Calculating bandwidth...
Select a PCIe version and lane count above to calculate bandwidth.
PCIe Bandwidth by Version and Lane Count
The table below shows theoretical unidirectional bandwidth for every PCIe generation across common lane configurations. Real-world throughput is roughly 80% of the theoretical figure due to protocol overhead, encoding inefficiency, and retransmission handling. PCIe is full-duplex, so bidirectional totals are double the values shown.
| Version | ×1 | ×4 | ×8 | ×16 | Typical Use |
|---|---|---|---|---|---|
| PCIe 2.0 (2007) | 0.5 GB/s | 2.0 GB/s | 4.0 GB/s | 8.0 GB/s | Legacy platforms, HDDs, 1 GbE NICs |
| PCIe 3.0 (2010) | ~1 GB/s | 3.9 GB/s | 7.9 GB/s | 15.8 GB/s | GPUs, PCIe 3.0 NVMe, 10 GbE NICs |
| PCIe 4.0 (2017) | ~2 GB/s | 7.9 GB/s | 15.8 GB/s | 31.5 GB/s | Modern GPUs, PCIe 4.0 NVMe, 25 GbE NICs |
| PCIe 5.0 (2019) | ~3.9 GB/s | 15.8 GB/s | 31.5 GB/s | 63.0 GB/s | PCIe 5.0 NVMe drives, server GPUs, 100 GbE |
| PCIe 6.0 (2022) | ~7.9 GB/s | 31.5 GB/s | 63.0 GB/s | 126 GB/s | Data center, AI accelerators, future NVMe |
Does PCIe 4.0 vs 5.0 Matter for Gaming?
For gaming GPUs, the short answer is no. PCIe 4.0 ×16 delivers approximately 25 GB/s of real-world bandwidth in one direction. Even the most demanding current GPUs — the RTX 5090 and RX 9070 XT — use at most 15–20 GB/s during compute-heavy workloads, and considerably less during typical 1080p, 1440p, or 4K gaming. PCIe 5.0 adds zero measured FPS improvement for gaming GPUs.
PCIe 5.0 matters for NVMe storage. The fastest PCIe 5.0 SSDs — including the Crucial T705 and Seagate FireCuda 540 — deliver 12–14 GB/s sequential reads. A PCIe 4.0 ×4 slot peaks at ~7.9 GB/s theoretical (6.3 GB/s real-world), which is sufficient for current PCIe 4.0 drives but would throttle a PCIe 5.0 SSD by roughly 40–50%. If you bought a PCIe 5.0 NVMe drive, you need a PCIe 5.0 ×4 M.2 slot to unlock its full speed.
The practical upgrade path for most gamers: if your platform supports PCIe 4.0, you already have more than enough GPU bandwidth. Focus your PCIe 5.0 budget on a fast NVMe SSD rather than chasing marginally higher platform PCIe specs.
PCIe ×8 vs ×16 for GPU — What Do Benchmarks Show?
Running a GPU at ×8 instead of ×16 cuts the available PCIe bandwidth in half. On a PCIe 4.0 platform, this takes you from ~25 GB/s to ~12.6 GB/s real-world — still well above what any current gaming GPU needs. Real-world gaming benchmarks consistently show less than 1–2% FPS difference between PCIe 4.0 ×16 and ×8 across all GPU tiers.
On PCIe 3.0, ×8 gives ~6.3 GB/s real-world. High-end GPUs like the RTX 4090 can occasionally approach that ceiling in CPU-bound or VRAM-streaming scenarios, resulting in up to 3–5% performance loss. Mid-range cards are unaffected. PCIe 3.0 ×4 (~3.15 GB/s) is where performance loss becomes more noticeable — expect 5–10% lower FPS with a top-tier GPU in bandwidth-sensitive scenarios.
Common ×8 scenarios: PCIe risers in SFF builds, budget B650 motherboards that share lanes between GPU and M.2 slots, and multi-GPU setups on older platforms. for most gamers on PCIe 4.0, ×8 is perfectly fine.
| Configuration | Real-World BW | Est. GPU Impact |
|---|---|---|
| PCIe 4.0 ×16 | ~25.2 GB/s | Baseline |
| PCIe 4.0 ×8 | ~12.6 GB/s | <1% loss |
| PCIe 3.0 ×16 | ~12.6 GB/s | <1% loss |
| PCIe 3.0 ×8 | ~6.3 GB/s | 1–3% loss |
| PCIe 3.0 ×4 | ~3.2 GB/s | 3–8% loss |
| PCIe 2.0 ×16 | ~6.4 GB/s | 1–5% loss |
NVMe SSD Bandwidth by PCIe Generation
NVMe SSDs connect via PCIe ×4 slots (4 lanes). Choosing the right slot generation determines the storage ceiling for your drive. The table below shows what each slot can deliver, and where common SSDs fall.
