waferPolymer dielectric layers (such as polyimide) and plated copper wiring are stacked layer by layer on a carrierA small piece of silicon with fine wiring is embedded in the package substrate or in redistribution layers
Where silicon is placedOne large slab under all the chipsNo silicon; redistribution layers sit under all the chipsOnly under the adjacent edges of neighboring chips; other connections go through ordinary substrate wiring
Typical line width/spacingFinest, typically around 1 µm or belowCoarser than silicon, typically around 2 µm, with fewer layersAs fine as a silicon interposer in the bridge area
Size limitsGoing beyond one exposure field (about 26 mm × 33 mm) requires stitching exposures together, and cost rises with sizeEasier to make large than a silicon interposerBridges themselves are small, so placing several of them makes packages larger than one exposure field easier
Need for TSVsRequired, to connect the chips on top through the interposer to the package substrateNot needed; vias in the redistribution layers make vertical connectionsIn most substrate-embedded designs the bridge itself needs no TSVs; power and many signals connect directly to the substrate outside the bridge
CostHigh, needing a large silicon area and TSV processingEasier to keep below a silicon interposerEasier to keep below a silicon interposer because the silicon area is small, though embedding and alignment steps are added
Thermal expansion and warpageSilicon expands much less with heat than organic substrates, and the resulting stress and warpage become harder to manage as size growsThermal expansion is close to the package substrate, which helps reduce stress at the joints, but warpage and dimensional change of the polymer layers must be controlledMost of the package is organic substrate, so overall warpage is close to an ordinary substrate; local stress around the bridge and height control of fine and coarse bumps are challenges
Electrical characteristicsWires are fine and short; fine wires have high resistance, so they are not suited to long runsThicker wires are easier, so resistance is lower, but wiring density is lowerDense wiring covers only the short span between chip edges; power can be fed directly from the substrate without passing through the bridge
Typical usesConnections that need very many wires, such as a processor with several HBM stacksChiplet connections where wiring demands are less severe or a large area is neededChiplet and HBM connections in large packages that need dense wiring only between chip edges

Figures are typical orders of magnitude and vary by process generation and product.

When to use which

Silicon interposer
Suits cases where wiring density comes first. It has been widely used for configurations that connect a processor and several HBM stacks with thousands of wires.
Organic (RDL) interposer
Suits cases that do not need silicon-level wiring density and put cost or size first. It also helps when stress at the joints from thermal expansion mismatch should be reduced.
Silicon bridge
Suits large packages that need dense wiring only at the boundaries between chips. It saves cost by reducing the silicon area.

FAQ

What happens when an interposer is larger than the exposure field?

The area one exposure shot can cover (the reticle size) is limited to about 26 mm × 33 mm. Silicon interposers larger than that are made with stitching, which joins the wiring precisely across the boundaries of neighboring shots. The larger the interposer, though, the harder cost, yield and warpage become to manage, so organic interposers and silicon bridges are also considered.

What is the main difference between a silicon bridge and a silicon interposer?

Where the silicon goes. A silicon interposer places a large slab of silicon under all the chips, while a silicon bridge places a small piece of silicon only under the areas where chips meet. Using silicon only where dense wiring is needed makes it easier to save area and cost.

How different is the wiring pitch?

As a rough guide, line width and spacing are around 1 µm or below on silicon interposers and around 2 µm on organic interposers, and ordinary package substrates are coarser still. The numbers change with generation and approach, but silicon can go finer because it uses the chip interconnect process.

Can organic interposers and silicon bridges be combined?

Yes. Some designs embed small pieces of silicon in the redistribution layers and use silicon wiring only where high density is needed. Vendors often market these approaches under their own names, so instead of relying on the name, check where silicon is used and whether TSVs are present.

Check your understanding

Which correctly describes the role of the silicon interposer used in 2.5D packaging?

  1. AA heat sink for cooling chips
  2. BAn intermediate substrate on which several chips are placed side by side and connected at high density with fine wiring
  3. CA photomask used for exposure
Show answer and explanation

Answer: B. An intermediate substrate on which several chips are placed side by side and connected at high density with fine wiring

A silicon interposer is a silicon intermediary substrate on which wiring of 1 µm or finer is formed using semiconductor back-end processes. Multiple chips, such as a processor and HBM, sit side by side on it and are connected with wiring far denser than ordinary organic package substrates, whose lines are tens of micrometers wide, can provide. The interposer connects to the package substrate through TSVs. It is unrelated to cooling parts or photomasks.

Which approach, instead of a large silicon interposer, embeds small pieces of silicon only at the chip-to-chip connection areas to achieve high-density wiring?

  1. AWire bonding approach
  2. BSilicon bridge approach
  3. CLead frame approach
Show answer and explanation

Answer: B. Silicon bridge approach

The silicon bridge approach embeds small silicon pieces in the package substrate covering only the connection regions between adjacent dies, forming fine wiring on them for high-density die-to-die links. Eliminating a large interposer under all the dies cuts cost and allows assemblies larger than the lithography field. Intel's EMIB is a representative example. Wire bonds or lead frames cannot reach this density.

In 2.5D packaging, what is the main advantage of using an interposer made of redistribution layers (RDL) of organic dielectric and copper wiring instead of a silicon interposer?

  1. AIt can always form finer wiring than silicon
  2. BTSVs become unnecessary and no heat dissipation is needed at all
  3. CIt keeps cost down more easily and makes large-area interposers easier to build
Show answer and explanation

Answer: C. It keeps cost down more easily and makes large-area interposers easier to build

Organic RDL interposers avoid silicon wafer processing, so they are cheaper, more compliant and easier to scale up in size. Their thermal expansion is closer to the package substrate's, which also reduces joint stress. Their wiring pitch and density usually fall short of silicon interposers, however, so they are chosen per application or combined with silicon bridges only where high density is needed.

  1. Which method is used to make silicon interposers larger than the maximum area a lithography tool can expose at once (about 26 mm × 33 mm)?
  2. In AI GPUs, what is the main reason the GPU chip and HBM are placed side by side on one silicon interposer?
  3. Which correctly describes the difference between 2.5D and 3D packaging?

Practice Advanced packaging (3D and chiplets) questions

Key terms explained

Silicon interposer
A silicon intermediate substrate that holds chips side by side and connects them densely with fine wiring. It is used to connect GPUs and HBM, among other things.
Redistribution layer (RDL)
A wiring layer formed on a chip or wafer surface to route pads out to the positions of the external terminals. It is used in fan-out packages and interposers.
TSV (through-silicon via)
An electrode that passes through the silicon substrate. It connects stacked chips vertically over a short distance.
2.5D packaging
Placing several chips side by side on an intermediate substrate such as a silicon interposer and connecting them with dense, fine wiring. It is used, for example, to connect GPUs and HBM.
HBM
High Bandwidth Memory. Multiple DRAM dies stacked with TSVs, achieving very high bandwidth through an extremely wide bus.
Chiplet
A small chip made for a specific function. Several chiplets are connected inside one package to build a larger system.
Microbump
A bump far smaller than an ordinary solder ball, used to connect TSV-stacked chips or chips to an interposer. The pitch is a few tens of micrometers.

Related comparisons