All PCI Express (PCIe) types explained

PCI Express (PCIe) is the high-speed interface that connects hardware directly to your CPU—including graphics cards, NVMe SSDs, network cards, and sound cards.

The core philosophy of PCIe is simple: every new generation doubles the speed (bandwidth) of the previous one while maintaining 100% backward compatibility.

1. PCIe Speeds Across Generations

PCIe bandwidth depends on generation and lane width (written as x1, x4, x8, or x16). Below are the theoretical maximum speeds (one direction) for every major PCIe generation.

Generation	Launch	Raw Data Rate	Encoding	x1 (1 Lane)	x4 (M.2 NVMe SSDs)	x16 (GPUs)
PCIe 1.0	2003	2.5 GT/s	8b/10b	250 MB/s	1 GB/s	4 GB/s
PCIe 2.0	2007	5.0 GT/s	8b/10b	500 MB/s	2 GB/s	8 GB/s
PCIe 3.0	2010	8.0 GT/s	128b/130b	~1 GB/s	~4 GB/s	~16 GB/s
PCIe 4.0	2017	16.0 GT/s	128b/130b	~2 GB/s	~8 GB/s	~32 GB/s
PCIe 5.0	2019	32.0 GT/s	128b/130b	~4 GB/s	~16 GB/s	~64 GB/s
PCIe 6.0	2022	64.0 GT/s	PAM4 + FLIT	~8 GB/s	~32 GB/s	~128 GB/s
PCIe 7.0	2025	128.0 GT/s	PAM4 + FLIT	~16 GB/s	~64 GB/s	~256 GB/s

(Note: Real-world speeds are typically ~10–15% lower due to protocol overhead. Additionally, PCIe is full-duplex, meaning it can send and receive this amount of data simultaneously, which doubles the total bidirectional bandwidth.)

2. Key Concepts Explained

Lanes (x1, x4, x8, x16)

Think of PCIe as a highway system. A lane is a pair of point-to-point data traces (one for sending, one for receiving).

• x1: Small cards like Wi-Fi or sound cards.
• x4: M.2 NVMe SSDs.
• x8: High-speed network cards, RAID controllers, or secondary GPU slots.
• x16: Graphics cards (GPUs) use the full 16-lane physical slot for maximum bandwidth.

GT/s vs. GB/s

• GT/s (Gigatransfers per second): Measures raw signal rate—how many billion operations or signal transitions happen per second.
• GB/s (Gigabytes per second): Measures actual usable data throughput after factoring in line-encoding overhead.

Line Encoding & Architectural Shifts

How PCIe transmits bits changes across generations to squeeze more speed out of copper wires:

1. 8b/10b (Gen 1 & 2): For every 8 bits of actual data sent, 2 extra bits were added for clock sync and error management. This resulted in a 20% bandwidth penalty.
2. 128b/130b (Gen 3, 4, & 5): A much more efficient scheme where only 2 bits of overhead are added for every 128 bits of data (~1.5% penalty). This is why Gen 3 doubled the throughput of Gen 2 without doubling the GT/s clock rate!
3. PAM4 & FLIT (Gen 6 & 7): Moving past basic NRZ (1s and 0s), Gen 6/7 use Pulse Amplitude Modulation 4 to send 2 bits of data per clock cycle using 4 voltage states. Paired with Forward Error Correction (FEC) and fixed-size FLIT (Flow Control Units), it keeps bandwidth doubling without creating insane electrical noise.

Backward & Forward Compatibility

All PCIe slots and cards are cross-compatible.

• A PCIe 3.0 GPU plugged into a PCIe 5.0 slot works fine—it just operates at PCIe 3.0 speed.
• A PCIe 5.0 SSD plugged into a PCIe 4.0 slot works fine—it is simply capped at PCIe 4.0 speed.
• A smaller card (e.g., an x4 card) can physically plug into a larger slot (e.g., an x16 slot) and work without issue.

(Source: Google Gemini)