| Slot | Theoretical | Real-World | Example Drives |
|---|---|---|---|
| PCIe 3.0 ×4 | 3.94 GB/s | ~3.1 GB/s | Samsung 970 Evo (3.5 GB/s), WD Black SN750 (3.4 GB/s) |
| PCIe 4.0 ×4 | 7.88 GB/s | ~6.3 GB/s | Samsung 990 Pro (7.45 GB/s), WD Black SN850X (7.3 GB/s) |
| PCIe 5.0 ×4 | 15.75 GB/s | ~12.6 GB/s | Crucial T705 (14.5 GB/s), Seagate FireCuda 540 (10 GB/s) |
Installing a PCIe 4.0 SSD in a PCIe 3.0 slot will throttle its sequential reads from ~7.4 GB/s to ~3.1 GB/s. Random 4K performance and latency are largely unaffected by the slot generation.
PCIe Encoding Overhead Explained
PCIe does not transmit raw data at the full signaling rate. Each generation uses a line encoding scheme that wraps data bits in overhead bits for synchronisation and error detection. This encoding overhead reduces the usable bandwidth below the theoretical maximum.
| Version | Encoding | Overhead | Efficiency |
|---|---|---|---|
| PCIe 2.0 / 3.0 | 8b/10b | 20% | 80% |
| PCIe 4.0 / 5.0 | 128b/130b | ~1.5% | 98.5% |
| PCIe 6.0 | PAM4 / FLIT | ~1.5% | 98.5% |
This calculator uses 80% as the real-world efficiency factor, which accounts for encoding overhead, protocol framing, TLP headers, and real-world traffic patterns — not just the encoding loss alone.
Frequently Asked Questions
Does PCIe 4.0 vs 5.0 make a difference for gaming?
No measurable difference for GPU gaming. PCIe 4.0 ×16 provides ~25 GB/s real-world bandwidth, far exceeding the 8–15 GB/s that even the most powerful gaming GPUs ever use. PCIe 5.0 provides value only for PCIe 5.0 NVMe SSDs, where it unlocks speeds beyond 10 GB/s.
Will running my GPU at ×8 bottleneck my RTX 5090?
On PCIe 4.0 or 5.0, no. PCIe 4.0 ×8 delivers ~12.6 GB/s real-world, which is more than the RTX 5090 needs for gaming. Benchmarks show <2% FPS difference between ×8 and ×16 on PCIe 4.0. On PCIe 3.0 ×8 (~6.3 GB/s), a very small impact (1–3%) may appear in bandwidth-heavy scenarios.
How much PCIe bandwidth does an NVMe SSD need?
PCIe 3.0 SSDs need ~3.5 GB/s, fitting in a PCIe 3.0 ×4 slot. PCIe 4.0 SSDs need up to ~7.5 GB/s, requiring a PCIe 4.0 ×4 slot for full speed. PCIe 5.0 SSDs need 10–14 GB/s and must be in a PCIe 5.0 ×4 slot. Putting a faster SSD in a slower slot will cap sequential speeds at the slot's bandwidth ceiling.
What is the difference between PCIe lanes and PCIe version?
PCIe lanes (×1, ×4, ×8, ×16) are the number of parallel data paths between the CPU and device. More lanes = more bandwidth. PCIe version (3.0, 4.0, 5.0) is the speed per lane — each generation roughly doubles the per-lane bandwidth of the previous one. A PCIe 4.0 ×8 slot provides the same bandwidth as a PCIe 3.0 ×16 slot.
Does PCIe version matter for a network card?
Almost never for consumer NICs. A 2.5 GbE NIC needs ~0.31 GB/s, a 10 GbE NIC needs ~1.25 GB/s — both fit comfortably on PCIe 3.0 ×1 (~0.79 GB/s real-world) or ×4. Only 100 GbE cards (~12.5 GB/s) would require PCIe 4.0 ×4 or higher to avoid throttling.
Is PCIe 6.0 worth waiting for?
PCIe 6.0 doubles per-lane bandwidth again to ~7.9 GB/s per lane (×16 = 126 GB/s), using PAM4 signaling. for gaming and consumer workloads, it offers no practical benefit today — even PCIe 5.0 ×16 is massively underutilised by GPUs. PCIe 6.0 targets AI accelerators and data-center fabric where 100+ GB/s bandwidth is genuinely needed.
Related Calculators
CPU / GPU Bottleneck Calculator
Find which component limits your build
GPU FPS Estimator
Estimate gaming performance for any GPU and game
VRAM Requirement Calculator
How much GPU memory do you need for your workload?
GPU Price-to-Performance
Find the best bang-for-buck GPU for your budget
NVMe vs SATA Load Time
See how much faster NVMe loads games vs SATA SSD
PC Build Budget Planner
Plan a balanced build across all components